Strategic database marketing the masterplan for starting and managing a profitable, customer-based marketing program

1. 1Compute retention rate as the percentage of this period's customers who purchase again next period; track it period over period rather than as a single snapshot, since database marketing builds each initiative on prior contacts.
2. 2Feed the retention rate directly into the lifetime value calculation — lifetime value is the net present value of profit from the average customer over several years, and retention is one of its revenue inputs.
3. 3Set the measurement horizon by industry: measure over more years for a utility (long relationships) and fewer for a dress manufacturer (short ones), because the period over which lifetime value is calculated varies with product and industry.
4. 4Before spending serious money, build a lifetime value spreadsheet that shows both the immediate and long-term impact of a retention strategy on profit, so you can calculate whether the strategy could work before testing.
5. 5Design matched test and control groups: the test group receives the retention tactic (membership cards, newsletters, special benefits), the exactly matched control group receives nothing, and you compare purchasing activity to isolate the effect on retention.
6. 6Pair retention moves with direct-cost moves where possible — proactive database marketing in insurance, credit card, and cellular phone companies improved retention while also reducing direct costs, so model both effects together.
7. 7Use RFM analysis to target retention offers at customers with purchase history, since it is more accurate and less expensive than modeling; fall back to demographic modeling only for prospects or where purchase history is missing.
8. 8After executing, re-measure the retention rate against the control group and recompute lifetime value to confirm the strategy actually raised the long-term asset value.

Take the gift service strategy aimed at Lydia McCabe. The stated problem was increasing the customer retention rate and annual sales to existing customers in a cost-effective way. The solution used a database to open a new way of selling products and services that already exist, by providing a service that did not previously exist for the customer — one that saves her time and builds loyalty to the store at the same time. Because the offer saves time and remembers her, Lydia keeps coming back, which raises the retention rate rather than relying on discounts to hold her. That retained behavior then feeds two of the five lifetime-value levers at once: higher retention and higher sales volume to an existing, loyal customer. You would validate it by issuing it to a test group, holding a matched control group, and comparing repeat purchasing to see the precise effect on retention before rolling it out.

• If you have an extensive purchase history on a customer, then use RFM analysis to target retention offers — it pays dividends greater than the effort; if you lack history (prospects), then rely on demographic modeling and test the file first.
• If a strategy's lifetime value chart shows positive long-term impact, then test it with control groups before full spend; if the chart doesn't, don't test it.
• If you can change the channel or affect the renewal process (as software companies selling upgrades directly retain 90% of revenue vs. ~50% through retailers), then pursue retention tactics that cut direct costs simultaneously.
• If customers are loyal, then make upgrade, companion-product, and new-product offers — loyal customers buy more — timing the right offer to the right people at the right time.
• If a tactic can't be measured against a matched control group, then treat its claimed effect on retention as unproven.

• Trying to hold customers with discounts instead of real value — the tell: retention barely moves while margins erode, versus the gift/grocery approach that kept customers through personal service and memory of prior transactions.
• Measuring retention as a one-off number instead of tracking attrition period over period — the tell: you can't show the increase feeding lifetime value, so the project looks unprofitable and gets cut.
• Building the database but having no strategy to keep it going and reduce attrition — the tell: a project touted as a new trend 'sinks like a stone months after the accolades,' the central fault behind wasted millions.
• Rolling out retention tactics without matched control groups — the tell: you claim a sales lift you can't attribute, because only comparing test vs. control reveals the precise effect on retention.
• Applying demographic modeling where RFM would work, or trusting appended demographic data without testing — the tell: predictions that 'might not be any help at all' because the data is seldom accurate and the file was never tested for applicability.

• ▸Measurement period: define the period over which repeat purchase is counted
• ▸Industry horizon (years): set number of years for lifetime value based on your product (longer = utility-like, shorter = dress-like)
• ▸Current retention rate: % of this period's customers who purchased again next period
• ▸Target retention rate: the rate the proposed strategy aims to achieve
• ▸Retention tactic: name the database-driven value offered (e.g., service, memory of transactions, membership benefits)
• ▸Direct-cost effect: note any channel/renewal change that also lowers cost per retained customer
• ▸Sales-volume effect: list upgrades/companion/new products loyal customers will buy
• ▸Test vs. control design: define the test group tactic and the exactly matched control group
• ▸Lifetime value impact: immediate and long-term NPV of profit from the raised retention rate

1. 1Define the referral rate as a measurable variable inside your lifetime value model, not a vague goal — it is one of the five levers (alongside retention rate, sales volume, direct costs, marketing costs) by which database marketing augments lifetime value.
2. 2Store the customer data that lets you identify and motivate referrers: purchase behavior (dates, dollars, products, responses), demographics, location, and opinions from surveys and complaints. You cannot enlist loyal customers in referral programs without knowing who they are.
3. 3Enroll your loyal, high-value customers in a formal referral program built around a system of rewards and motivations — the passages state carefully crafted strategies can influence word-of-mouth and expand the referral rate.
4. 4Set up exactly matched control and test groups: the test group receives the referral promotion (offers, membership cards, newsletters), the control group receives nothing. Compare purchasing activity to isolate the referral effect.
5. 5Measure the program's effect on referral rates directly through group comparison — this type of analysis can be done only with a database.
6. 6Run the numbers through a lifetime value spreadsheet before full rollout, charting both immediate and long-term impact, so you know whether the strategy could work before serious money is spent.
7. 7Budget for three years. Referral-driven loyalty effects take two to three years minimum to affect customer behavior, so secure the long-range budget and bank short-run wins to defend it.

Take the membership-card process the book describes and apply it to referrals. You issue membership cards to your customers, track sales to cardholders, and give special benefits to members who use their cards. Within that tracked population you can identify your loyal, repeat buyers and enroll them in a referral program with defined rewards. You then split comparable cardholders into a test group that receives the referral offer and an exactly matched control group that does not. By comparing the purchasing activity — and the new customers each group brings in — you learn precisely the effect the referral promotion has on referral rates and sales volume. Finally you feed those revenue and cost effects into a lifetime value chart to confirm the program raises net present value before rolling it out.

• If a referral tactic cannot be shown to raise customer lifetime value on a spreadsheet, then do not roll it out — lifetime value is the criterion of strategy.
• If you want to know the true effect of a referral promotion, then run an exactly matched control group against the test group; without it you are guessing.
• If a customer shows high loyalty and value (strong RFM, repeat purchase), then prioritize enlisting them in the referral program — loyal customers are the word-of-mouth engine.
• If you need budget for a referral program, then secure short-run successes first, because you can't justify the long-term budget without immediate results.
• If a referral tactic stands alone without fitting a coherent long-range strategy, then it is all dressed up with no place to go — integrate it with retention, sales, and cost levers.

• Running a referral program as an isolated clever tactic rather than inside a coherent long-range strategy — the tell: many marketers can build individual tactics but few assemble them into strategy, so the program sinks like a stone months after the accolades.
• Launching with no matched control group, so you credit the program with sales that would have happened anyway — the tell: you can state activity but cannot state the precise effect on referral rates.
• Judging the program on first-quarter numbers — the tell: behavior change requires two to three years, so early flat results trigger premature cancellation.
• Never defining referral rate as a variable in the lifetime value model — the tell: you treat word-of-mouth as uncontrollable rather than a lever that rewards and motivations can expand.
• Failing to store the customer information needed to identify referrers — the tell: no purchase behavior or survey data on file, so you cannot target your loyal customers for the program.

• ▸Target segment: which loyal/high-RFM customers you will enroll
• ▸Reward/motivation offered: the specific incentive for referring
• ▸Test group definition: who receives the referral offer
• ▸Control group definition: exactly matched group receiving nothing
• ▸Tracking mechanism: how you attribute new customers (e.g., membership card)
• ▸Baseline referral rate: current referrals per customer before test
• ▸Measured referral rate: test group result vs control
• ▸Revenue & cost effects: new-customer value minus reward/mailing cost
• ▸Lifetime value impact: net present value change, immediate and long-term

1. 1Identify which lever you are pulling before you act: more frequent purchases, higher-priced items (up-sell), or additional product categories (cross-sell). Each needs a different offer, so name the target explicitly.
2. 2Pull each customer's full purchase history from the database and flag the gaps — categories never bought, upgrades never taken, products bought by similar profiles but not by this customer. The Family Club database does this by auto-generating a suggested gift for every recipient from inventory reports plus prior-gift records.
3. 3Segment the base into affinity, lifestyle, and geographic groups so that every offer is relevant to the recipient's needs — relevance is what drives the incremental buy, not volume of mailings.
4. 4Make the right offer to the right people at the right time: upgrades and companion products to loyal buyers, new categories to profiles that match but have gaps. This is the stated mechanism for how database marketing lifts sales volume.
5. 5Shift the sale to a lower-cost, higher-retention channel where possible — software companies sell upgrades direct and keep 90% of revenue versus 50% or less through retailers, so the same volume increase yields more profit.
6. 6Split the target into a test group that receives the offer and an exactly matched control group that receives nothing. Compare purchasing activity of the two groups to measure the precise sales-volume effect of the promotion.
7. 7Roll out only after the test proves the lift, then recalculate customer lifetime value to confirm the increased spend persists rather than pulling future purchases forward.

Harald's department store runs a Family Club through its customer marketing database. Andrea and five other gift counselors each handle about 2,000 clients like Lydia; the database holds names, profiles, and prior gifts for each recipient, and automatically generates a suggested gift for every person on a client's list using inventory reports and marketing's creative ideas. Lydia buys 20-plus birthday and Christmas presents a year through it because it saves her the work. The volume payoff is not the club purchases themselves — the marketing manager points to a 12 percent growth in sales to Family Club members from purchases they made unrelated to the Family Club system. That is cross-category lift: the relationship and recognition pulled spending across the whole store, not just the service that created the tie. Together the counselors generate more than $5 million a year in sales.

• If a loyal customer has never bought an adjacent category that matching profiles buy, then cross-sell that category with a targeted, relevant offer rather than a discount.
• If you can sell an upgrade or companion product direct to an existing customer, then do so — retaining 90% of revenue beats 50% or less through a retailer at the same volume.
• If a promotion has not been run against a matched control group, then do not credit it with any sales-volume increase; you cannot separate lift from what would have happened anyway.
• If the offer is not relevant to the segment's demonstrated needs, then suppress it — irrelevant communications erode the loyalty that drives repeat volume.
• If a strategy raises short-term sales but lowers recalculated lifetime value, then kill it — you are borrowing from future purchases.

• Blasting offers to the whole list to chase volume. Tell: marketing costs rise with revenue and profit stays flat; the passages note profits increase by mailing to a targeted fraction, not millions.
• Measuring gross sales without a control group. Tell: you can report a number but cannot state the precise effect of the promotion, because you never isolated it.
• Treating database marketing as a one-off campaign. Tell: each initiative stands alone with no memory of previous contacts, so cross-sell and frequency plays never compound across the customer's history.
• Counting only the sales directly tied to a program. Tell: you miss the unrelated spillover — Harald's would have missed the 12 percent growth on purchases outside the Family Club.
• Pushing volume through the highest-cost channel. Tell: revenue grows but margin shrinks because you kept 50% instead of 90% selling direct.

• ▸Lever: name one — frequency / up-sell / cross-sell
• ▸Target segment: affinity, lifestyle, or geographic group
• ▸Observed gap: category or upgrade this segment has not bought
• ▸Offer and timing: the specific right offer at the right time
• ▸Channel: where the sale happens and % revenue retained
• ▸Test group vs. matched control: who gets the offer, who does not
• ▸Measured lift: incremental sales volume from test minus control
• ▸Spillover: unrelated purchases this program triggered
• ▸Recalculated lifetime value: before vs. after

1. 1Compute lifetime value for your customer base before designing any promotion — this is the criterion that tells you in advance whether a targeting strategy holds the promise of success. Use the simplified set of numbers Paul Wang's techniques reduce it to, so any competent marketer can master it.
2. 2Score prospects and customers by propensity to respond. Build this score from the data you hold: start with RFM (Recency, Frequency, Monetary) analysis, then refine with customer profiles and statistical models to separate high-responders from non-responsive segments.
3. 3Rank your file by that score and cut off the mailing or promotion at the point where the marginal response no longer covers marginal cost — mail only the high-propensity deciles, suppressing the rest.
4. 4Set up a control group that receives no promotion (or the standard treatment) so you can measure the actual lift your targeting produces against what would have happened anyway.
5. 5Run the test before spending at scale — the focus is on theory that tells you whether an idea will work before millions have been spent. Verify the cost-per-order and revenue-per-dollar against the control.
6. 6Roll out only the segments that beat the control on cost efficiency, then feed response results back into the database to sharpen the next round's scoring.
7. 7Track cost per order and marketing cost per dollar of revenue over a multiyear period, because the real test is reducing attrition and increasing repeat sales, not a single campaign's numbers.

Frank Perdue's chicken empire, the book notes, was not so much a triumph of production — which was superb — but of marketing, which was even better. Apply the efficiency logic to a Perdue-style promotion: instead of a mass mailing to every household, you score your file by RFM and model propensity to respond. Suppose scoring splits a 1,000,000-name file into deciles, and historically the top three deciles produce nearly all profitable orders while the bottom seven barely respond. You mail only the top 300,000 and hold back a control group within them. Because you eliminated 700,000 non-responsive names, your promotion cost falls roughly 70% while most of the orders remain — driving cost per order down and revenue per promotional dollar up. The control group proves the lift is real rather than a seasonal coincidence, so you roll out confidently and feed the response data back to refine next round's cutoff.

• If a segment's modeled response won't cover the marginal cost of reaching it, then suppress it — don't mail to prove the database works.
• If you cannot measure lift against a control group, then you cannot claim cost efficiency — any apparent improvement may be the Hawthorne Effect, where results rise simply because attention is being paid.
• If lifetime value analysis shows a strategy won't pay before rollout, then kill it at the test stage rather than after millions are spent.
• If you're tempted to compete purely on drastic price reductions, then shift to targeted relationship building instead — constant price cuts cannot be a successful long-range strategy and can be self-defeating.
• If a customer scores high on profitability and interest, then lavish special attention on them, since those are the customers most likely to build the bottom line.

• Treating the database as a cost-reduction gimmick only — the tell is a program with no lifetime value model and no retention metric; it reduces costs once then stalls because it never built relationships.
• Mailing everyone to be safe — the tell is a flat, undifferentiated file with no RFM scoring, which keeps wasteful spending on non-responsive segments and leaves cost per order high.
• Confusing database marketing with market research — the tell is spending to collect statistically accurate truth about the market rather than to begin a dialogue that reduces attrition and drives repeat sales.
• Skipping the control group and crediting the program for a lift that was really the Hawthorne Effect — attention alone raised response, and the next unattended campaign disappoints.
• Launching at full scale without testing — the tell is a project touted as a new trend that sank like a stone months after the accolades because no one verified the strategy before millions were spent.

• ▸Segment / decile: name each propensity tier from your RFM-plus-model score.
• ▸Names in segment: count of prospects/customers in this tier.
• ▸Modeled response rate: expected percent who respond, from history or model.
• ▸Cost to reach: promotional cost per name times names in segment.
• ▸Projected revenue: orders times average revenue per order.
• ▸Cost per order: cost to reach divided by projected orders — your efficiency metric.
• ▸Control group lift: response in treated minus response in holdout, to confirm the gain is real.
• ▸Mail / suppress decision: 'mail' only if cost per order beats target and lift is positive.
• ▸Lifetime value check: multiyear projected value confirming the segment pays beyond this campaign.

1. 1Profile your existing customers first, not your prospects. Segment the database by profitability and compute lifetime value for each segment so you know which customers actually build the bottom line versus which drain it.
2. 2Build a demographic/behavioral profile of your most profitable existing customers from the database — the attributes shared by the segment with the highest lifetime value, since these are the people you want to clone.
3. 3Use that profile to find new prospects that match it, rather than mass-marketing to everyone. Apply the database to qualify prospects before you spend acquisition dollars on them.
4. 4Set the customer acquisition budget with precision by anchoring it to lifetime value: the amount you can afford to spend acquiring a customer is bounded by the net present value of future profits that customer type produces.
5. 5Study the lifetime value of customers obtained from different sources and concentrate spending on the best sources — track which channels deliver prospects who match the profitable profile and stay.
6. 6Target the type of customer most likely to stay with the seller, not just the one cheapest to acquire; a cheap acquisition that churns fast destroys the economics.
7. 7Verify before scale: enter estimated acquisition costs and projected benefits (retention, annual sales, referrals) into a lifetime value chart to see whether a new acquisition strategy raises long-term profit before testing it with real money.

Treat a membership-card program as an acquisition-and-retention lever and run it through a lifetime value chart before launch. The costs are concrete: card issuance, promotion to get members to use the cards, capturing and storing purchase information, and premiums given for card use. The benefits are estimated as well: increase in the retention rate as cardholders accumulate equity, increase in annual sales to card members, and — the acquisition piece — increase in referrals of new customers by cardholders. Ukrops in Richmond, Virginia ran a frequent-shopper card program, analyzed its customer base regularly, and when a regular shopper appeared to drop the supermarket, went after them with letters and telemarketing. They sustained this for five years, which signals it produced results. The lesson: enter costs and benefits into the chart; if costs outweigh benefits lifetime value falls and the strategy is a loser that should not be introduced. If lifetime value dips short-term from promotional cost but rises long-term, the program is worth acquiring customers through.

• If a prospect does not match the profile of your most profitable existing customers, then do not spend acquisition budget on them — qualify first using the database.
• If a source produces customers with low lifetime value or high attrition, then cut it and reallocate to the sources delivering profitable, loyal matches.
• If the lifetime value chart shows costs outweighing benefits, then the acquisition strategy is a loser and should not be introduced, no matter how innovative it sounds.
• If a prospect type is cheap to acquire but unlikely to stay, then weight toward the customer most likely to remain with the seller, because retention drives lifetime value.
• If you cannot measure a strategy's effect on customer lifetime value, then you cannot evaluate it — build the measurement before you spend.

• Building the database but never profiting from it. The tell: the project is touted as a new trend, then sinks like a stone months after the accolades because no logical, practical, winning acquisition strategy sat underneath it.
• Treating all customers alike instead of segmenting by profitability. The tell: acquisition spend is spread evenly and you cannot say which segment or source builds the bottom line.
• Chasing cheapest-to-acquire prospects instead of profile-matched ones. The tell: acquisition numbers look good but attrition is high and lifetime value stays flat.
• Setting the acquisition budget by gut rather than lifetime value. The tell: millions get spent on a database project that did not survive because costs were never tested against projected future profit.
• Expecting the strategy to be served on a silver platter. The tell: marketers buy tools and consultants but never do the upfront work to come to grips with who their customers are and are not.

• ▸Most profitable segment: name the existing customer segment with highest lifetime value.
• ▸Profile attributes: list the demographics/behaviors shared by that segment.
• ▸Target prospect definition: describe the matching prospect you will pursue.
• ▸Acquisition source: name the channel and why it should yield matches.
• ▸Estimated acquisition cost per customer: enter the per-prospect spend.
• ▸Lifetime value of this customer type: enter net present value of future profits.
• ▸Retention likelihood: rate how likely this prospect type is to stay.
• ▸Acquisition budget ceiling: set max spend bounded by lifetime value.
• ▸Go/No-go: compare cost to benefit on the lifetime value chart and decide.

1. 1Define the time horizon first — LTV is net present value of profit on the average new customer over a given number of years, so fix the number of years (e.g., 3) before computing anything. Without a horizon the number is meaningless.
2. 2Build the number on the average new customer, not a single buyer. Start with an acquired cohort (e.g., 1,000 new customers) and track what happens to that group across each year of the horizon.
3. 3Lay out the retention reality year by year: of the cohort that starts, how many remain in Year 2, Year 3, and so on. Attrition is the central variable — reducing attrition is explicitly named as the hard part of making money with a database.
4. 4Attach revenue and all costs to each year. The passages insist on detailed cost accounting for every step; include acquisition cost, cost of goods, service calls, surveys, mailings, and the cost of the loyalty-building events (receptions, driving schools, pickup/dropoff service).
5. 5Compute net profit per year, then discount future-year profit back to present value — LTV is a present-value figure, not a raw sum of future dollars.
6. 6Reduce the whole thing to the simple set of numbers Paul Wang's technique produces: a spreadsheet any competent marketer can master, not a mysterious model.
7. 7Use the resulting LTV as the single pass/fail criterion. Before spending real money on any new strategy, test whether it raises LTV versus the control — if the strategy doesn't lift lifetime value, it doesn't ship.
8. 8Re-run LTV with and without each relationship investment so you can see whether events, clubs, dialogue, and super service actually pay back by lowering attrition and raising repeat purchases.

Take the Land Rover program. Lydia buys the car and is enrolled in the 'Royal Welcome': a dealership tour, introductions to key staff, and a British picnic basket. Follow-up phone calls, an extensive survey, maintenance reminders, and free drive-to-work service follow. Then come the golf-club reception, a yacht-club race invitation for her daughter, and the Off-Road driving school trip to Scotland. Each of these is a real cost line that must be entered into the LTV model. The payoff is on the retention side: Lydia will make the transition to her next Land Rover 'a matter of course' rather than defecting — and the cellular example states flatly that after such service she will not switch for a discount offer ('Not on your life'). LTV is how you prove that the picnic basket and the Scotland trip are cheaper than losing her and re-acquiring a replacement buyer.

• If a strategy does not raise the LTV of the average new customer, then reject it — LTV is the ultimate criterion by which all database marketing strategies are appraised.
• If you cannot assign a cost to a marketing step, then do not run it until you can — detailed cost accounting must be performed for every step.
• If a tactic lowers attrition even at added cost, then compare the service cost to re-acquisition cost; keep it only if retained lifetime profit exceeds the spend.
• If you are tempted to compete on price, then first check whether the buyer actually values price — buyers may value service, recognition, or reliability more, and those raise LTV at lower cost.
• If projected LTV supports it, then invest in finding prospects who resemble your most profitable existing customers rather than chasing anyone who responds.

• Talking about lifetime value but never computing it — the book names this directly: 'everyone talks about, but few people actually ever compute.' The tell: strategy decks cite LTV with no spreadsheet behind them.
• Summing future profit without discounting to present value, overstating the number. Tell: the horizon-year dollars are treated the same as Year 1 dollars.
• Building the database but having no strategy to make money from it. Tell: the project is launched on excitement, then 'sinks like a stone' months after the accolades.
• Ignoring attrition in the model. Tell: the cohort is assumed to stay intact, so retention investments look worthless and get cut.
• Omitting relationship-program costs (events, surveys, free service) from the calculation, so LTV looks artificially high. Tell: the picnic baskets and driving schools never appear as line items.

• ▸Time horizon (years): set the number of years LTV will cover
• ▸Starting cohort size: number of new customers acquired in Year 1
• ▸Acquisition cost per customer: total spend to win the cohort ÷ cohort size
• ▸Retention / customers remaining each year: how many of the cohort stay active in Year 1, 2, 3...
• ▸Revenue per retained customer per year: average purchases × price
• ▸Cost per customer per year: cost of goods, service calls, surveys, promotions, events (e.g., receptions, driving schools)
• ▸Net profit per year: (revenue − all costs) for the customers remaining that year
• ▸Discount factor per year: present-value rate applied to each future year's profit
• ▸Lifetime value: sum of discounted net profit across the horizon ÷ starting cohort size — the number every strategy is judged against

1. 1Segment your database by department or customer type before planning any relationship efforts — upper-income women, value shoppers, and do-it-yourself buyers respond differently, so compute lifetime value separately for each and target the relationship accordingly rather than treating all customers alike.
2. 2Fund the dialogue, not the discount: budget relationship costs per customer per year (Robin budgets $10 each), which buys the database, data entry, point-of-sale purchase capture, a newsletter, and personalized communications on birthdays, anniversaries, and holidays, plus telemarketing to the best customers.
3. 3Build recognition into every contact by referencing the customer's own history — their tenure, cumulative value, or prior purchases — the way the insurance example does: 'you have been a customer of ours since 1976. The cash value of your policy has now grown to more than $28,000.'
4. 4Run satisfaction surveys as a dialogue tool, not a reporting tool: store each respondent's answers in the database and use them to open a one-on-one conversation with that individual, since the overall tabulated results matter less than the per-customer follow-up.
5. 5Establish institutional identification structures — a president's club, a gold card group, or a frequent buyer/flyer/shopper club — that build loyalty and give customers a reason to buy more in order to belong.
6. 6Model each new relationship effort as a change to the retention rate and the average yearly sale, then project both forward across your computation period before spending.
7. 7Test every new strategy (gold cards, president's clubs, newsletters) by measuring lifetime value before and after; if the calculation shows no increase after your computation period, drop or modify the effort.

Robin plans database activities for Ridgeway and assumes her relationship-building efforts will lift the average customer's Year 1 clothing purchases from $150 to $180 — a 20 percent increase (Line E). She concedes this may be too ambitious but treats it as a testable proposition. She also assumes a retention rate of 50 percent of last year's customers plus 5 percent new referrals (Line B), a reasonable goal her well-designed programs might beat. For later years she projects loyal remaining shoppers will buy about $20 more per year each, a less questionable assumption. She budgets database costs at $10 per customer per year (Line I), which covers the database, data entry, POS purchase capture, a newsletter, personalized birthday/anniversary/holiday communications, and some telemarketing to her best customers. Total revenue (Line F) is simply average yearly sale times customer count (Line C), and the resulting lifetime value lines let her see whether the $10 spend pays off before committing serious money.

• If the buyer values service, information, reliability, recognition, or friendship over price, then build those into the relationship rather than cutting price — discover each buyer's subjective value and fill it.
• If one lifetime-value number covers all departments or customer types, then re-segment, because lumping them together leads to false strategies that turn off some customers while turning on others.
• If a relationship effort's before-and-after lifetime value shows no increase after the computation period, then drop or modify it.
• If a customer resembles your most profitable existing customers, then use survey data, appended external data, and modeling to find and win over prospects who match that profile — and concentrate acquisition on the sources that produced them.
• If you run a satisfaction survey, then store results per respondent and follow up individually; if you only tabulate and discard, you have wasted the relationship opportunity.

• Treating relationship building as discounting — the tell is that your program competes on price, when competitors can always outspend, underprice, and overdistribute you; only a database-implemented relationship resists that.
• Launching the database without a strategy — the tell is a project 'touted as a new trend' that sinks like a stone months later; building a database is easy, making money and reducing attrition over multiple years is the real challenge.
• Lumping all customers and products into one lifetime-value figure — the tell is a single average that hides that upper-income, value, and do-it-yourself shoppers want different things.
• Collecting survey or customer data but never using it — the tell is tabulated results that get discarded once reported instead of stored and used to open dialogue.
• Talking about lifetime value without ever computing it — the tell is that 'everyone talks about' it but few recalculate it month to month as competitors, obsolescence, and saturation change the numbers, so innovations go untested before serious money is committed.

• ▸Customer segment: name the department or customer type this sheet covers
• ▸Current average yearly sale: dollars this segment buys now, pre-effort
• ▸Projected average yearly sale: dollars after the relationship effort, with your assumed percent lift
• ▸Retention rate assumption: percent of last year's customers kept, plus any referral percent
• ▸Relationship efforts funded: list specific activities (newsletter, birthday/anniversary mailings, club, telemarketing, personalized recognition)
• ▸Database cost per customer per year: dollar budget covering data, entry, POS capture, and communications
• ▸Computation period: number of years over which you will measure value for this industry
• ▸Before/after lifetime value: cumulative net profit per original customer, pre- and post-effort
• ▸Keep/modify/drop decision: your call based on whether value rose after the computation period

1. 1Pull a cohort from your database — all customers who came on board at about the same time — using 1,000, 10,000, or more records depending on base size, so your analysis rests on comparable buyers.
2. 2Run RFM analysis to score every customer on Recency, Frequency, and Monetary value, then rank cells by predicted response and dollar sales before any offer is mailed.
3. 3Suppress the cells whose predicted response is low — mail nothing to them — and schedule the high-responding cells to receive promotions more often; expect lower total revenue but higher profit.
4. 4For prospecting, profile your existing responsive or most profitable customers by surveying them on demographics (age, income, presence of children) and overlaying the base with external rented-list data.
5. 5Run ad hoc cross tabulations or neural network modeling on the combined overlaid-plus-existing data to isolate which variables actually separate respondents from nonrespondents.
6. 6Clone the winning profile against outside lists, rent names matching it, and market to those clones using methods already proven in tests.
7. 7Set exactly matched control and test groups: test groups get the offer, controls get nothing, so you can read the net effect on sales, retention, and referral.
8. 8Run paper lifetime-value calculations, then a small-scale test, and only commit serious marketing dollars after both confirm the strategy pays.

A company that mails millions of offers a year wants higher profit, not just more mail. It scores its file with RFM and finds low-predicted-response cells that would lose money if mailed. It mails only a targeted fraction of the list — the high-responding cells — which cuts both cost and total revenue but raises net profit. For prospecting it profiles its best responders, overlays external data, and models to find the separating variables, then rents matching names and mails only those clones. Ukrop's applies the retention side of this logic: when a regular shopper drops the supermarket, it goes after that specific shopper with letters and telemarketing, and it has run the program for five years — evidence it produces results.

• If a cell's predicted RFM response is low, then mail it nothing and reallocate that budget to high-responding cells.
• If you have existing customers, then profile responders with a test mailing; if you have none, then still run a test mailing and profile respondents versus nonrespondents.
• If an initiative adds marketing cost (newsletter, membership card, survey), then keep it only when the paper lifetime-value gain across the five retention/referral/spend/direct-cost/marketing-cost levers exceeds that cost.
• If paper calculations look favorable, then test on a small scale before committing serious dollars; never roll out on projection alone.
• If you can sell upgrades direct to existing customers (retaining 90% of revenue versus 50% or less through retailers), then shift the channel rather than discount price.

• Mailing the whole list instead of the profitable fraction — the tell is high gross revenue paired with flat or falling net profit.
• Treating low price as the loyalty strategy — the tell is constant price reductions that still fail to hold market share in competitive industries.
• Skipping the control group so you cannot isolate a promotion's effect — the tell is 'sales went up' claims with no matched non-mailed baseline.
• Modeling without first surveying and overlaying external data, so you lack the variables that separate respondents — the tell is a profile that won't match rented lists.
• Building the database but failing to develop a logical, practical, winning strategy — the central fault behind wasted database projects; the tell is spend with no lifetime-value test behind it.

• ▸Cohort source: which same-start-date customer group and how many records
• ▸RFM cell & score: recency/frequency/monetary rank for this segment
• ▸Predicted response & dollar sales: pre-mail estimate for the cell
• ▸Mail decision: suppress / mail / mail more often
• ▸Profile overlay variables: demographic and external fields used to define the segment
• ▸Modeling method: ad hoc cross tabs or neural network, and separating variables found
• ▸Test vs control setup: matched groups and the metric compared (sales, retention, referral)
• ▸Paper lifetime-value result: projected net profit vs. initiative cost before spend
• ▸Go/no-go: small-scale test outcome and rollout decision

1. 1Build the customer database first, then compute each customer's lifetime value using the LTV method (direct cost percent, sales commission, retention across years) so you know what a retained customer is actually worth.
2. 2Determine the profile of your most profitable customers by analyzing which attributes correlate with high LTV — this is the characteristic set you will search for in outside prospect pools.
3. 3Build a model (or linked models) from your response and profitability data; Metromail's Ken Hartmann links five models to predict response to solicitation, number of loans over a lifetime, average loan amount, profitability of payments, and lifetime value of each new customer.
4. 4Score outside prospect lists against the model and select only prospects whose scores match your profitable-customer profile.
5. 5Never roll out on the model alone: mail simultaneously to a model-scored group and a control group picked the old way, in the same period, same offer, same economy, so the only variable is the model.
6. 6Measure the lift between the two groups, then subtract the cost of the variable data the model required (name-rental agencies seldom let you rent the extra variables cheaply) to confirm the model pays for itself.
7. 7Layer advance telemarketing after the mailing — call respondents, score 'hot leads' from the calls, make appointments, and send reps only to appointments — to cut cost per lead dramatically.
8. 8Compute the maximum you can spend to acquire a lead by dividing customer LTV by your conversion rate, and manage spend against that ceiling.

Susan Hancock of Hancock Information Services shows the payoff on a mainframe software company selling $60,000 purchasing/materials software. The company had prospected by direct mail followed by a sales visit at a cost per lead of about $150. As a test, telemarketers called likely prospects already on the direct-mail list: 133 hours of calls produced 617 completed 'research survey' interviews, averaging 4.5 interviews per hour, at a total calling cost of $6,650. Of those survey calls, 19 percent — 117 leads — qualified as hot leads. That drove cost per lead to $57, about one-third the cost of direct-mail prospecting. The lesson: matching prospects to a profile and qualifying them by phone before a sales visit makes the sales force cost-effective.

• If a qualified lead is worth $227 (LTV $2,273 at a 1-in-10 conversion), then every discarded lead throws away $227 — $45 already spent plus the rest in anticipated profit — so build reporting to recover them.
• If the control group responds at 2 percent and the model group at 3 percent, that is a 50 percent lift, but accept the model only if that lift exceeds the cost of the variable data it needs.
• If you have a narrow, finite market (the high-tech firm had only 200 prime prospects for its $2M+ product), then call deep into each organization — Hancock called up to 17 levels deep, spent $200,000, and got 23 sales worth $46 million.
• If product cost and project size are large, telemarket all prospects; if small, telemarket only respondents, deciding by cost of the product versus cost of calling.
• If you want sales reps to report lead status, then make it in their economic best interest — people respond to what pays them.

• Testing this month's mailing against last month's instead of running simultaneous control and model groups — the tell: you can't separate the model's effect from seasonal, offer, or economic shifts.
• Turning sales reps loose on a territory instead of following the eight-step process (database, profile, match, mail, telemarket, hot leads, appointments, visits) — the tell: reps burn hours on unqualified prospects.
• Treating last month's response behavior as a permanent fact about how people react — the tell: forecasts that assume stable response and then miss.
• Building the database well but having no profit strategy — the tell: a technically sound project that sinks 'like a stone' months after the accolades.
• Ignoring the cost of variable data when chasing model lift — the tell: a statistically better model that loses money because name-rental surcharges eat the gain.

• ▸Profitable-customer LTV: compute from direct cost %, commission %, and multi-year retention (example: $2,273 at 5 years).
• ▸Conversion rate: leads-to-paying-customer ratio from sales experience (example: 1 in 10).
• ▸Value per lead: LTV / conversion rate = max acquisition spend per lead (example: $227).
• ▸Profile attributes: list the characteristics shared by your highest-LTV customers.
• ▸Model score threshold: minimum prospect score that qualifies for mailing.
• ▸Control vs. model mailing plan: equal groups mailed same period/offer; record response % of each.
• ▸Variable-data cost: per-name cost of the extra data the model requires.
• ▸Net lift: (model response − control response) value minus variable-data cost.
• ▸Telemarketing follow-up metrics: calls, completed interviews, interviews/hour, hot-lead %, cost per lead.

1. 1Define satisfaction operationally from the customer's point of view: it is the feeling you produce by providing recognition, service, friendship, and information — build your program around delivering these four, not around pushing offers.
2. 2Capture the raw data that makes recognition possible: tenure (customer since-date), cumulative value, product holdings, and dates/events tied to the household, so you can speak to each customer as an individual rather than as a segment.
3. 3Run satisfaction surveys, but change their purpose: store every individual response in the database instead of discarding results after tabulation, and treat the survey as an opener for dialogue with each respondent, not just an aggregate score.
4. 4Trigger event-driven communications off stored data — affinity-product announcements when a particular product is ordered, or anniversary/birthday reminders — so the customer experiences timely usefulness rather than generic mail.
5. 5Close the loop by acting on each respondent's stored answer: follow up individually so the customer sees you remembered, which converts a one-time survey into an ongoing relationship.
6. 6Measure whether satisfaction is actually rising by watching its downstream signals in the database: retention/attrition rate, repeat purchase, and average yearly sale per customer.
7. 7Verify any satisfaction-driven initiative with testing and control groups before scaling, so you prove the lift (e.g., higher average purchase) rather than assume it.

An insurance company uses stored tenure and cash-value data to open a personal communication: "As you know, Mr. Sawyer, you have been a customer of ours since 1976. The cash value of your policy has now grown to more than $28,000. In view of this..." This single line delivers three satisfaction drivers at once — recognition (naming his 1976 start), information (his current $28,000 value), and service (a relevant next step). It works precisely because the database preserved the start-date and running value rather than treating them as dead records. Contrast this with the common pattern of running a satisfaction survey, tabulating a company-wide score, and discarding the individual answers — that produces a number but no relationship. The disciplined approach stores each respondent's answer and uses it to open a dialogue, because the overall results matter less than the per-respondent conversation. The payoff is the documented chain: make the customer happy, and they reward you with loyalty, reduced attrition, and increased sales.

• If a customer survey has been tabulated into an aggregate score, then store the individual responses and open a per-customer dialogue — the overall result matters less than the use of the survey with each respondent.
• If a particular product is ordered, then trigger an affinity-product announcement or personal message, because timely useful information is a satisfaction driver.
• If you hold relevant household data (tenure, value, dates), then reference it by name in communications to deliver recognition — the one satisfaction factor fully under your control.
• If retention rate or average yearly sale is not moving after a satisfaction initiative, then treat the initiative as unproven and re-test with a control group rather than scaling it.
• If a customer would be glad you have a database and included them in it, then you are doing it right; if the contact feels intrusive or generic, you are missing recognition, service, or information.

• Running satisfaction surveys and discarding the results once tabulated — the tell is you can report a company score but cannot say what any single respondent answered.
• Treating database marketing as cost-cutting and more selling only, forgetting it is a tool to make customers happy — the tell is rising mail volume with flat or worsening attrition.
• Chasing the overall survey number instead of per-customer follow-up — the tell is dashboards full of aggregate metrics and zero individual dialogue.
• Assuming satisfaction automatically lifts sales without testing — the tell is projections like a 20% jump in average purchase stated as fact, with no control group behind them.
• Generic, un-personalized contact that ignores stored tenure and value — the tell is communications that could go to anyone, delivering no recognition and no useful information.

• ▸Customer tenure / since-date: the exact start year to use for recognition.
• ▸Cumulative value or holdings: the stored figure (e.g., cash value, lifetime purchases) to cite as information.
• ▸Trigger event: the ordered product, date, or milestone that launches the contact.
• ▸Satisfaction survey response stored: the individual answer you will act on, not the aggregate.
• ▸Recognition line: one sentence naming what makes this customer individual.
• ▸Useful offer or information: the affinity product or next step relevant to this record.
• ▸Follow-up action + owner: who closes the loop with this respondent and when.
• ▸Downstream metric watched: retention/attrition, repeat rate, or average yearly sale to confirm lift.
• ▸Test/control group flag: yes/no, whether this contact is being verified before scaling.

1. 1Collect per-customer data that is relevant to your product: name and address, complete purchase history, service calls/complaints/returns/inquiries, survey results, and household demographics. Store it in a relational database accessible to every unit that touches the customer.
2. 2Open the information to all customer-contact points — customer service, technical support, sales — so any interaction is informed. Lydia's '*611 SEND' call worked because the rep already knew her name, vehicle, and plate before she finished speaking.
3. 3Build the bond that inoculates against discounts. Discover each buyer's subjective definition of 'profitable' (price, service, reliability, information, recognition, friendship) and supply that, so a temporary competitor price drop no longer triggers switching (churning).
4. 4Design proactive relationship events, not just statement stuffers: welcome programs, follow-up calls, surveys, clubs, and member-only experiences that make the customer identify with the company as an institution.
5. 5Use the purchase cycle to pre-stage the next sale. Record SKUs, dates, amounts, and the serving clerk so you can reach customers with personalized, data-rich communications timed to maintenance, replacement, or related-interest triggers.
6. 6Budget for 2–3 years minimum before customer behavior shifts, and secure corporatewide commitment plus leadership/teamwork across MIS, service, sales, and any outside agencies — database loyalty programs fail without this integration.
7. 7Measure loyalty economically by lifetime value and attrition reduction, and verify every retention tactic with tests and control groups before scaling it to millions.

Lydia McCabe bought a Land Rover and, unknowingly, entry into an exclusive club. The dealer's 'Royal Welcome' included a dealership tour, introduction to key personnel, and a British picnic basket, followed by phone calls to confirm all was well, an extensive survey on every aspect of the vehicle and service, and a maintenance reminder with free drive-to-work-and-pickup during service. The relationship extended beyond service: her daughter was invited to an agency-sponsored yacht club race, and a year in, Lydia and Jack attended the Land Rover Off Road driving school in Scotland. The stated problem was ensuring she buys another Land Rover at the end of her three-to-four-year ownership and recommends it to friends. The solution was a customer database driving super service, surveys, communication, and events that make the next purchase 'a matter of course.' The cellular case shows the same logic defensively: after the '*611 SEND' rescue, the question 'Will she switch for a discount in a few weeks? Not on your life' — because she bought friendship, service, and security, not just a phone.

• If a competitor's lower price can pull your customer away, then you have a transactional relationship, not loyalty — build a service/recognition bond to inoculate against the discount rather than matching the price.
• If you are in a highly competitive industry, then treat constant price reductions as unsustainable long-range strategy and shift spend to relationship marketing that reduces attrition.
• If a customer makes a significant or personalized purchase, then follow up with a data-rich personal communication (name, product, date, benefit) — like the Nordstrom china letter — not a generic promotion.
• If a product has a known ownership cycle (e.g., 3–4 years), then schedule proactive contact before the cycle ends to make the repeat purchase automatic.
• If you cannot commit a 2–3 year budget and cross-unit cooperation, then do not launch the database loyalty program — half-measures sink 'like stones.'

• Treating loyalty as newsletters and statement stuffers alone — the tell: no integration with service, tech support, or sales, and no behavior change after a year.
• Chasing market share with drastic price cuts — the tell: a price-war spiral like the airlines that eliminated carriers and brought huge losses instead of loyalty.
• Expecting fast results — the tell: pulling the plug before the 2–3 year minimum needed for a database to affect customer behavior.
• Building the database but having no logical, practical, winning strategy to profit from it — the tell: millions spent on a project 'that did not survive' and sank after initial accolades.
• Making data inaccessible to contact staff — the tell: a service rep who doesn't already know the caller's name, account, and history, so interactions feel generic instead of personal.

• ▸Customer name & address — identity anchor for every contact
• ▸Complete purchase history — SKUs, dates, quantities, dollar amounts, serving clerk
• ▸Service record — calls, complaints, returns, inquiries to date
• ▸Subjective 'profit' driver — what this buyer values most (price, service, reliability, information, recognition, friendship)
• ▸Ownership/replacement cycle — expected timing of next purchase and the pre-sale contact date
• ▸Relationship events delivered — welcome, follow-up calls, surveys, club/event invitations
• ▸Attrition/switch risk — exposure to competitor discounts and current inoculation level
• ▸Next personalized action — the specific data-rich communication and its trigger date