Web analytics demystified a marketers guide to understanding how your web site affects your business

1. 1Define the conversion as a specific, measurable goal tied to a business objective. The passages reject 'increase traffic' and endorse 'increase visitor traffic by 10 percent and visitor conversion by 50 percent from June to July by refining the marketing message in our email campaigns.' Name the action, the change you will make, the expected effect, and the period.
2. 2Sort the goal into one of three categories the source names: activities leading to an online purchase; activities leading to collection of valuable information, such as registration; or activities leading visitors to information that reduces operational costs, such as fewer calls to technical support. Do not assume a purchase is the only valid conversion.
3. 3Break the conversion into discrete steps. A single step may span more than one page. Use process measurement tools to record entries and completions at each step, so you can measure both the conversion rate and the abandonment rate.
4. 4Set up the goal in your analytics application. The source lists valid conversions: starting a download of a document or application, submitting information (lead generation), locating a support document or FAQ, navigating from general to specific information, spending a defined time or viewing a set number of pages, and viewing key pages such as pricing or a service agreement. Use multiple-goal flexibility if your tool offers it.
5. 5Take baseline measurements before you change anything, for example the visitors arriving from email campaigns in June. Roll them into a regular report shared across the organization, so everyone sees the baseline and later the effect of the change.
6. 6Measure more than one conversion rate. The source stresses that many conversion rates exist for any site, including the site-wide rate, and warns against fixating on a single rate. Report the rate for each defined goal alongside the site-wide rate.
7. 7Diagnose before you change. If response is low, the passages name two candidate explanations, poor targeting and poor messaging. List the candidates, prioritize them, change one thing, and record the expected effect.
8. 8For form-based conversions, audit the form fields. The source notes an anecdotal correlation between the number of fields and the likelihood of abandonment. Cut questions like 'How did you hear about us?' that invite garbage answers and make the form look too long, unless the offer is exceptional.
9. 9Judge success against your own prior rates. The source advises ignoring published averages and competitor figures, because they are not comparable unless you market the same way to the same audience with the same products and prices.

The source's own illustration is an email campaign goal. You define it as: raise visitor traffic 10 percent and visitor conversion 50 percent from June to July by refining the marketing message in email campaigns. This is specific enough to tell you which part of the site or marketing program changes (the email campaigns) and what effect to expect. You first measure June's baseline, meaning the visitors who arrive from email campaigns and the share who complete the defined activity, and put it in a shared report. If response is low, you consider the two explanations the source gives, poor targeting and poor messaging. You pick one, refine the message, and measure July against June using the same definition. Because you compare against your own baseline and not a competitor's published average, any movement is attributable to the change you made. If conversion misses 50 percent, you reprioritize and repeat the cycle.

• If the goal is only 'more traffic' or 'more conversions', then rewrite it with a target, the changed element, and a time period before you measure anything.
• If the conversion is a multi-step process, then define each step and measure abandonment per step, since one step can span several pages.
• If the conversion is a form submission and abandonment is high, then cut the optional or low-value fields first, unless the deliverable is exceptional enough to justify them.
• If the business is not a retailer, then still define conversions: lead submission, document download, registration, or a support-page visit that reduces support calls.
• If someone asks you to benchmark against published industry averages, then decline and track against your own baseline, because audience, products and prices will differ.

• Vague goals: the target is 'increase traffic', so you cannot tell which change caused a movement. The tell: after a change, no one can say what was expected.
• Measuring without acting: reports circulate but no change follows. The source cites ill-defined goals, poor information sharing, and inaction as the three reasons analytics fail. The tell: the same metrics appear unchanged month after month.
• Fixating on a single conversion rate: you track only the site-wide rate and miss step-level or goal-level problems. The tell: the overall rate is flat while individual funnels move in opposite directions.
• Treating only purchases as conversions: you ignore lead, download and support-deflection value. The tell: the analytics goal list contains a single entry.
• Bloated forms with 'nice to know' questions: you collect junk answers and abandonment rises. The tell: high form starts, low completions, and free-text fields filled with meaningless entries.

• ▸Conversion action: state the exact visitor action (purchase, lead submission, download start, registration, key page view).
• ▸Business objective served: name the core objective it supports, such as revenue, lead flow, or lower support cost.
• ▸Category: choose purchase-related, valuable information collection, or operational cost reduction.
• ▸Process steps: list each discrete step from entry to completion, noting pages in each step.
• ▸Baseline measurement: record current conversion and abandonment rates per step, with the period measured.
• ▸Planned change: describe the single change you will make (message, targeting, form fields).
• ▸Expected effect and period: write the target (for example +50% conversions, June to July).
• ▸Hypothesis for current shortfall: list candidate causes, such as poor targeting or poor messaging, in priority order.
• ▸Report owner and audience: name who publishes the shared report and who receives it.

1. 1Place retention as the fourth stage of the lifecycle the book uses: reach, acquisition, conversion, retention. Retention starts after a visitor has converted or at least visited once. Don't count it as more acquisition. Write one sentence stating who counts as 'retained' for your business, because the source passages give only section titles for Chapter 14, not the definitions.
2. 2Split every traffic and order report into new visitors and returning visitors. The author's own entry test is whether you have thought about the difference between a 'new visitor' and a 'returning visitor'. If your tool cannot make that split, retention is unmeasurable, so fix that first.
3. 3For the visitors you classify as retained, measure how often they come back. Chapter 14 lists 'Average Frequency of Return for Retained Visitors' as its own metric. Report it only for retained visitors, so first-time traffic does not dilute it. Track the trend per period, not the one-off number.
4. 4Tie retention to money using the two conversion-chapter metrics: 'Percent of Orders from New and Repeat Customers' and the 'New and Repeat Customer Conversion Ratio'. These show whether returning people are producing orders and whether they convert better than newcomers.
5. 5Apply visitor segmentation to the retained group. Chapter 14 includes 'Visitor Segmentation Metrics for Retained Visitors', and Chapter 7 covers segmentation tools. Compare retained visitors with the rest by source, behavior and value, to see which kinds of visitors come back.
6. 6Run and track every campaign aimed at bringing people back. The book's tip is to track every campaign you run, carefully. Give return-oriented campaigns their own tracking so response can be attributed to retention work, not to acquisition.
7. 7Decide deliberately whether to require registration. The book has a web analytics tip on when to require registration, and identifying visitors is what lets you recognize them when they return. Weigh that measurement benefit against the friction it puts in front of first-time visitors.
8. 8Carry retention into the Chapter 15 KPI worksheet. Select the retention KPIs the book recommends for your business model (commerce, advertising, lead generation or customer support), and review them alongside conversion KPIs on a fixed schedule.

The passages contain one explicit retention action from the buyer's side. When selecting a page-tagging analytics vendor, the author advises asking what support and training the vendor provides, how often, and at what cost. He notes that strong vendors give free training and support, at least on a limited basis, 'to improve customer retention and satisfaction'. If training is only available for a fee, you ask what they provide to most other customers and have it included at no extra charge in the contract negotiation. The vendor is applying the retention logic itself: it invests in the customer after the sale so the customer stays. You can copy the move for your own site. First identify what makes post-sale value tangible (training, support, follow-up). Then offer it to converted customers up front instead of selling it afterward. Then check the effect in the Chapter 14 and 13 metrics: returning visitors, their frequency of return, and the share of orders from repeat customers. The passages give no numeric benchmarks, so use your own baseline period as the comparison.

• If your analytics tool cannot separate new from returning visitors, then fix identification before reporting any retention figure.
• If repeat customers convert at a lower rate than new customers (New and Repeat Customer Conversion Ratio), then examine what the returning experience offers before spending on more acquisition.
• If frequency of return for retained visitors is falling while the retained count holds steady, then returning visitors are cooling. Segment them and target the group that is dropping off.
• If a campaign targets existing customers, then track it separately from acquisition campaigns, so its response is not mixed into reach and acquisition numbers.
• If you are considering mandatory registration, then first check what you lose in first-visit conversion against what you gain in recognizing returners.

• Treating total traffic growth as retention. The tell: visits rise but the returning-visitor share and frequency of return are flat or unreported.
• Averaging frequency of return across all visitors. The tell: a low, unmoving number that never reflects the retained group, because first-time visitors drag it down.
• Reading repeat orders without a conversion ratio. The tell: the percent of orders from repeat customers rises only because new-customer orders fell.
• Mixing acquisition and retention campaign tracking. The tell: you cannot say which campaign brought a lapsed visitor back.
• Forcing registration too early. The tell: identified returning visitors are easy to count, but first-visit conversion drops.

• ▸Retained visitor definition: state in one sentence who counts as retained for your business and over what window.
• ▸Returning vs new visitors (period): record both counts and the returning share, from your tool's new/returning split.
• ▸Average frequency of return (retained only): record the figure for retained visitors, excluding first-time visitors.
• ▸Percent of orders from repeat customers: record the share of orders from repeat versus new customers.
• ▸New vs repeat customer conversion ratio: record both conversion rates side by side.
• ▸Retained-visitor segments: list the segments (source, behavior, value) and which are growing or shrinking.
• ▸Return campaigns tracked: list each campaign aimed at existing visitors and its tracked response.
• ▸Registration decision: note whether registration is required, the reason, and the observed effect on first-visit conversion.
• ▸Review cadence and owner: name who reviews these KPIs, how often, and which Chapter 15 KPI worksheet row they sit in.

1. 1Name your business model first: online commerce, advertising, lead generation, or customer support. The source organizes KPIs by model, and ROI means a different thing in each. Write one sentence stating what 'return' is for your site (a completed purchase, a qualified lead, an ad impression, a resolved support need).
2. 2Lay the metrics along the customer life cycle: reach, acquisition, conversion, retention. List the KPI you already track at each stage. An empty stage means you cannot trace spend to return, so fill it before you report ROI.
3. 3Define the cost side next to the return side. ROI is net value to the bottom line, so record the site and marketing investment for the same period. Without a cost figure you have only revenue, not return.
4. 4Measure the conversion process step by step. For commerce, track cart conversion (completed carts ÷ cart starts) and checkout conversion (completed carts ÷ checkout clicks). Add revenue per visitor, revenue per customer, items per order, and browse-to-buy by product, category, and brand. For lead generation, use process measurement to see where each form step loses visitors.
5. 5Find the biggest leak and fix only that. In the source's example, 50 percent abandonment between 'form 2' and 'form 3' sends you back to form 2 to remove every field that is not absolutely required. Redeploy, then measure abandonment at that step and the overall conversion rate.
6. 6Connect search and traffic data to money. Credit internal search terms, including terms from earlier visits, when the purchase finally happens. Compute browse-to-buy per term to learn which searches are most valuable financially, which moves you from the average visitor's intent to the average customer's intent.
7. 7Feed the result back into merchandising and content. Feature products, categories, and brands that carry higher browse-to-buy or cart-to-purchase ratios on key pages. Retire or fix pages that waste traffic, such as heavily requested error pages, which you should redirect to the correct or a closely related page.
8. 8Re-run the same KPI worksheet each period and compare to the prior one. Practice continuous improvement: change one thing, deploy, measure the effect on the step metric and on the overall conversion rate, then report the net change against the investment.

A lead generation site finds, in its process measurement report, 50 percent abandonment between 'form 2' and 'form 3'. A continuous-improvement team re-examines form 2 and asks of every field whether the business absolutely needs it. Fields like 'How did you hear about us?' and 'How many people do you supervise?' are easy to answer with garbage and make the form look too long, so they go. Anything that could lead to identity theft, such as a social security number or mother's maiden name, is never requested, because the site is not a financial institution. If many required fields remain, the team splits the form into a small number of logical pages, each quick to load and labeled with progress such as 'Step 2 of 4'. It adds client-side validation so an error does not clear fields such as passwords. After deploying, the team measures abandonment at that step and the overall conversion rate. The ROI case rests on that overall rate: the same marketing spend now produces more completed leads. The return is that gain net of the cost of the change and of the traffic.

• If the step-level abandonment report shows one step losing a large share of visitors (the source's example is 50 percent between two forms), then fix that step before adding traffic. More acquisition spend just feeds the leak.
• If a form asks for anything not absolutely required, then remove it. If many required fields remain, split them into short, labeled pages.
• If a product, category, or brand has a high browse-to-buy ratio, then feature it on key pages. If it is browsed often but rarely bought, then investigate the product page and the purchase process.
• If an internal search term leads to purchases, then treat it as financially valuable and track it across visits until the sale. If a search term attracts volume but no purchases, then review what the visitor finds.
• If a page is requested in error with significant traffic, then redirect it to the correct or a closely related page, and correct the linking page or ask the other site's owner to fix the link.

• Reporting revenue as ROI. The tell: no investment or cost figure appears anywhere in the report, so you cannot show net value.
• Tracking traffic metrics (hits, page views, visitors) and calling them return. The tell: your dashboard has no conversion, revenue-per-visitor, or customer-value line.
• Crediting a purchase only to the last search term or visit. The tell: terms entered in earlier visits never appear in your purchase attribution, so the most valuable searches look worthless.
• Measuring overall conversion without process-step conversion. The tell: you know the rate fell but cannot say between which steps, so fixes are guesses.
• Adding questions to forms for marketing curiosity. The tell: abandonment climbs at the form step while fields like 'How did you hear about us?' or intrusive identity data remain.

• ▸Business model: commerce, advertising, lead generation, or customer support, and the one-sentence definition of 'return' for it.
• ▸Period and investment: the dates covered and total site plus marketing cost for that period.
• ▸Reach KPI: your current reach measure and its value.
• ▸Acquisition KPI: your current acquisition measure and its value.
• ▸Conversion KPIs: cart conversion, checkout conversion, overall conversion, revenue per visitor, revenue per customer, or the equivalent for your model.
• ▸Biggest leak: the process step with the highest abandonment, and its rate.
• ▸Top search terms to purchase: the internal search terms with the best browse-to-buy ratio, tracked across visits.
• ▸Change deployed: the single fix made, and the date.
• ▸Result: step conversion and overall conversion before and after, plus net return against investment.

1. 1Write one specific goal per marketing effort before launch. The source rejects 'increase traffic to our Web site' because you can never be sure what moved it. Use its model: 'increase visitor traffic by 10 percent and visitor conversion by 50 percent from June to July by refining the marketing message in our email campaigns.' That names the channel, the change, the metric, the size and the window.
2. 2Name the single thing you will change. In the source's example only the email marketing message changes, so any shift in traffic or conversion has one plausible cause. If you change several things at once, you can't say which one worked.
3. 3Take baseline measurements before you change anything. Measure visitors arriving from that channel in the prior period (the source's example uses June email traffic). Without a baseline you can't tell whether a result is an improvement.
4. 4Tag every effort to a business objective. BackcountryStore.com sorts every page into six objectives. Two are reach objectives: 'Introduce BackcountryStore.com to new visitors' and 'Gather leads for their online marketing group.' Group the landing pages each campaign feeds under one of these, so you can compare page views and unique visitors per group before and after a change.
5. 5For SEO, measure crawler behavior as well as human visits. Log-file tools report which robots and spiders visit and how often. The goal of SEO is to attract more indexing agents and have them traverse the site deeply and broadly, so compare that activity before and after the work. This matters most when SEO is outsourced, because it lets you quantify what you paid for.
6. 6Roll the measurements into a regular report and share it across the organization. The source calls this critical so that everyone understands the change you are about to make and can see its effect. Include the baseline, the target and the actual result.
7. 7Act on the result, then repeat. After the test window, compare actual to expected. Keep, refine or drop the change, and set the next goal. Sterne's point in the source is that measurement alone isn't enough; the results must feed back into sales and marketing processes.
8. 8If you need to tie campaign activity to visitors and customers, plan for data integration. The source shows vendors partnering with email providers (Coremetrics with CheetahMail and Digital Impact) and affiliate platforms (Coremetrics with BeFree) so visitors and customers can be linked by a common identifier. Decide whether that matters at your stage; the source says it is either irrelevant or critical.

The source's own model campaign is an email effort. You define the goal as a 10 percent increase in visitor traffic and a 50 percent increase in visitor conversion from June to July, by refining the marketing message in your email campaigns. You take a baseline first: the visitors who arrived from email activity in June. You change only the email message, and you report the numbers in a regular report that people across the company can read. Because only the message changed, any difference between June and July points back to it. Then you compare the result to the two targets. In the same spirit, a retailer like BackcountryStore.com would file the landing pages that email feeds under its 'Introduce BackcountryStore.com to new visitors' group, so the before and after view comes from the same content group. That group view is the same one that showed 220 daily unique visitors generating 888 page views on Men's Ski Jackets, nearly four views per visitor.

• If a goal reads 'increase traffic,' then rewrite it with a channel, a change, a percentage and a date range before spending anything.
• If you want to know whether a message or creative change worked, then change only that element and keep the baseline period's measurement to compare against.
• If SEO work is outsourced, then require crawler data (which spiders visit, how often, how deeply) as evidence, not just a ranking claim.
• If a report is generated but nobody changes anything, then stop adding metrics. The source says measurement without thoughtful action is wasteful. Fix sharing and action first.
• If you need campaign visitors linked to later customer behavior, then check whether your analytics and email or affiliate vendors share a common customer identifier before you commit to a stack.

• Vague goals: the objective is 'more traffic' or 'more awareness.' The tell is that after the campaign nobody can say whether it worked or why.
• Changing many things at once: new message, new design and new channel in one month. The tell is a traffic bump nobody can attribute, which is Leverhulme's 'I'm just not sure which half.'
• No baseline: you launch first and measure later. The tell is that you can't state last period's number for the same channel.
• Reports that don't circulate: data stays with the analyst. This is one of Burby's three reasons analytics fail (information is ineffectively shared). The tell is that campaign decisions are made without referring to the report.
• Collecting without acting: this is another of Burby's reasons (companies don't take action on the data). The tell is dashboards full of reach numbers and no record of a change made because of them.

• ▸Campaign / channel: name the single channel (e.g., email, SEO, PPC, offline) and the effort.
• ▸Business objective served: which objective group this feeds (e.g., 'Introduce us to new visitors' or 'Gather leads').
• ▸Goal statement: one sentence with metric, percentage change, and date window.
• ▸Single change being made: the one element you will alter (e.g., the email message). Do not list more than one.
• ▸Baseline measurement: the prior-period number for this channel, with dates and source tool.
• ▸Crawler / SEO indicator (if SEO): spider visit frequency and depth of site traversal before the work.
• ▸Report owner and audience: who produces the regular report and which people in the organization receive it.
• ▸Result vs. target: the actual figure at the end of the window next to the goal.
• ▸Next action: keep, refine or drop the change, and the next goal.

1. 1Weigh your home page and three or four key pages first. Page weight is the total kilobyte size of the text, script, images, multimedia files and banner ads that the browser reassembles. Treat the result as a baseline for likelihood of success: heavy pages plus other design problems means real trouble, and light pages mean you can worry less about the rest.
2. 2Test download time against Nielsen's three limits. 0.1 second feels immediate, 1 second keeps the user's flow uninterrupted, and 10 seconds is the upper limit for keeping focus on one task. Past 10 seconds users start side tasks or get frustrated. Flag any key page that risks crossing 10 seconds.
3. 3Monitor page performance while you build new pages, not after launch. The source suggests an inexpensive service such as NetMechanic.
4. 4Make analytics reports readable for navigation review. Cryptic directory and file names ("invtinfo", "default.html") force a translation step that slows analysis. Map pages to content groups by function, for example /shoes/availability, /shoes/pricing, /shoes/compare. For dynamic pages driven by query strings such as page.asp?pid=1234&cid=8&vid=41A, build the mapping from parameter values back to what is actually displayed.
5. 5Define the process you care about (checkout) as an ordered series of steps, and apply process measurement to it. Record the count and percentage of visitors who reach each step, so you can see where abandonment concentrates.
6. 6Inspect the step with the largest drop. Ask whether it is a required action that stands between the visitor and the goal, such as forced registration before shipping and billing information. Then decide whether to remove it or offer an optional path instead.
7. 7Ship the fix together with any other usability improvements the analytics identified. Re-measure site-wide conversion before and after, and separately during peak periods such as sales.
8. 8Run usability studies with likely visitors or actual customers alongside the quantitative data. The source calls them a necessary component of a Web analytics program, though a science of their own.

BackcountryStore.com, an outdoor-gear retailer, sent shoppers who clicked "Checkout Now" to a required registration page (create an account, or log in) instead of the first step, shipping and billing information. Process measurement tools showed significant abandonment at that registration step. The funnel counts in the source's figure are partly garbled, but they show roughly 48% and 38% of visitors at successive stages and about 34% at a later one. Requiring registration is not always a bad idea, so the data made the call, not a rule. BackcountryStore.com removed the step and gave customers an optional route to register or log in, a "Returning customer login, please click here" link on the checkout page. Together with a small handful of other usability improvements found through Web analytics, the change produced a 20 percent increase in site-wide conversion rate. During the twice-yearly sales, the lift is estimated at as much as 30 percent versus pre-change numbers.

• If page weights are high and you find other design problems, then treat the site as at real risk and fix weight first. If weights are low, worry less about the other problems.
• If a key page could take 10 seconds or more to load, then expect visitors to switch to side tasks or abandon, and cut page weight before anything else.
• If process measurement shows heavy abandonment at one step, then examine that step for friction. Forced registration is the example, so make it optional and keep a login path for returning customers.
• If URLs or file names are cryptic or generated from query strings, then build content groups that name what the page shows (/shoes/pricing) before you analyze navigation.
• If a fix bundles several changes, then credit the combined result, not the single change. The source attributes the 20 percent lift to registration removal plus a few other usability improvements.

• Judging design without a performance baseline. The tell: you debate layout and navigation but nobody can state the weight of the home page or key pages.
• Treating a general best practice as a verdict. The tell: you remove or keep registration by opinion instead of reading abandonment at that step.
• Reporting on raw URLs and file names. The tell: analysts spend time translating codes like "invtinfo" or "default.html" and reports stay unreadable to the business.
• Measuring only overall conversion. The tell: you see a flat site-wide rate but cannot say which checkout step loses people, so you miss that a step is the culprit.
• Relying on analytics alone and skipping usability studies. The tell: you know where visitors leave but not why, because no one has watched likely visitors or customers use the site.

• ▸Page and URL: list the home page and three or four key pages being weighed.
• ▸Page weight (KB): record the total size of each page's components.
• ▸Estimated load time: note whether each page is under 1 second, under 10 seconds, or at risk of 10 seconds or more.
• ▸Content group name: give each page a readable group (e.g., /shoes/pricing) in place of codes or query strings.
• ▸Process steps in order: list each checkout step from cart to receipt.
• ▸Visitors and % at each step: enter counts and percentages from process measurement.
• ▸Largest-drop step and suspected cause: name the step losing the most visitors and what it demands (e.g., required registration).
• ▸Change made and date: describe the fix and when it shipped.
• ▸Conversion before/after (site-wide and peak periods): record both rates to measure the lift.

1. 1Define the outcome as a visit. Visitor acquisition is a visitor arriving at the site from a marketing channel, the step after reach and before conversion in the reach, acquisition, conversion, retention life cycle. Count visits or visitors. Do not count hits or page views as acquisition.
2. 2Capture the referrer as a complete, undifferentiated URL. The source defines it as the exact page on the referring site that contained the link, so store the URI and the query string. Without both you cannot reconstruct the referring page.
3. 3Report referring domains individually, with no grouping or categorization. The source warns that grouping loses granularity and should be avoided if possible. Keep any roll-up in a separate view, never in the base referrer report.
4. 4Handle missing referrers explicitly. A request with no referring source is reported as "unknown referrer" or left out of the referrer report. Never reassign it to a channel you prefer.
5. 5Report referring sources as accurately and without bias as possible. The source calls this information critical to measuring both reach and acquisition (Chapters 11 and 12). Do not let a channel owner adjust the numbers.
6. 6Split new visitors from returning visitors before judging acquisition, and look at the top pages and content they request. Chapter 12 treats these as acquisition metrics: pages requested by new visitors, and navigation paths by referring source.
7. 7Track every campaign you run, carefully, so each campaign's visits can be tied to its referrer. Chapter 8 flags this as a core campaign analysis tip.
8. 8Apply continuous improvement. Before changing a landing page, link, or campaign, write down the reason and the effect you expect on the acquisition indicators. Afterward, compare the result to that expectation and act on it.

The source has no acquisition-specific data set, so this example adapts its segmentation passage, which compares two ways of reporting visitors to content group "A" in one week. The raw report reads 27,000 visitors and 43,000 page views. The segmented report reads 8,000 visitors in the male 18-to-25 segment, generating 15,000 page views. To apply this to acquisition, tag each of the 27,000 visits with its complete referring URL and report "unknown referrer" where none exists. Then cut the 8,000 target-segment visits by referring source. You can now see which referring pages brought the audience you wanted, not just the most visits. Count visits or visitors, not page views, as the acquisition number. The 43,000 page-view figure describes activity after arrival and is not itself acquisition. The comparison shows why the raw total hides which sources reach the targeted segment.

• If a request has no referring source, then label it "unknown referrer" or omit it from the referrer report. Do not guess a channel.
• If a report groups referring domains into categories, then also pull the individual domains and complete URLs. Grouping loses granularity and should be avoided if possible.
• If you are computing a rate for a campaign or source, then use visits or visitors as the unit. Use views only when nothing else is available, and even then it is not a good idea.
• If you are about to change a page, link, or campaign, then record the reason and the expected acquisition effect first. Without both, you cannot judge the change afterward.
• If a total visitor count looks healthy, then segment it by who arrived and from where before reporting. The aggregate can hide whether the target audience came.

• Counting hits or page views as acquisition. The tell: the acquisition number moves when site navigation changes, not when channel traffic changes.
• Reporting only referring domains, or grouped categories. The tell: you cannot say which exact page carried the link, and the query string is absent from the report.
• Filling unknown referrers into a favored channel. The tell: one channel's numbers rise while the unknown share quietly shrinks, with no change in marketing activity.
• Running campaigns without tracking each one. The tell: you cannot attribute a visit spike to a specific campaign afterward.
• Measuring without acting. The source cites undefined goals, poor sharing of results, and no action on data as the reasons analytics fail. The tell: reports are produced each week, but no acquisition change has a recorded reason or expected effect.

• ▸Period and channel/campaign name: Identify the campaign and dates so each campaign is tracked separately.
• ▸Unit counted (visits or visitors): State which one. Do not use page views.
• ▸Complete referring URL (URI + query string): Record the exact page that held the link, not just the domain.
• ▸Referring domain (individual, ungrouped): List each domain on its own line with no categorization.
• ▸Unknown referrer count: Record requests with no referring source, shown separately.
• ▸New vs. returning visitors: Split the acquired visits so new-visitor acquisition is visible.
• ▸Top pages requested by new visitors: List the first content these arrivals request.
• ▸Target segment visits (if segmented): Record visits from the audience you wanted versus the total.
• ▸Change made, reason, expected effect: Write these before the change, then record the observed effect and the next action.

1. 1Fix your definitions first. Write down what a page view, a visit and a unique visitor mean in your tool. The source says vendors define these terms differently, and engagement ratios built on mismatched definitions mislead. A visit is a period of activity separated by 30 minutes of inactivity.
2. 2Decide what a 'page' is on your site. If one template serves many products, such as one Active Server Pages template that can define 100 products, choose whether you count 100 views of the template or each product view as its own page view. Use the product-level count, because it gives the detail managers need. Keep the choice constant over time.
3. 3Pick engagement signals from what the passages name as measurable: how long visitors stayed on a page or on the whole site, which links they clicked, which pages they viewed in sequence (top paths), and when they left. Add page views per visit as a derived ratio, computed from your own page-view and visit counts.
4. 4Split every engagement number by visitor type: new vs. returning visitors, and by any form-based segment you collect, such as demographics. An aggregate hides the difference. For example, a targeted segment can be small yet generate more page views per visitor than the total.
5. 5Check page weight and download time before you read low engagement as a content problem. Measure the home page and three or four key pages. Heavy pages plus design problems predict abandonment, so slow pages can depress engagement with good content.
6. 6Check the visit boundary. A pause of more than 30 minutes splits one real session into two visits. Look at visits that start with a cart or form already begun and ask whether the visitor is just resuming.
7. 7Compare segments and content groups against each other, and against your own prior periods, in the same time window. The passages give no benchmarks, so your own history is the only baseline.
8. 8Log every change you make to content or design and its effect on the engagement metrics. Assign one gatekeeper for this log so the same change is not repeated and judged again.

The source gives a visit-boundary case that matters for engagement. A visitor arrives at 1:00 PM and starts putting items in a cart. At 1:15 PM the phone rings and the visitor does not click for 31 minutes. At 1:46 PM they return and finish the purchase by 1:57 PM. Because inactivity exceeded 30 minutes, the tool records two visits: 1:00 to 1:15 and 1:46 to 1:57. One highly engaged person therefore looks like two shorter visits, and visit-duration measures understate their engagement. The source also gives a segmentation example: an advertiser could hear '8,000 visitors in the male 18-to-25 segment generated 15,000 page views' instead of '27,000 visitors generated 43,000 page views'. Dividing those figures (our arithmetic, not the source's) gives about 1.9 page views per visitor for the segment and about 1.6 for the whole group. The segment looks more engaged than the aggregate, and only the segmented view shows it. The way to read both cases is to confirm the definitions of visit and page view first, then compare segments rather than totals.

• If engagement is low and key pages have high page weights, then fix download time before revising content, because slow pages push users to abandon.
• If a visitor's activity pauses just over 30 minutes and then resumes, then treat the two visits as potentially one engaged session before you judge visit counts or duration.
• If you report engagement to someone who cares about a specific audience, then segment by form-based data such as demographics instead of quoting total visitors.
• If someone quotes a high click-through rate as engagement, then ask what share of those clicks reached the site and how many did something important.
• If a metric changes after a content or design change, then check your change log and wait long enough that the result is likely to persist before concluding it worked.

• Confusing hits, page views, visits and unique visitors. The tell is a claim like 'we got X hits last month' presented as visitor interest.
• Counting template views instead of product or content views on dynamic sites. The tell is page-view totals that look right but every product page shows the same flat count.
• Blaming content for weak engagement when pages are slow. The tell is high page weights on the home page and key pages alongside high exit or abandonment.
• Reading only aggregate totals. The tell is a report with total visitors and page views and no segment, new-vs-returning or content-group breakdown.
• Repeating past changes and mistaking their effect. The tell is someone asking 'isn't this the same change we made last year?' with no log to answer.

• ▸Period and content group: state the dates and the section or page group being measured.
• ▸Definitions in use: write how your tool defines page view, visit (30-minute timeout) and unique visitor.
• ▸Segment: name the visitor group, such as new, returning or a demographic form-based segment.
• ▸Visitors / visits / page views: enter the three raw counts for this segment.
• ▸Page views per visit or per visitor: calculate and record the ratio, stating which denominator you used.
• ▸Time on page or site and top paths: note how long visitors stayed and which sequences of pages they followed.
• ▸Page weight and download time of key pages: record home page and three or four key pages.
• ▸Comparison baseline: enter the same figures for the prior period or another segment.
• ▸Change log entry and gatekeeper: record any change made, its date, the observed effect and who owns the record.

1. 1Classify the process into one of the three conversion categories the source names: activities leading to an online purchase, activities leading to the collection of valuable information (such as registration), or activities leading the visitor to information that reduces operational costs (such as support calls).
2. 2Define the process as a series of discrete steps. A single step may span more than one page, so list each step with every page that belongs to it. Example FAQ flow: support home page, page of support categories, page of commonly asked questions, page with the specific answer.
3. 3Configure your process measurement tool (multiple goal-page flexibility in your analytics application) with the ordered steps and the final goal page, so it reports step-to-step abandonment and overall completion.
4. 4Record a baseline for each step transition and for end-to-end completion before changing anything. Treat these as several distinct conversion rates, not one site-wide number.
5. 5Locate the step with the largest drop-off and change one thing there, following the source's form rules: ask only for information that is absolutely required, with no optional fields.
6. 6On forms, drop requests for personally identifiable information unless you are a trusted brand and financial institution (and then only after the visitor intentionally starts a process). If required fields are numerous, split the form into a small number of logically grouped pages, keep each page quick to load, and show progress such as 'Step 2 of 4'.
7. 7Use client-side form validation so errors do not force a page reload, which saves time and avoids clearing other required fields (password fields most often).
8. 8Re-measure the same transitions after the change. Judge success against your own prior baseline, not published averages or competitor figures. Repeat as part of the continuous improvement cycle.

The source's FAQ model illustrates the setup. A support site defines a four-step sequential flow: support home page, page listing support categories, page listing commonly asked questions, page containing a specific answer. You set the answer page as the goal and measure how many visitors move from each step to the next. The reason to measure is that self-service via the Internet is, by the source's account, the most cost-effective way to support customers, so each visitor who reaches an answer is a potential avoided support call. Suppose the categories-to-questions transition shows the heaviest abandonment. Your next move is to change that one step, then compare the same transition afterward to your own baseline. The same method applies to a registration form: remove non-required fields such as 'How did you hear about us?', split a long form into grouped pages labeled 'Step 2 of 4', and add client-side validation. The source offers no numeric results for these changes, so your measured before/after difference is the evidence.

• If a field is not absolutely required, remove it: the source ties more form fields to higher abandonment, and says to skip optional information unless the offer is exceptional and free.
• If the required field count is large, split the form into a few logically grouped pages with fast loads and a visible 'Step X of Y' indicator.
• If you are not a trusted brand and a financial institution, do not request personally identifiable information that could enable identity theft; if you are a financial institution, ask only after the visitor intentionally starts the process.
• If visitors see errors after submitting, switch to client-side validation so the page does not reload and clear fields.
• If you are tempted to compare your rate to published averages or competitors, do not: unless audience, marketing, products and prices match, the comparison is meaningless. Compare against your own baseline.

• Tracking one blended conversion rate and treating it as the answer. The tell: the number moves but you cannot say which step caused it. The source says there are many conversion rates and warns against fixating on one.
• Defining steps by page rather than by logical step. The tell: the funnel report shows phantom drop-offs between pages that belong to one step.
• Keeping marketing-driven questions on the form ('How did you hear about us?', 'How many people do you supervise?'). The tell: high abandonment on the form page and junk answers from those who do finish.
• Using server-side validation that reloads the page. The tell: complaints or drop-offs right after submission, with password or other fields cleared.
• Measuring without acting, or benchmarking against competitors' published figures. The tell: reports are produced each period but no step has been changed, or targets are set from industry averages.

• ▸Process name and category: purchase, valuable-information collection (e.g., registration), or cost-reducing information (e.g., FAQ/support).
• ▸Ordered steps: list each discrete step and every page that belongs to it.
• ▸Goal page: the final page that marks successful completion.
• ▸Baseline step-to-step rates: record the continuation (or abandonment) rate for each transition before any change.
• ▸Baseline overall completion rate: visitors who finish divided by visitors who start.
• ▸Largest drop-off step: the transition with the highest abandonment.
• ▸Change to test: one specific change (remove a non-required field, split the form, add 'Step X of Y', add client-side validation).
• ▸Post-change rates: the same transitions measured after the change, over a comparable period.
• ▸Next action: keep, revert or iterate, with the date of the next review.