The truth about search engine advertising

1. 1Set the objective for each campaign before you split anything. The source says position, click volume, or CPC average/cap are valid objectives, but an efficient campaign also folds post-click behavior into the ongoing work. Write one line per campaign, such as 'hold CPA/ROAS at X' or 'win position Y', and refuse to segment without it.
2. 2Start from your converting keywords. Group them by business sector and intent so each group's ads and bids serve one kind of searcher. The source says to use every campaign permutation you can get 'while keeping your conversion high', so a split that cannot hold conversion does not belong in the structure.
3. 3Cut the converting set by geography and by day-part/day of the week. Look for 'clusters where conversion is already there'. These cuts are the cherry-picking technique the source names for finding lift when the whole campaign shows none.
4. 4Measure each segment on post-click results (conversion rate, order size, CPO/CPA, ROAS), not clicks alone. Bid management only buys clicks, so segment-level profit data has to come from your own reporting.
5. 5Set bids per segment from that profit data. Raising conversion rate or revenue per order is leverage for a higher position. The source's levers: lift conversion (3% to 3.96% is a 32% gain) or raise average order size, for example by moving the free-shipping threshold from $50 to $59 or $65.
6. 6If a lever fails across the whole clickstream on a high-potential keyword, run a segmentation analysis. The source says to do this for both conversion and revenue. Find the sub-segment where it works and bid up only there.
7. 7Expand to more keywords and segments in stages. Low-volume keywords add up, but diminishing marginal returns 'can creep up on you'. Because of inverse volume discounts, a larger budget buys progressively costlier clicks, so check each new segment against the profit target before the next one.
8. 8Watch for intent shifts and keep bids reactive. The source's example is 'Janet Jackson' during the wardrobe malfunction, when conversion rate and potentially Quality Score fell as the makeup of search intent changed. Review segments on a regular cadence and when news or search behavior shifts, and let the bidding engine react between reviews.

The source gives two numbers that show why structure matters. First, a $5,000 month-long test works: ROI, CPO/CPA and ROAS are in line. The team wants $20,000. Unless it sat at very low positions before, that inventory may not exist, and what does exist costs more per click because it displaces advertisers above it who earn the engine more revenue. Broadening to more keywords helps but is capped by the keyword list your business sector supports. Second, suppose a high-potential keyword converts at 3% and needs a 32% lift to reach the position you want. That means reaching 3.96%. If no whole-campaign change gets there, the source's instruction is to look for clusters where conversion is already higher by geo or day-part and raise bids only for those. Applied here: instead of pushing the whole campaign from $5,000 to $20,000 at rising CPCs, split the keyword's traffic by geo and day-part, then fund the segments already near 3.96% and hold the rest. A promotion change, such as moving free shipping from $50 to $59 or $65, can raise average order size and add bid headroom in the same segments. The segments, not the aggregate, decide where extra budget is profitable.

• If a high-potential keyword needs a conversion lift to win position and the whole campaign cannot deliver it, then look for geo or day-part clusters that already convert and bid up only there.
• If you cannot raise revenue across the entire clickstream on a keyword, then run a segmentation analysis for the sub-segments where order size or profit is higher, and test a promotion change such as a higher free-shipping threshold.
• If the budget is going up sharply (say $5,000 to $20,000), then expect higher per-click costs from inverse volume discounts. Add keywords and segments in stages and judge each against CPO/CPA or ROAS, not volume.
• If a new segment or low-volume keyword cannot hold conversion, then drop it from the structure. The source's test for a permutation is that it keeps conversion high.
• If a keyword's search intent shifts (news events, behavior change) and conversion drops, then rely on reactive bidding or manual review to cut or re-segment it, and do not wait for a monthly cycle.

• Segmenting to the point of noise. Low-volume keywords and thin segments add up, but diminishing marginal returns creep in. The tell: more segments, flat or falling blended ROAS, and bids set from too little data.
• Optimizing clicks instead of post-click outcomes. Bid tools buy clicks only, so segments get bid on click volume or position while conversion and profit are never reported at segment level. The tell: position goes up and CPO/CPA gets worse.
• Scaling budget linearly. Moving from $5,000 to $20,000 assumes the same inventory at the same CPC. The tell: costs per click rise and impressions or position stall as you displace higher-earning advertisers.
• Ignoring intent shifts within a segment. A news event changes who is searching, and conversion and Quality Score fall under unchanged bids, as with 'Janet Jackson'. The tell: a sudden conversion drop on a stable keyword with no change on your side.
• Treating the account as one pool. Hunting for conversion lift across the whole campaign finds nothing when the lift lives in particular geos or day-parts. The tell: whole-campaign tests come back flat while a few clusters quietly outperform.

• ▸Campaign objective: state position, click volume, CPC cap, or CPO/CPA/ROAS target in one line.
• ▸Keyword group / intent: name the group and the searcher intent it serves.
• ▸Segment cut: geo, day-part, day of week, or other, and the specific value.
• ▸Current conversion rate: the segment's post-click conversion rate.
• ▸Target conversion rate: the rate needed for the desired position (for example 3% to 3.96% is +32%).
• ▸Average order size / profit per order: current figure, plus any promotion change under test (for example free shipping at $59 or $65).
• ▸Segment CPO/CPA or ROAS vs. target: record the gap.
• ▸Bid action: raise, hold, or cut, with the reason.
• ▸Intent-shift check date / trigger: set the review date and the news or behavior event that forces an early review.

1. 1Segment the keyword list by buying-cycle stage before you set a single bid. Use the passage's linguistic logic: generic terms ("laser printer") are a mix of stages; modifiers like "review" or "compare" signal research; "best" signals pre-purchase evaluation; "cheap" signals near-purchase and price sensitivity; a specific model ("Hp 1200 laser printer") signals close to purchase.
2. 2Decide which stages you want to reach based on your role. A manufacturer or brander wants early-stage consumers; a retailer prefers those ready to buy or ready to register for a specials newsletter. Write this decision down as the campaign objective, because it determines which keyword groups get budget.
3. 3Write ad copy that answers the need the keyword expresses. For "review" and "compare" or "best" terms, lead with the benefits of a particular unit. For "cheap" terms, lead with price. For model-number terms, lead with the exact product and a purchase call to action. Treat the listing as the envelope cover and the landing page as the catalog page the psychic mailman hands over.
4. 4Respect engine format limits when you tailor the message. Google allows 25/35/35 characters for title, description line 1, and description line 2; Yahoo! allows 40/190. Use Yahoo!'s extra room for prequalification or a fuller offer explanation. In Google, keep copy compelling, because a poor CTR demotes or deactivates the listing.
5. 5Send each stage to a matching landing page, not the home page. A research-stage click lands on comparison or benefit content; a close-stage click lands on the specific product. The passage warns that ignoring buying phases produces an inefficient site and an inefficient campaign regardless of traffic source.
6. 6Value early-stage traffic quantitatively. Assign values to post-click actions (purchase, registration, newsletter subscription, information request, pageview, content depth, webinar view), set the most valuable action to 1, and weight the others relative to it. Use hard data (how many catalog requesters actually buy) or an intercept survey for lift in purchase intent or awareness.
7. 7Adjust CPC or position per keyword using that blended value. Each listing delivers a mix of immediate buyers, earlier-stage visitors, and some careless clickers, so bid on the traffic's true value rather than on last-click purchases only.
8. 8Test creative per stage and read the results quickly. On a high-volume keyword you can have results in as little as a week. Re-check the stage assumptions against the post-click behavior you observe.

The passage's laser printer keyword set shows the mechanics. "Laser printer" is generic and a mix of stages, so you bid moderately and use a broad benefit message. "Laser printer review" and "best laser printer" are likely not quite ready to buy but susceptible to messages about a particular unit's benefits, so you write benefit-led copy and send clicks to comparison-style content. "Cheap laser printer" is probably close to purchase but price sensitive, so you lead with price. "Hp 1200 laser printer" is close to purchase because of the model number, so you send it to that product page with a buy call to action. Value the clicks with the passage's index arithmetic: 20 newsletter registrations (value 1) and 10 brochure requests (value 0.7) on $1,000 spend give (20*1 + 10*0.7)/1000 = 0.027. If registrations rise to 40 in another segment, the index is 0.047. You then raise bids on the higher-scoring stage segment and lower them on the lower-scoring one.

• If you are a manufacturer or brander, then fund early-stage keywords (review, compare, best) and value their post-click actions; if you are a retailer, then weight spend toward model-number and price-modifier terms.
• If a keyword is generic and ambiguous, then bid conservatively and let post-click behavior data decide whether it deserves more.
• If a keyword carries a model number or "cheap," then treat it as near purchase and send it to a product-specific page, not the home page.
• If a Google ad gets a poor CTR, then rewrite for compelling copy before anything else, because the CTR bonus affects position and low CTR risks demotion or deactivation.
• If you cannot measure the value of early-stage actions, then derive it from hard data (the average purchase rate of catalog requesters) or an intercept survey before adjusting bids.

• Bidding on last-click purchases alone, which starves research-stage terms; the tell is that review/compare keywords look unprofitable and get cut, even though they feed later buyers.
• Sending every keyword to the home page or one generic page; the tell is a high bounce rate on research-intent terms and weak registration or information-request rates.
• Using one message for all stages; the tell is benefit copy on a "cheap" query or a hard sell on a "review" query, with CTR or conversion lagging the segment.
• Setting action values arbitrarily; the tell is that the blended index ranks segments in a way that contradicts eventual sales.
• Ignoring engine differences in format and editorial rules; the tell is Yahoo!-length prequalification copy that cannot fit in Google, or Google listings demoted for noncompelling copy.

• ▸Keyword group: list the keywords sharing one intent signal (e.g., review/compare/best).
• ▸Inferred stage: early research, evaluation, near purchase, or ambiguous, with the modifier that signals it.
• ▸Campaign objective by role: state whether you want early-stage reach (brander) or ready-to-buy traffic (retailer).
• ▸Ad message angle: benefits of a particular unit, price, or specific model with call to action.
• ▸Engine and character limits: Google 25/35/35 or Yahoo! 40/190, and where prequalification fits.
• ▸Landing page assigned: the specific page matched to the stage (not the home page).
• ▸Post-click action values: list each action with its value relative to the top action set at 1.
• ▸Index calculation: sum of (actions × value) divided by spend, per group.
• ▸Bid/position adjustment: the change you make based on the index, and the review date.

1. 1Define Quality Score as the engine's prediction, not your own grade. Treat it as predicted CTR normalized for position, plus post-click relevance variables. The engine compares your ad's CTR against the norms for that position and possibly against neighboring competitors on the same query. Your only lever is to raise predicted CTR.
2. 2Stop judging ads by the position they won. Normalization is meant to put an ad in position one or six on a level playing field. Record CTR together with average position for every ad so you can see whether a CTR gain is real or just a position effect.
3. 3Structure AdGroups around tight keyword and creative sets. The passages tie campaign structure, ad relevance, keyword relevance, and starting position to Quality Score over time. Keep each AdGroup's keywords and ad copy matched so the engine can learn a clean CTR signal.
4. 4Plan the launch window deliberately. New listings in the top three positions often get a temporary Quality Score boost because data accrues faster there and the normalization curve may be imperfect. Many practitioners start aggressively to learn Quality Score quickly. Set a hard ROI ceiling for the launch bids, because aggressive bidding in the hope that Quality Score pays off later can blow your ROI.
5. 5Re-test after the launch period ends. The top-position boost may fade as the engine gathers more data on your ads and the ads around you. Compare Quality Score and CTR before and after the window and do not treat the launch-period score as your baseline.
6. 6Watch for real-time variation. Since the 2008 change, Quality Score is calculated at the time of each search query, and broad matches may be scored in real time. Segment CTR by day-part, geography, and ISP block. The source speculates these may act as adjusting factors, though Google has not confirmed them.
7. 7Treat content (contextual) traffic as a separate Quality Score regime. Google gives content ads a different Quality Score, and targeting happens at the AdGroup level, not the keyword level. Write ad copy for browsers rather than late-stage buyers. Build variant AdGroups with different keyword combinations, synonyms, phrases, and match types, and test them.
8. 8Check the landing page against the traffic source. A page tuned for search visitors may not suit contextual visitors, and post-click relevance feeds the score. Match the page's information mix to what that visitor wants to learn.

The passages describe a contextual-campaign case rather than numbers. A marketer's search campaign hits its targets, so the same account is extended to Google content placements. Content ads receive their own Quality Score, and targeting works at the AdGroup level. The marketer finds the search ad copy underperforms, because surfers are not scanning for a keyword and are not in the final stage of the buying cycle. The marketer rewrites the ads, builds several AdGroups with similar keywords and creative but different combinations (added keywords, synonyms, phrases, match types, with Broad Match likely giving the most impressions), and adjusts the landing page for browsers rather than shoppers. Because any gain in perceived relevance on content lowers costs and raises click volume at once, the marketer tracks each variant's CTR and cost. Day-parting and geo-segmentation then bring cost per lead or cost per order into compliance with target. The passages note that the firm AdMetrica specializes in creating many iterations of a Google content campaign to gain volume, so you can run a small version of that approach yourself.

• If a CTR gain appears only when your ad sits in the top three positions, then treat it as possibly temporary. Re-check after the engine has gathered more data on you and your competitors.
• If you are launching a new AdGroup or campaign, then bid aggressively enough to learn Quality Score fast, but cap spend at your ROI limit. Do not overbid on the theory that a high score will repay it later.
• If an ad improves predicted CTR at any position, then keep it. Improving predicted CTR regardless of position is always good.
• If content-network results are weak, then change the ad copy, AdGroup keyword mix, and landing page. Do not copy search settings over, because content ads have a different Quality Score.
• If CTR swings by day-part, geography, or ISP, then segment by those variables and adjust bids and ad tests accordingly. Do not rely on whole-account averages.

• Bidding hard at launch to buy a long-run Quality Score. The tell: ROI falls in the first weeks and the score advantage does not persist once the top-position data advantage fades.
• Reading Quality Score as position-dependent or position-independent without checking. The tell: you explain CTR changes by position alone, yet the engine says it normalizes for position and you see unexplained variation.
• Mixing many unrelated keywords in one AdGroup, especially for contextual targeting at the AdGroup level. The tell: content impressions and CTR are weak even though the individual keywords perform in search.
• Reusing search ad copy and landing pages on content traffic. The tell: content CTR and conversion rates trail search while the copy still reads as late-funnel.
• Assuming the score is static after the 2008 change. The tell: CTR on broad-match terms swings by time of day, location, or ISP without any change on your side, because the score is calculated at each query.

• ▸AdGroup / campaign name: identify the unit being scored and whether it is search or content.
• ▸Launch date and starting position: record when it began and the position it first held, since early top-three data can inflate the score.
• ▸CTR by ad and average position: log both together so you can separate relevance from position.
• ▸Quality Score reading (date-stamped): record the reading at launch, after the first data period, and after the launch boost may have faded.
• ▸Keyword and creative set: list the keywords, match types, and ad copy in the group to check tight relevance.
• ▸Segment CTR (day-part / geography / ISP): note any variation suggesting real-time adjustment.
• ▸Landing-page match: state who the visitor is (browser or shopper) and what the page tells them.
• ▸Launch ROI cap: the maximum overspend you accept while learning Quality Score.
• ▸Decision and next test: keep, rewrite, split the AdGroup, or reduce the bid, with the reason.

1. 1Treat CTR as an input to position, not just an outcome. In Google, the passages say a Click-Through Rate position bonus is factored in when the engine decides which position a given bid earns. Write down for each priority keyword: current bid, current position, current CTR. Without this baseline you cannot tell whether a position change came from price or from clicks.
2. 2Respect the engine rules that shape clicking. Google: no 'click here', and listings with noncompelling copy and poor CTR risk demotion or deactivation. Yahoo!: no superlatives. Yahoo! allows more characters, so use the room for a bit of prequalification. Google rewards the click, so prequalification there costs you CTR.
3. 3Write each listing to the stated format limits: title from 25 to 40+ characters, description from 70 to 190 characters, depending on engine. Treat the listing as the 'cover of the catalog or envelope'. It is the first thing the searcher sees after typing a need such as 'cashmere sweater' or 'air purifier', so it must mirror that need.
4. 4Run creative tests like envelope tests, but faster. The source says results can come in as little as a week on a high-volume keyword. Change one listing element at a time, run variants simultaneously on the same keyword, and compare CTR.
5. 5Do not judge a high-CTR ad on CTR alone. The source says all clicks are not created equal. After the click, send the searcher to the best landing page for the search terms entered, not the home page. Then check cost per order or cost per action (CPO/CPA) against target. A listing earns its place only if clicks also convert within the cost objective.
6. 6Rebuild your volume forecast around CTR. Spend predictions for Yahoo!, Google and Looksmart assume predicted positions, CTRs, CPCs and click volumes. Those assumptions are guesses until you have your own CTR data. Replace them with measured CTR, and plan budget flexibly around ROAS or CPO/CPA rather than a fixed click number.
7. 7Adjust position keyword by keyword using post-click results. Scale back keywords (lower position) that exceed your conversion cost objective. Look for keywords coming in under CPO/CPA target and raise their positions, since CTR and position move volume together.
8. 8Expect no guaranteed position in XML feeds or directory-paid inclusion. Since position is not guaranteed, no click volume or spend can be fixed there. Do not apply auction CTR logic to those feeds.

The source gives a scale-up problem. A $5,000 month-long test works: ROI, CPO/CPA, ROAS or branding metrics meet expectations. You want $20,000 a month. The source warns that unless you started at very low positions, that inventory may not exist. If it does, you pay more per click as you displace advertisers above you who earn the engine more revenue. Raising bids alone is the expensive route. The CTR route (my application of the source, not a quoted procedure) is to improve the listing first. Because Google factors a CTR position bonus into the position a bid earns, a better-clicked listing can hold position at a given bid. Take the source's second illustration: 40% of initial spend comes from 20 high-volume keywords, and half of them (10) exceed post-click conversion cost objectives. You cut position or eliminate those ten. Then you shift budget to keywords under their CPO/CPA target. Broaden to more keywords only while conversion holds, because diminishing marginal returns creep up. Check the high-volume keyword tests weekly, since the source says a week can be enough to read results.

• If a Google listing has poor CTR, then rewrite the copy before raising the bid. Poor CTR risks demotion or deactivation, and a better CTR earns a position bonus.
• If a keyword's CTR is high but its CPO/CPA exceeds your objective, then lower its position or drop it. Clicks do not pay if post-click conversion fails. Check the landing page before cutting.
• If a keyword converts below its CPO/CPA target, then raise its position to capture more clicks. Watch the cost per click, since extra volume comes at a higher per-click price (inverse volume discount).
• If you are forecasting spend for a new campaign, then treat CTR, position and click volume as assumptions. Set a flexible budget tied to ROAS or CPO/CPA rather than a fixed click count.
• If you want to add Yahoo! prequalification text, then do it there (more characters, no superlatives). Do not copy the same text into Google if it lowers CTR.

• Optimizing for CTR alone: the creative promises more than the page delivers. The tell is rising clicks with CPO/CPA above target and weak conversion.
• Using banned or weak copy ('click here' in Google, superlatives in Yahoo!, vague copy). The tell is a low CTR, then lost position or a deactivated listing.
• Forecasting from assumed CTRs. The tell is actual spend and click volume landing far from the pre-launch estimate, because the prediction assumed positions, CTRs, CPCs and volumes that did not materialize.
• Scaling budget by bid increases alone. The tell is cost per click climbing as you displace advertisers above you, and CPO/CPA drifting up as budget goes from $5,000 toward $20,000.
• Sending clicks to the home page. The tell is a good CTR and decent traffic but poor conversion, because the searcher is left to 'flip to the right page'.
• Ending tests too early or running them one after another. The tell is conflicting CTR conclusions. Run variants at the same time on a high-volume keyword, where the source says a week can be enough.

• ▸Keyword: the exact term and engine (Google or Yahoo!) so the right editorial rules apply.
• ▸Current bid and position: record both before any test.
• ▸Baseline CTR: measured CTR for the current listing, with the date range.
• ▸Listing variant tested: title (25-40+ chars) and description (70-190 chars) text; note which element changed.
• ▸Variant CTR and days run: the source says a high-volume keyword can read in about a week.
• ▸Editorial check: confirm no 'click here' (Google) and no superlatives (Yahoo!).
• ▸Landing page used: confirm it matches the search term, not the home page.
• ▸Post-click CPO/CPA vs. target: the actual cost per order or action against your objective.
• ▸Action: raise position, hold, scale back, or eliminate, with one-line reason.

1. 1Set a goal CPO/CPA for the campaign before touching bids. If listings sell different products or attract different buyer profiles, set the goal per listing. Pair it with the CPO/CPA you know across all your other marketing media, and with the level at which you still make a reasonable profit.
2. 2Manage each listing against that goal, not against position, click volume, or CPC caps alone. Post-click behavior goes into the ongoing campaign so spend concentrates on keywords that work.
3. 3Sort keywords by post-click result. Treat the high-volume words first: in the source, 40% of initial spend came from 20 high-volume keywords. Where a keyword's cost per conversion exceeds your objective, scale it back by lowering its position, or remove it. Where it comes in under target, look for ways to buy more of that inventory.
4. 4Raise or lower bids keyword by keyword, then re-measure conversion after every position change. A new position changes the network makeup within vendor distribution, and top positions often attract more compulsive clickers, so a conversion rate observed at one position does not carry to another.
5. 5Look for the balance point between ROI and total profit. The highest-ROI campaign is your best listing at minimum CPC, but it produces almost no volume. Accept a lower per-unit profit in exchange for more conversions until the added profit from extra spend approaches zero. Raise a CPC too far and profit goes negative. Drop it too low and you lose gross profit to opportunity cost.
6. 6Add conversion types beyond the immediate trackable sale. Source gaps include sales from dropped cookies, sales after cookie expiry, untracked offline sales, and brand lift from ad exposure and site interaction. Fold offline behavior into the analytics so it informs bids.
7. 7Profile your most profitable customers. Find what they share in purchase cycles: multiple brand touch points over time, and prepurchase research searches (most purchase decisions are preceded by two to six or more searches, per the comScore reference). Use those patterns to decide which earlier-stage actions count as meaningful progress.
8. 8Plan budget around ROAS or CPO/CPA rather than a fixed spend. Spend forecasts depend on assumed positions, CTRs, CPCs, and volumes, and XML and directory feeds guarantee no position or volume, so a flexible budget tied to conversion economics is the more reliable approach.

The source's own illustration: a $5,000 month-long test worked well, with ROI, CPO/CPA, and ROAS in line with expectations. You now want $20,000 a month. That inventory may not exist, and if it does it costs more per click because you displace advertisers above you. So broadening volume has to be steered by post-click behavior. Suppose 40% of your spend sits on 20 high-volume keywords. Half of them exceed your CPO/CPA goal, so you reduce their positions or remove them, and you put the freed budget toward keywords converting under target. The source also shows the opposite trap: a single best listing at $25 a month that drives $3,000 in sales at a 50% profit has superb ROI but no volume. You accept a lower ROI on more listings in exchange for higher total profit.

• If a keyword's CPO/CPA exceeds your goal, then lower its position or remove it; if it beats the goal, then look for ways to increase its volume.
• If you change a keyword's position, then re-measure its conversion behavior before trusting the old rate, because network makeup and click quality shift with position.
• If a listing's ROI is excellent but its conversion volume is negligible, then accept a lower ROI to buy more traffic until marginal profit approaches zero.
• If you can't forecast spend (new campaign, XML or directory feeds), then budget against ROAS or CPO/CPA rather than a fixed amount.
• If your conversion tracking covers only immediate online sales, then add cookie-expiry, offline, and brand-lift behavior before cutting keywords that look unprofitable.

• Running a search engine sales approach, judging only immediately trackable sales. The tell: keywords get cut because of missing credit for dropped-cookie, post-expiry, or offline sales, and opportunity cost goes unseen, so net search profit falls.
• Optimizing to the single highest ROI. The tell: a tiny, efficient listing with trivial conversion volume, like $3,000 in sales off $25 of spend.
• Treating all clicks as equal. The tell: one blended CPA for a keyword list where half of the high-volume terms are above target and subsidized by the rest.
• Pushing to higher positions without re-checking conversion. The tell: volume rises but conversion quality drops, since top positions bring more compulsive clickers and a different network mix.
• Scaling budget linearly from a successful test. The tell: the $5,000 test result is assumed to hold at $20,000, ignoring inverse volume discounts and rising per-click costs.

• ▸Keyword / listing: name the term and its match position.
• ▸Goal CPO/CPA: the target set for this listing or campaign, by product or buyer profile.
• ▸Current position and CPC: record both; every change resets the conversion read.
• ▸Share of total spend: shows whether this is one of the high-volume words driving spend.
• ▸Tracked conversions: online sales, leads, registrations, and micro-conversions counted.
• ▸Untracked or offline conversions: estimated offline sales, post-cookie sales, dropped-cookie sales.
• ▸Actual CPO/CPA vs. goal: over or under, and by how much.
• ▸Action: scale back, remove, hold, or boost.
• ▸Marginal profit check: is the added spend still earning incremental profit, or has it reached zero?

1. 1Treat search clicks as fixed inventory. Before setting a budget, note that only a certain number of people search each keyword per day at the major portals, so the ceiling on clicks sits outside your control. Set position, click-volume, or CPC-average/cap objectives against that ceiling.
2. 2Pay for position, not for clicks in the abstract. In an auction, the per-click rate that produces your desired position is the price of the inventory. Record the bid-to-position relationship for each keyword group so you know what each position costs.
3. 3Plan scale-ups with inverse volume discounts in mind. When you raise spend, expect a higher per-click price because you displace advertisers above you who earn the engine more revenue. For a proven test, check first whether the added inventory exists, since it may not unless you started at very low positions.
4. 4Expand volume through keywords before bids. Use every relevant campaign permutation you can find while keeping conversion high. Add low-volume keywords in aggregate, and watch for the point of diminishing marginal returns, which the source says can creep up on you.
5. 5Model CTR as a driver of price and position, not only as a result. Click-Through Rate affects position and therefore click volume, so any spend forecast based on assumed positions, CTRs, CPCs, and volumes is approximate. Label forecasts that way and revise them as live data arrives.
6. 6Manage in near real time without negotiating. The auction cannot be bargained with, so adjust listings (or have a listings-management firm adjust them) as price and position move. Bring post-click behavior into the adjustments and concentrate on the keywords that work.
7. 7Route premium or fixed placements through reps. If you need negotiated terms, use the premium keyword placements portals offer through reps, and expect longer commitments.
8. 8Check the competitive mix around your listings before cutting paid. Review the paid and organic listings that surround yours and your channel partners'. If competitors surround you, good PPC visibility matters more. Also confirm your organic rank holds up under personalization, universal search elements, and syndication partners, since true visibility may be much lower than rank suggests.

The source's own illustration: a $5,000 month-long test works, with CPO/CPA, ROAS, or branding metrics in line with expectations. The advertiser then wants $20,000 a month. Unless the first test ran at very low positions on those keywords, the extra inventory may not exist. If it does, each click costs more because the advertiser displaces competitors in the positions above, who earn the engine more revenue. The advertiser widens the keyword list to every relevant permutation that keeps conversion high and adds low-volume terms. The source warns that diminishing marginal returns can arrive sooner than expected. The practical path is to scale in steps, tracking post-click performance by keyword, rather than assuming the $5,000 results will quadruple with spend.

• If you want a much larger budget on the same keywords, then assume per-click cost rises and some inventory may not exist; expand keyword breadth first and test in steps.
• If you are surrounded by competitors in paid and organic listings, then treat PPC visibility as strategically more important rather than turning it off because of high organic rank.
• If your messaging on organic pages does not convert and changing them risks rank, then rely on paid titles, descriptions, and landing pages to carry the message.
• If you need predictable placement and negotiated terms, then use rep-managed premium placements and accept longer commitments; otherwise accept the auction's real-time pricing.
• If pure search is hard to extend after several rounds of expansion and segmentation, or you are in a highly competitive category, then begin testing contextual or behavioral media, provided you can monitor interaction effects with search.

• Linear scaling: assuming a $5,000 result becomes $20,000 at the same ROI. The tell is a sharp rise in CPC or flat click volume after the increase.
• Treating a spend forecast as a commitment. Predictions assume specific positions, CTRs, CPCs, and volumes. The tell is actuals drifting from forecast because CTR moved position.
• Ignoring post-click quality while chasing position or volume. The tell is hitting click or position targets while CPO/CPA worsens.
• Believing high organic rank equals real visibility. The tell is personalization, universal search elements, or syndication partners displaying you lower than the rank you see.
• Keyword expansion without a stopping rule. The tell is a growing list of low-volume terms whose combined return flattens or whose conversion falls.

• ▸Keyword group: name the set of keywords being evaluated.
• ▸Current position and CPC: record the average position and per-click rate delivering it today.
• ▸Current budget and result: state spend and the CPO/CPA, ROAS, or branding metric achieved.
• ▸Target budget: the higher spend you want to reach.
• ▸Inventory exists? (Y/N/unknown): note whether the target volume is plausible at your current positions.
• ▸Competitors above you: list advertisers in the positions you would displace.
• ▸Added keyword permutations: list new relevant keywords and their expected conversion.
• ▸Post-click performance by keyword: record conversion or CPA to decide which keywords to keep.
• ▸Review trigger: define the CPC or CPA level at which you stop scaling or revise the plan.

1. 1Treat position as a purchased asset per keyword. Each position on each keyword has its own clearing price, like its own stock, so set a target position and a maximum bid for each keyword, not one bid for the campaign.
2. 2Derive each keyword's bid from its post-click conversion attributes and your ROI objective (CPO/CPA, ROAS, or branding metric). The source says you need a system that sets the right bid for every keyword at every point in time. Bid only what the keyword's conversion economics support.
3. 3Write the listing to earn position, not only clicks. Google factors a Click-Through Rate position bonus into the position it awards for a given bid, so a compelling title and description can win a better position at the same bid. Observe the format limits: Google 25/35/35 characters for title and two description lines; Yahoo! 40 for title and 190 for description. Observe the editorial rules too: no 'click here' in Google, no superlatives in Yahoo!.
4. 4Re-rank keywords by post-click behavior on a recurring schedule. Move positions down, or cut the keyword, where cost exceeds your CPO/CPA target. Raise positions where cost is under target.
5. 5Test brand terms even if your organic rank is high. Buy the top paid position on the brand keyword to gain extra SERP real estate, full control of title and description, and a tunable landing page. Top brand positions are usually cheap because of the engines' relevance algorithms.
6. 6Plan scale-ups against inverse volume discounts. Moving from a $5,000 test to $20,000 a month may find no inventory at your position. Where inventory exists, you pay more per click as you displace advertisers above you. Expand to more relevant keyword permutations first, and watch for diminishing marginal returns.
7. 7Set the objective before bidding: position, click volume, or CPC average/cap. Fold post-click performance into it. You cannot negotiate in the auction, so objectives must be mapped in advance. Where you need negotiation, use premium placements, which involve reps and longer commitments.
8. 8Budget flexibly. Spend forecasts assume positions, CTRs, CPCs and volumes that may not hold. XML feeds and directory paid inclusion guarantee no position, so no spend or traffic can be predicted. Anchor the budget to ROAS or CPO/CPA rather than a fixed dollar figure.

The source describes a $5,000 month-long test whose ROI, CPO/CPA or ROAS met expectations. Management then asks for $20,000 a month. Unless the test sat at very low positions, the extra inventory may not exist at that position. If it does, you pay more per click because you displace advertisers above you who earn the engine more revenue. The response is to broaden to more relevant keyword permutations while keeping conversion high, and to stop where diminishing returns appear. Separately, the source describes 40% of initial spend concentrated in 20 high-volume keywords, half of which exceed post-click conversion cost objectives. Those ten get lower positions or are removed. Keywords under the CPO/CPA target get boosted. For a brand term, the source reports that every test showed a very high net ROI from paying for the top position even with strong organic rank. The gains were more SERP real estate, control of the offer text, a cheap top position and a higher combined conversion rate.

• If a keyword's post-click cost exceeds your CPO/CPA target, then lower its position or remove it. If it comes in under target, then raise the position and boost that inventory.
• If you ask for several times the spend (for example $5,000 to $20,000), then expect higher CPC and possibly no inventory. Add relevant keywords before raising bids on the same terms.
• If you hold high organic rank on your brand, then still bid on the brand keyword, because the extra SERP real estate, offer control and cheap top position produced a high net ROI in every test.
• If your position is poor at a bid that is otherwise competitive, then improve ad copy CTR. Google's CTR position bonus can raise the position that a bid wins.
• If position and spend cannot be predicted (XML feeds, paid inclusion), then budget on ROAS or CPO/CPA rather than on fixed volume.

• Chasing the top slot on every keyword. The tell: spend rises while CPO/CPA overshoots target on high-volume terms.
• Linear scale-up assumptions. The tell: a plan to multiply budget fourfold at the same CPC, then flat clicks or sharply higher average CPC.
• Skipping brand terms because organic rank is high. The tell: competitors or other listings appear above you on your brand, and you have less SERP real estate and offer control.
• Weak, non-compelling copy. The tell: low CTR leaves you in a worse position for the same bid, and in Google the listing risks demotion or deactivation.
• Ignoring that position changes who clicks. The tell: conversion rate shifts after a position change because network makeup changes and top positions attract more compulsive clickers.

• ▸Keyword: the exact term being bought.
• ▸Current average position: where the ad appears today.
• ▸Target position: the position justified by economics, not by ego.
• ▸Max CPC bid: the highest bid your CPO/CPA or ROAS target supports.
• ▸Post-click CPO/CPA vs. target: actual cost per conversion against your objective.
• ▸CTR and ad copy status: current CTR and whether copy meets character and editorial limits.
• ▸Action: raise position, hold, lower, or remove, per the post-click result.
• ▸Brand term? (Y/N): if Y, record SERP real estate gained and the paid-plus-organic conversion rate.
• ▸Next review date: when you re-rank this keyword.

1. 1Fix the objective before you bid. Choose position, click volume, CPC average/cap, or a return metric (CPO/CPA, ROAS), because the auction reprices in near real time and you cannot negotiate inside it.
2. 2Estimate each keyword listing's true value to you. Research post-click behavior per keyword, then calculate the maximum price you can pay for traffic while still maximizing profit. The source says you have this advantage over competitors only if you do the homework.
3. 3Build the value model at the level your business supports. A simple ROI formula works for simple businesses. A richer one folds in immediate profit, multiple conversions, cross-selling, and lifetime value, which explains why a competitor's bid can look ludicrous but be rational.
4. 4Set the budget as a flexible ceiling tied to ROAS or CPO/CPA, not a fixed spend. Predicting spend is hard because it depends on assumed positions, CTRs, CPCs, and click volumes. XML and directory-paid inclusion feeds guarantee no position, so no spend or traffic volume can be predicted.
5. 5Run a controlled test (the source's example is $5,000 for a month) and confirm CPO/CPA, ROAS, or branding metrics hit expectations before you scale.
6. 6Scale in steps and watch the inverse volume discount. More clicks and visibility cost more per click, and the extra inventory at higher positions may not exist. Check that the incremental volume still clears your maximum allowable cost, not just the blended average.
7. 7Expand keywords to find volume at lower cost, using every relevant permutation that keeps conversion high. Watch for diminishing marginal returns as low-volume keywords accumulate.
8. 8Test elasticity. Ask whether a slightly higher CPA buys meaningful extra volume at low incremental marginal cost, and use a bid engine that is predictive (learns from repeated patterns) and reactive (responds to shifts in conversion rate or search intent).
9. 9Keep refining beyond bidding. Bid management only buys clicks. Profit also requires continuous campaign refinement for conversion and profitability, and ongoing removal of keywords that do not work post-click.

The source's own scenario: a $5,000 month-long test works well, with CPO/CPA and ROAS in line with expectations, and everyone wants to move to $20,000 a month. The source warns that unless you were at very low positions on those keywords, the inventory may not exist at that level. If it does, you pay more per click because you displace advertisers who earn the engines more revenue. So you do not multiply the $5,000 results by four. You raise spend in steps, compare the marginal CPA of each added tranche against your maximum allowable CPA, and add relevant keyword permutations to find cheaper volume. You stop where the extra volume costs more than it returns, even if the blended average still looks fine. The source adds that the best budget answer is often a flexible one anchored to ROAS or CPO/CPA rather than a fixed dollar figure.

• If a spend increase pushes you toward higher positions, then judge the added clicks on marginal CPA, not blended CPA, because per-click price rises as you take more volume.
• If you cannot predict spend (first-time search, XML or paid-inclusion feeds), then set the budget by a ROAS or CPO/CPA target instead of a fixed amount.
• If a competitor's bid looks irrational, then do not follow it into a bid war. Bid only up to your own calculated value per keyword, since they may be valuing lifetime value or cross-sell that you are not.
• If conversion rate on a keyword shifts because search intent changes (the source's 'Janet Jackson' example), then let a reactive bid engine cut the bid or pause the keyword rather than waiting for a monthly review.
• If a slightly higher CPA yields a large volume gain at low marginal cost, then accept it. If keyword additions drag conversion down, then stop adding them.

• Linear extrapolation: you assume $5,000 results scale to $20,000. The tell is blended CPA holding while marginal CPA on new tranches balloons, or the inventory is simply not available.
• Bidding wars: you raise bids to regain position out of emotion. The tell is bids exceeding your computed maximum price per keyword while profit per order falls.
• Optimizing clicks only: you rely on bid management without refining conversion. The tell is strong click and position metrics with flat or falling profit.
• Using past data as a guarantee: you ignore pricing volatility and intent shifts. The tell is a sudden conversion drop on a keyword that had a stable history.
• Judging by ROAS or first-order ROI alone while ignoring multiple conversions, cross-selling, and lifetime value. The tell is cutting keywords that competitors keep bidding on profitably.

• ▸Keyword/segment: name the listing or group being evaluated.
• ▸Test spend and period: the spend level and timeframe that produced your baseline (e.g., $5,000 per month).
• ▸Baseline CPO/CPA and ROAS: results from the test, measured post-click.
• ▸Value per order/lead: immediate profit, plus any multiple-conversion, cross-sell, or lifetime value you can defend.
• ▸Maximum allowable CPA / max price per click: the most you can pay while still maximizing profit.
• ▸Proposed new spend level: the budget you want to move to.
• ▸Marginal CPA of the added tranche: the cost per order of the incremental spend only, not the blended figure.
• ▸Elasticity note: volume gained for each increase in CPA or CPC.
• ▸Decision: hold, scale one more step, or cut, with the reason.

1. 1Start from the principle the source states: a bid should come from a business model, not a hunch. Models range from simple ROI formulas to 'sophisticated valuation formulas that take immediate profit, multiple conversions, cross-selling, and lifetime value into account.' Decide which tier you run, and write the formula down.
2. 2Separate the profit you earn on the first order from the profit that follows it. The follow-on profit comes from repeat purchases, cross-sells, and extra conversions by the same customer. The source gives no formula or retention rates, so pull these from your own order history. Do not borrow benchmarks.
3. 3Tag each keyword with the post-click behavior of the customers it brings in. The source tells you to 'do research into the post-click behavior of each keyword listing.' Track which keywords produce first-time buyers who come back or cross-buy, not only which produce first orders.
4. 4Compute a maximum allowable CPA or CPC per keyword: the most you can pay for traffic 'while still maximizing profits.' Run it twice, once on immediate profit only (your floor) and once with lifetime value included (your ceiling). Label the gap between them as the investment you are choosing to make in future profit.
5. 5Test the click price against volume. The source calls this market elasticity: the tradeoff between click price and sales or lead volume. Raise bids in steps toward the lifetime-value ceiling and record the extra orders each step buys. Stop where the marginal customer's expected profit, lifetime included, no longer covers the marginal cost.
6. 6Check channel partners and competitors against lifetime value. The source notes that a partner may deliver a lower immediate and lifetime profit than capturing the customer yourself. If your own PPC listing earns more lifetime profit per customer than the partner route, that supports bidding for the position.
7. 7Make the bid decision per keyword, 'at every point in time.' Put the lifetime-adjusted ceiling into whatever sets bids, whether manual or campaign-management software. Re-run the calculation on a schedule, because 'past data does not guarantee future results.'
8. 8Use landing-page testing to raise first-order conversion without touching the ceiling. The source suggests testing a results page, a top-four best-sellers page, and a product page with cross-sell to alternatives. A cross-sell page can also lift the multiple-conversion component of lifetime value.

The source's own illustration is the 'New York Lawyer' keyword, where CPCs exceed $10. A service with great lifetime value can justify that price. The same price makes the keyword a target for competitive click fraud: a competitor clicking once a day from five people on a listing in each of the two major engines adds about $1,500 a month per engine, $3,000 in total, with no chance of benefit. A bidder who counts only the first fee would call a $10+ click absurd. A bidder who values the whole client relationship sees a rational price. The source makes the same point about auctions: a bid that looks 'ludicrous' may come from a marketer with a sophisticated valuation model. A second data point is the account rep who pushes spend from $12,000 to $15,000 a month. Test that extra $3,000 against the incremental customers it buys, valued at lifetime profit rather than first-order profit. The source gives no retention figures, so supply them from your own records. Then accept or reject the increase on that basis.

• If a keyword's maximum CPA is justified only by lifetime value and not by first-order profit, then bid up to the lifetime ceiling only after your own data shows the repeat or cross-sell behavior. Otherwise cap at the immediate-profit floor.
• If raising the bid adds orders whose lifetime-inclusive profit still exceeds the added cost, then raise it. If the next step buys fewer orders than it costs, hold the bid. That is the elasticity limit.
• If a competitor's bid looks irrational, then do not match it until you know your own ceiling. The source warns that bidding wars are destructive and that the other bidder's data and formulas are invisible to you.
• If your CPCs are high because customer value is high, then add click-fraud monitoring, since competitors can inflate a budget fastest on exactly those terms. Review server logs and IP addresses for repeat visits through paid links.
• If a channel partner would deliver a lower immediate and lifetime profit than capturing the customer yourself, then prioritize your own PPC visibility on that keyword.

• Setting bids on first-order ROI alone. The tell: you lose auctions to competitors whose bids look 'ludicrous' to you, while your own acquisition volume stalls.
• Assuming lifetime value without evidence. The tell: the lifetime ceiling is a round number nobody can trace to repeat-purchase or cross-sell records, and past data is treated as a guarantee.
• Letting lifetime value justify an unlimited bid. The tell: a bidding war that escalates past the position's value and ignores the click-price versus volume tradeoff.
• Applying one lifetime value to every keyword. The tell: the same ceiling on all terms, with no research into the post-click behavior of each keyword's buyers.
• Ignoring click fraud on high-CPC terms. The tell: spend rises on expensive keywords with no matching orders, and the logs show repeat visits from the same IPs through paid links.

• ▸Keyword and match scope: name the single keyword or tightly grouped set this sheet covers.
• ▸First-order profit per conversion: profit on the initial sale, from your own data.
• ▸Repeat/cross-sell profit per customer: profit from later purchases and cross-sells, taken from your order history. Mark 'unknown' if you lack it.
• ▸Evidence source for the follow-on figure: note which post-click or order data supports it and its date, since past data does not guarantee future results.
• ▸Immediate-profit CPA floor: the maximum CPA using first-order profit only.
• ▸Lifetime-inclusive CPA ceiling: the maximum CPA with follow-on profit added.
• ▸Elasticity test log: each bid step, the click price, and the order or lead volume it produced.
• ▸Chosen bid and reason: the bid you set, between floor and ceiling, with the one-line rationale.
• ▸Fraud/competitor check: note repeat-IP visits through paid links and the competitive mix around your listing, and your review date.

1. 1Declare branding as an explicit objective before any spend. The source says to map objectives clearly (position, click volume, CPC average or cap) because you cannot negotiate in an auction. Write down which keywords are bought for awareness or favorability and which for orders, so branding spend is not judged on CPO alone.
2. 2Separate search sales from search marketing. Marketing creates interest, awareness, and demand; sales only fulfills demand that already exists. Tag each keyword group as demand-creating or demand-fulfilling, then apply a different success test to each.
3. 3Treat impression visibility as an asset with its own value. The source says even the visibility of a listing being served may have a measurable branding value. Record impressions and average position by keyword so you can price visibility apart from clicks.
4. 4Pick your proxy metric and hold its definition fixed. The Branding Effectiveness Index is a proxy for awareness, favorability, and purchase intent. The passages give no formula, so use your vendor's stated definition and do not change it mid-test. Record a baseline before the test period.
5. 5Write copy that carries the brand within each engine's rules. Yahoo! allows some prequalification or branding in copy, though it bars superlatives. Google demotes or deactivates listings with poor CTR, so branding copy must still earn clicks. Respect the character limits (Google 25/35/35 for title and two description lines) and the "no click here" rule.
6. 6Test position changes against the branding proxy, not clicks alone. Position changes alter conversion behavior because network makeup changes and top positions draw more compulsive clickers. Log the proxy, CTR, and CPO side by side for each position tested.
7. 7Scale in steps and expect inverse volume discounts. More visibility costs more per click, because you displace advertisers who earn the engine more. Raise the budget in increments and re-read the proxy at each step. Stop when the cost per unit of lift exceeds what the brand objective is worth.
8. 8Add behavioral or contextual media only when the source's conditions hold: pure search is hard to improve further, the category is highly competitive, you are building awareness alongside direct response, and you can monitor the interaction effects between search and quasi-search buys.

The source's own scaling scenario shows the mechanics. A $5,000 month-long test works: ROI looks good, and CPO/CPA, ROAS, or branding metrics are in line with expectations. Management wants $20,000 a month. The source warns that unless your original positions were very low, that inventory may not exist. If it does, it costs more per click because you displace advertisers who currently sit above you. Applied to branding lift: record the proxy index and impression position at $5,000 as your baseline. Move to a mid-step, for example $10,000 (our illustration, not from the source), and re-read the index. Broaden to more relevant keywords, since the list is limited by your business sector and low-volume terms add up. Compare the index change per added dollar against the baseline. If the cost per point of lift rises sharply, you have hit the inverse volume discount. The source does not report a resulting index figure, so your own measurements decide where to stop.

• If a keyword group is meant to build awareness, then judge it on the branding proxy and visibility, not CPO alone. If it is meant to fulfill demand, judge it on CPO/CPA and ROAS.
• If a position change moves the branding proxy but conversion behavior shifts too, then treat them as separate effects, because the click population changes with position.
• If a budget increase raises the CPC and the proxy gain does not keep pace, then you have hit the inverse volume discount. Hold spend, or widen the keyword set instead of buying higher positions.
• If you are on Google and branding copy lowers CTR, then rewrite it, because poor CTR risks demotion or deactivation. Put heavier prequalification or branding copy on Yahoo!, where it is more tolerated.
• If pure search stops yielding and you are building awareness in a competitive category, then test behavioral or contextual media. Do it only if you can account for the interaction effects with your search buy.

• Judging every keyword by CPO. The tell: awareness-oriented listings get paused as unprofitable while branding metrics fall after the cuts.
• Changing the proxy definition between periods. The tell: the index moves but nothing else about the campaign changed.
• Assuming scaled spend repeats test results. The tell: the proxy and CPO that looked good at $5,000 degrade at $20,000 because of inverse volume discounts and displaced competitors.
• Writing branding copy that ignores engine rules. The tell: low CTR on Google, then demotion or deactivation. Or a Yahoo! rejection for superlatives.
• Treating the engines as the only counterparty. The tell: you read a rising CPC as an engine fee increase, when a competitor is buying the position and the visibility out from under you.

• ▸Keyword group and intent: name the group and mark it demand-creating (awareness) or demand-fulfilling (sales).
• ▸Baseline proxy value: record the Branding Effectiveness Index (or your vendor's equivalent) before the test, with its definition.
• ▸Baseline impressions and average position: log the visibility you hold today.
• ▸Monthly spend and CPC: record the budget level and average CPC for this test step.
• ▸Copy variant and engine rules check: note title and description text, and confirm the character limits and banned phrases are met.
• ▸Post-test proxy value: record the index after the period, using the same definition.
• ▸CTR and CPO/CPA at this step: record these to see whether position changes altered conversion behavior.
• ▸Cost per point of lift: divide the spend increase by the proxy change versus the previous step.
• ▸Next action: scale, hold, rewrite copy, or test behavioral/contextual media, with the reason.

1. 1Treat each keyword as a slice of fixed inventory. Only so many people search a given phrase each day, so list the keywords that fit your business and accept that the list has a ceiling. Your relevant keyword list, not your budget, sets the real limit on volume.
2. 2Group keywords by searcher intent and keep one tight theme per ad group. Quality Score responds to campaign structure, ad relevance and keyword relevance, so a loose group weakens all three at once.
3. 3Write each ad to match the keywords in its group. Google's Quality Score estimates CTR normalized for position, plus post-click relevance. The ad and the landing page must both answer the query you bought.
4. 4Add every permutation of relevant keywords that still holds your conversion rate. Low-volume terms add up, but drop any term whose conversion falls, and watch for diminishing marginal returns as the list grows.
5. 5Split networks into separate campaigns. Search-mode clicks and content/contextual clicks can convert differently (Danny Sullivan's "search mode" versus "surf mode"). Measure conversion as a percentage of visitors for each, and keep contextual only where it matches search quality.
6. 6Test geography and day-part as segments. Google may use day-part, geography or even ISP IP block as adjustments to real-time Quality Score. These dimensions can show CTR and conversion variation that the keyword alone hides.
7. 7If you run behavioral retargeting of searchers, set its window by how long your category takes from search to purchase. That interval differs for car buyers, consumer electronics buyers and downloadable-file buyers, and it drives how much conversion lift you get.
8. 8Before scaling, expect inverse volume discounts. Moving a proven test to a larger budget means displacing advertisers above you at a higher CPC. Scale by widening relevant targeting first, then by raising bids, and re-check CPO/CPA or ROAS at each step.

The source describes a $5,000 month-long test that worked: ROI looked good, and CPO/CPA or ROAS met expectations. The temptation is to jump to $20,000. Unless you started at very low positions, that inventory may not exist. If it does, you pay more per click because you displace advertisers who earn the engine more revenue. The sound response is to widen targeting before raising bids. Add more relevant keywords, including low-volume ones, but only while conversion holds. Add each new batch as its own tightly themed ad group so Quality Score stays healthy. Keep the content network separate and check its conversion against search. eBags' Chris Seahorn reported no drop in conversion quality on AdWords, but the source says other marketers do see a drop, so verify it yourself. Stop expanding when added terms lower conversion or marginal cost per order climbs past target.

• If a keyword's conversion rate drops after you add it, then pause it. More volume is not worth losing conversion.
• If content-network clicks convert worse than search clicks, then cut their bids or budget and run them as a separate campaign. If they match search, keep them.
• If you need more volume than your keyword list can supply, then expand through new relevant keywords before raising bids. Raising bids alone pays more per click for the same limited inventory.
• If your category has a long gap between searching and buying, then lengthen the retargeting window. If the gap is short, as for downloadable files, shorten it.
• If CTR or conversion varies by day-part or geography beyond what position explains, then split that segment out and manage it separately.

• Scaling budget linearly from a successful test. The tell: CPCs and CPO jump when spend goes from $5,000 to $20,000 because you are displacing higher-paying advertisers.
• Mixing broad themes in one ad group. The tell: ads match only loosely, CTR sags, and Quality Score erodes over time.
• Treating contextual or content clicks as equal to search clicks. The tell: overall conversion falls while search-only conversion holds.
• Judging a keyword's quality from a temporary top-three Quality Score boost. The tell: early strong numbers fade as the position-normalization advantage wears off.
• Applying one retargeting window to every product. The tell: car-style long-cycle buyers lapse before they return, or quick-purchase buyers get chased long after they have bought.

• ▸Keyword / phrase: the exact term and its intent (research, compare, buy).
• ▸Match type and ad group theme: which group it sits in and the single theme that group serves.
• ▸Network (search vs content/contextual vs behavioral): where it runs; keep separate campaigns.
• ▸Ad-to-keyword match: does the ad text use the searcher's wording? Yes/No.
• ▸Landing page relevance: does the page answer the query? Note the gap if not.
• ▸Segment (geo / day-part / audience): the slice you are testing or excluding.
• ▸CTR and conversion rate by segment: compare each against the account average.
• ▸Retargeting window (days): based on the category's search-to-purchase interval.
• ▸Action: keep, pause, split out, or expand, with the reason.

1. 1Write down the constraints for each engine before drafting. Per the source, Google allows 25 / 35 / 35 characters for title, description line 1 and description line 2. Yahoo! allows a title of 40 and a description of 190. Draft inside the tightest format first, then expand for the longer one.
2. 2Open every ad with the searcher's own query term. A searcher who types "cashmere sweater," "air purifier" or "Panasonic plasma HDTV" is looking at the listing as the cover of your catalog, so the title should reflect that term.
3. 3Split the copy job by engine. In Google, spend the limited characters on a compelling call to action, because poor CTR can get a listing demoted or deactivated. In Yahoo!, use the extra space to explain the offer and add a bit of prequalification or branding.
4. 4Check each variant against editorial rules before launch: no "click here" in Google, no superlatives in Yahoo!. Strip these out in the draft so a disapproval doesn't stall a test.
5. 5Run at least two ad versions per keyword group at the same time, as you would test envelope designs. On a high-volume keyword, results can arrive in as little as a week. Judge the winner on CTR and on post-click conversion, not CTR alone.
6. 6Match the ad to the page it lands on. Send the click to the page that best fits the search terms, never the home page by default. If you can't build custom pages, look for existing pages on your site that already fit the query.
7. 7If you opt into content (surf-mode) placements, write or tune the creative and offer for both searchers and surfers, since the two groups are in different mindsets. Compare the creative's performance by placement type rather than assuming search results carry over.
8. 8Feed results back by keyword. Keep the creative that wins on CTR and on conversion cost against your CPA/CPO target, and retire the rest.

Illustrative build (the copy is mine; the limits and rules come from the source). Keyword: "cordless drill" for a tool catalog site. Google version: title "Cordless Drill Selection" (24 characters, limit 25); line 1 "Compare cordless drills by brand" (32, limit 35); line 2 "See current prices and specs." (29, limit 35). It contains no "click here." Yahoo! version: title "Cordless Drills: Compare Models" (31, limit 40), plus a description of up to 190 characters that adds offer detail and a bit of prequalification, with no superlatives. Both ads point to the cordless drill category page rather than the home page. This follows the source's own question of whether the home page, a results page listing all cordless drills, or another page converts best. Run two title/line variants against this one for about a week on a high-volume term. Keep the winner on CTR and conversion cost. The Google CTR matters twice: it drives the click-through-rate position bonus Google weighs when awarding position for a given bid, and weak CTR risks demotion.

• If the engine is Google, then spend the characters on a compelling call to action and relevance, because noncompelling, low-CTR copy risks demotion or deactivation.
• If the engine is Yahoo!, then use the longer description to explain the offer and prequalify clicks, and keep superlatives out.
• If a keyword is high-volume, then expect a creative test result within about a week; if it is low-volume, lengthen the test before declaring a winner.
• If the traffic comes from content placements, then rewrite the creative and offer for surf-mode readers, because curiosity clickers behave differently from someone who types "music download." If the ad and offer are compelling, the quality gap may be small.
• If an ad wins on CTR but its keyword exceeds your CPA/CPO target after the click, then fix the ad-to-landing-page match or scale the keyword back; don't keep the ad on CTR alone.

• Copying one ad across engines. The tell: Yahoo! copy has unused space and no offer detail, or Google copy is truncated or rejected for length.
• Violating editorial rules, such as "click here" in Google or superlatives in Yahoo!. The tell: disapprovals, or listings that are deactivated or demoted.
• Writing generic copy that ignores the query term. The tell: weak CTR, which costs the position bonus Google applies and raises your price for the same position.
• Pointing every ad at the home page. The tell: decent clicks but poor conversions, because the searcher must hunt for the right catalog page.
• Shipping one ad and never testing, or declaring winners on CTR alone. The tell: no variants running, or a high-CTR ad sitting on a keyword that misses your CPA/CPO target.

• ▸Keyword / query term: the exact term the ad group targets (e.g., "cordless drill").
• ▸Engine and limits: Google 25/35/35 or Yahoo! 40/190; note which applies.
• ▸Title (with character count): must carry the query term and fit the limit.
• ▸Description lines (with counts): call to action for Google; offer detail and prequalification for Yahoo!.
• ▸Editorial check: confirm no "click here" (Google) and no superlatives (Yahoo!).
• ▸Landing page URL: the page that best matches the search term, not the home page by default.
• ▸Variant B hypothesis: the one element that differs (title wording, offer, call to action).
• ▸Test window and result: start date, run time (about a week on high-volume terms), CTR and conversion cost per variant.
• ▸Decision: keep, retire or rewrite, judged against your CPA/CPO target.

1. 1Stop sending paid traffic to the home page. Map each ad group's search term to the best page for that term, so the searcher lands where the query is answered. The passage's analogy: don't hand the customer your regular catalog and make them flip to the right page.
2. 2Audit existing pages before building new ones. The source says several pages already on your site may deliver dramatically improved conversions at no added development cost. List the candidates for each high-value term.
3. 3Test candidate page types against each other for the same keyword. For "cordless drill" the source names four: the home page, a search results page listing all cordless drills with descriptions, a landing page listing only the top four best-sellers, and the most popular drill product page modified to cross-sell alternatives. Only testing shows which wins.
4. 4Where resources allow, build custom landing pages so you can vary images, copy, price, and offers. Treat the landing page as the inside of the catalog or the letter in the envelope; it is the more critical piece of creative than the ad.
5. 5Segment campaigns by the predictable needs, profiles, and buying-cycle stages of searchers, then check whether your site serves each major segment. A site that is non-optimal for any major segment makes the campaign mediocre.
6. 6Align ad copy and landing page. Write ads compelling enough to earn the Google CTR position bonus. On Yahoo!, use the extra characters to explain the offer and prequalify, so the click arrives already informed.
7. 7Update keyword lists, ad creative, and landing pages for seasonality, as a catalog marketer would not mail the same catalog all year. Test and prove each change before rolling it out.
8. 8Convert the gains into bidding power. Higher conversion makes keywords that were too pricey for a top-four position affordable, so re-run bid targets after each landing page win.

The source's tool-catalog case: a retailer buys the term "cordless drill" and must pick a destination. Four options already exist on the site: (1) the home page, (2) a search results page listing every cordless drill with descriptions, (3) a page listing only the top four best-selling drills, (4) the most popular drill's product page, modified to cross-sell the alternatives. No new development is needed for any of them. You split the traffic across the four, hold the ad constant, and compare conversion for the same keyword. The source does not say which wins; it says only testing will determine the answer. Once a winner emerges, its higher conversion rate can make a top-four position affordable for that keyword. That produces a volume increase the source describes as orders of magnitude greater than lower-tier listings deliver.

• If the destination for a paid keyword is the home page, then replace it with the page that best fits the exact search term, even an existing page, before spending on a redesign.
• If you cannot build custom pages, then test existing pages (results list, top sellers, modified product page) against each other; if you can, vary images, copy, price, and offers.
• If a keyword is too expensive to hold a top-four position at current conversion, then work on the landing page first; improved conversion can make the bid affordable.
• If an agency or consultant recommends a usability or conversion project, then take it seriously even when the site was expensive or award-winning; good design does not guarantee optimal conversion.
• If a major searcher segment (by need or buying-cycle stage) converts poorly, then treat the campaign as mediocre until the landing experience for that segment is fixed.

• Using the home page as the campaign landing page; the tell is searchers who must hunt for the product they just searched.
• Assuming a beautiful, flashy site converts well; the tell is high design quality alongside weak conversion, with Flash used where it does not serve the user's objective.
• Treating the web agency as the conversion owner; the tell is creative briefs with design preferences (color, fonts, tone) but no usability parameters or user objectives, and a media team that stops at buying and trafficking.
• Writing the ad and landing page separately; the tell is a strong CTR followed by poor conversion because the page does not match the offer or query.
• Running the same pages and lists year-round; the tell is no seasonal updates to keywords, creative, or landing pages.

• ▸Keyword / search term: the exact term and ad group being tested.
• ▸Searcher segment and buying-cycle stage: the need or profile this term signals.
• ▸Current landing page: the URL now receiving the traffic (flag if it is the home page).
• ▸Candidate pages (up to 4): existing pages such as results list, top sellers, or modified product page with cross-sell.
• ▸Custom page variables: images, copy, price, offers you will vary if you build new pages.
• ▸Ad copy promise: the offer stated in the ad, to confirm the page delivers it.
• ▸Test duration and traffic split: how long and how evenly traffic is divided.
• ▸Conversion result per page: conversion rate for each candidate.
• ▸Bid implication: whether the winner makes a higher position affordable, and the new bid target.

1. 1Set the objective in profit terms before touching a bid. Choose CPA, CPO, ROAS or ROI as the governing metric, not position or click volume alone. Position, volume and CPC caps can be secondary constraints, but the post-click result decides whether a keyword stays.
2. 2Estimate each keyword listing's true value to you. Each position is its own 'stock' with its own value and clearing price. Build the estimate from your own data: immediate profit, and where you can measure them, multiple conversions, cross-selling and lifetime value. That value is your ceiling for the bid.
3. 3Raise the ceiling by raising conversion. Run simultaneous ad copy tests and landing page and offer tests. A higher conversion rate is the leverage you use when setting bids, because it lets you pay more per click and still hit your CPA target.
4. 4Bid by keyword and position, not by campaign average. Rank keywords by post-click cost against your CPO/CPA objective. Scale back (lower position) or remove those over target. Boost those coming in under target.
5. 5Check the cost of volume before you buy it. Per-click price rises as you ask for more clicks and visibility (inverse volume discount). Extra spend, such as moving from $12,000 to $15,000 a month, buys traffic that would have gone to competitors. Expect a higher cost per click, and the inventory may not exist at all.
6. 6Re-measure after every position change. A move changes the network mix within a vendor's distribution, and top positions attract more compulsive clickers. Conversion behavior at the new position is not the old behavior.
7. 7Use any available targeting bid boosts, and protect the base bid. On MSN adCenter, bid-boost by age, gender and geography for searchers you will pay more for. The base bid covers searchers MSN has no registration data on, so do not set it too low on non-targeted ads, since some of them are in your sweet spot.
8. 8Choose automation to match scale, and budget flexibly. Use a specialized vendor to adjust bids on measured conversion, or manage manually if the campaign is small. Because spend forecasts depend on predicted positions, CTRs, CPCs and volumes, set the budget as flexible and tied to ROAS or CPO/CPA rather than a fixed sum.

The source describes a campaign where 40% of initial spend comes from 20 high-volume keywords. Post-click data shows half of them, 10 keywords, exceed the CPO/CPA objective. The rule is to scale those back by dropping their positions, or eliminate them. Meanwhile, keywords already under the CPO/CPA target are candidates for a boost. Because a higher position pulls a different mix of clickers, you re-check conversion after each change rather than assuming the old rate holds. You also avoid a blanket jump in spend. In the source's example, going from $12,000 to $15,000 a month takes traffic from competitors and costs more per click, so you only buy up the positions whose value to you clears the price. The competitor who bids what looks like a crazy amount may be valuing lifetime value and cross-sales that you are not counting.

• If a keyword's post-click CPO/CPA exceeds your objective, then lower its position or remove it. If it comes in under target, then raise the bid to buy more of that inventory.
• If the next position's clearing price is above your estimated value of that position to you, then do not match a competitor's bid. Let them have it, because bidding wars are destructive and only make sense up to your valuation.
• If you add volume and the per-click price climbs faster than conversion value, then stop at the point where cost per acquisition meets your target. Do not chase maximum ROI, since the highest-ROI campaign is rarely the profit-maximizing one.
• If you improve a landing page or ad and conversion rises, then raise bids in proportion, since the conversion lift is the leverage for setting them.
• If a competitor's bids look irrational, then assume they model lifetime value or the buying cycle, and check whether your own value model is too narrowly tied to the immediate sale.

• Maximizing ROI instead of profit. The tell: your best-converting listing sits at the minimum CPC with a great ratio but little volume, and total profit is flat.
• Search engine sales thinking: valuing only the immediate trackable sale. The tell: you lose bids to competitors you call crazy, and ignore later-cycle, after-cookie and offline sales.
• Escalating spend on emotion or against a rep's push. The tell: spend rises, such as from $12,000 to $15,000, with no matching rise in profit, and per-click cost climbs.
• Treating a position change as a pure volume dial. The tell: conversion cost shifts after you move a keyword up, because the network mix and clicker type changed.
• Setting a base bid too low where targeting data is missing. The tell: on MSN, non-targeted traffic dries up even though some of those unidentified searchers are in your sweet spot.

• ▸Keyword and engine: the listing under review.
• ▸Current position and max CPC: where you sit and what you bid now.
• ▸Value per conversion: your estimate of true value, including immediate profit and any measured lifetime value or cross-sell.
• ▸Target CPO/CPA or ROAS: the limit that decides keep, raise or cut.
• ▸Actual post-click CPO/CPA: measured cost per order or action at the current position.
• ▸Conversion rate after last test: the latest landing page or ad result, used to justify a bid change.
• ▸Cost of next position up: the extra per-click price and expected extra volume.
• ▸Targeting bid boosts: any age, gender or geography adjustments and the base bid for unknown searchers (e.g., MSN adCenter).
• ▸Action and re-check date: raise, hold, lower or eliminate, with a date to re-measure after the position change.

1. 1Baseline first: if paid campaigns already run, use their results as the baseline; if not, use organic-only traffic, conversion, and revenue/profit/lead-quality numbers. Record clicks, conversions, revenue/profit, and lead score (for lead-gen) per source before changing anything.
2. 2Run the pulse test: one week on, one week off, with the PPC listings paused in the off week. Watch all key post-click metrics for both organic and paid throughout.
3. 3Size the window to the data: if a week does not produce a statistically significant number of orders or leads, extend to a couple of weeks or longer before drawing conclusions.
4. 4Switch states and compute net click gain. Take the delta in total clicks, then the delta by source. Where the paid listing takes clicks from the organic listing, that is cannibalization.
5. 5Compute true cost per click: add the cost of the cannibalized clicks to the net gain of clicks, and calculate CPC on the net clicks the paid listing actually delivered, not on the clicks it reported.
6. 6Compare total conversions and revenue/profit/lead quality between the two states, separating organic from paid and then looking at the net gain. Check whether paid brings a different, more valuable customer.
7. 7Close the offline gap: tag customers by marrying the online cookie to the offline customer number or credit card data; issue unique offer codes for phone or in-store redemption; use unique online pricing as a tracking code that the searcher must request.
8. 8Attach the media source ID to customer IDs and have your tech team crunch the numbers. Repeat-purchase products (the passage cites adult diapers and insulin test kits) or listings with unusually large checkout totals may justify higher CPO targets.
9. 9Screen traffic for fraud signals: use IP address lookups, cookie tracking, session tracking, and normalized benchmarking. Repeat visits from competitors in server logs are a flag, not proof.

The passages' own test is the pulse test. Say your paid listing sits above a strong organic listing. Week one, both are live; you log clicks, orders, and revenue per source as the baseline. Week two, you pause the PPC listing and log the same numbers. The delta in total clicks is your net click gain; some of the paid clicks simply moved over from the organic listing, which is cannibalization. You add the cost of those cannibalized clicks to the net gain to get the true cost per net click. You then compare total conversions and revenue/profit/lead quality across the two states. Because you have more flexibility with ad copy and landing pages in PPC than in organic, you also account for conversion rate differences. If a week yields too few orders to be significant, you extend the test to a couple of weeks or longer. The passages do not supply numbers for this test, so the figures are yours to collect.

• If a week of pulse-test data does not produce a statistically significant number of orders or leads, then extend the test to a couple of weeks or longer rather than decide on thin data.
• If the paid listing cannibalizes organic clicks, then price the paid listing on net click gain, with cannibalized click costs added, not on gross paid clicks.
• If paid-sourced customers show higher lead scores, repeat purchases, or larger checkout totals, then raise CPO targets for those listings; if you can't see this, attach media source ID to customer IDs first.
• If a meaningful share of your sales closes offline (comScore's 2006 study for Google found 63% of purchases tied to a search query completed offline), then do not manage search on online-only conversions; add customer tagging, offer codes, or unique pricing.
• If a competitor with a holistic on/offline view bids against you while you count only online conversions, then expect to be priced out of the positions that give you scale; fix your measurement before cutting bids.

• Judging the test on too little data: a one-week test on a small site that declares a winner from a handful of orders. Tell: results swing sharply between runs.
• Using reported paid CPC instead of true CPC: ignoring cannibalized organic clicks. Tell: total clicks barely move when paid is paused while paid reports plenty of clicks.
• Online-only conversion counting for a multichannel business: Tell: low reported ROI on keywords that visitors use for research before buying in store or by phone.
• Merging organic and paid results into one number, missing that paid may bring a different, more valuable customer. Tell: the net gain looks flat while per-source quality scores or revenue differ.
• Treating suspicious repeat clicks as certain fraud (or ignoring them entirely): the author admits he can never know whether competitor visits were fraud or curiosity. Tell: conclusions drawn from IP lookups alone, without cookie, session, and benchmark checks.

• ▸Baseline source: state whether baseline is existing paid results or organic-only, with dates.
• ▸Test window: record start/end dates for the on and off weeks and note any extension for significance.
• ▸Clicks by source (on vs off): enter organic, paid, and total clicks for each state.
• ▸Net click gain and cannibalized clicks: compute the total-click delta and the organic clicks lost to the paid listing.
• ▸True CPC: paid spend plus cannibalized click cost, divided by net clicks gained.
• ▸Conversions and revenue/profit by source: enter both states, organic and paid, plus net gain.
• ▸Lead quality score (lead-gen): record average score per source in each state.
• ▸Offline tracking method: list customer tagging, offer code, or unique pricing in use and the redemption count.
• ▸Source-to-customer link and CPO target: note whether media source ID is attached to customer IDs and any CPO target changed as a result.