The Customer Centricity Playbook Implement a Winning Strategy Driven by Customer Lifetime Value

1. 1Separate goodness from outcomes. Write down that goodness is the baked-in part of a customer's value (preference, propensity, potential), and that behaviors such as repeat purchase, referral, and longevity are how it shows up. You are reading behavior to infer the trait, not crediting the behavior to your marketing.
2. 2Define the three components for your business in one line each. Preference is how well your offering fits the customer's needs and how often they choose you over a competitor. Propensity is the likelihood of loyalty, referral, or buying higher-valued offerings. Potential is future value: what you can maintain, enhance, or extract.
3. 3List the behavioral propensities you can observe at the individual level. Start with those the passages name: how long the customer stays, how much they spend over the life cycle, referrals, and cost to serve. Add nonfinancial behaviors where they apply, such as volunteering, social sharing, and endorsements.
4. 4Find the proxy that best predicts future value. EA found that in-game engagement was the best single predictor of whether a customer would buy other games later. Test candidate behaviors against later purchasing and keep the one that predicts best.
5. 5Tie each behavior to a customer ID. Make sure customers are consistently defined and uniquely identified in the CRM, and link as many purchases and activities as possible to specific customers. Without this, you cannot see heterogeneity.
6. 6Decide where nonfinancial behaviors live. Usually keep them out of the CLV formula and use them for profiling and targeting. Ask of each one: does it improve long-run value predictions, and do you know how to model it, or is it just nice to know?
7. 7Profile your highest-value customers and look for shared inherent characteristics. Use individual-level data rather than demographics or personas, which fall victim to the streetlight effect. Then find and acquire customers who look like them.
8. 8Allocate resources by goodness. Direct retention and development spend toward the best customers. Check where you are spending heavily on customers who are not valuable enough to deserve that attention. Do not fire or ignore the rest.

Electronic Arts treated its players monolithically for most of its history and spent 22% of revenue on marketing, mostly in pre-launch broadcast advertising with no clear return measurement. A small data team on the sales planning team showed through a marketing mix algorithm that the spend was inefficient. After Andrew Wilson became CEO in 2013 and championed a "Player First" mantra, a Global Analytics & Insights group consolidated what the company knew about players. It found that in-game engagement was a good proxy for the future value of individual players, and the single best predictor of whether a customer would buy other games. That is customer goodness read through behavior: players who look alike on the surface differ in propensity and potential. EA then designed games to attract and retain its best customers, and game designers used engagement data to see what was working. The turnaround took more than a decade to show on Wall Street. The stock rebounded 1,000% over six years and 2017 profit cleared a billion dollars.

• If a behavior (such as engagement) predicts future purchasing better than other measures, then use it as your goodness proxy and design around it, as EA did.
• If a nonfinancial activity (referral, volunteering, sharing) improves long-run value predictions and you know how to model it, then consider including it in CLV. If it is merely nice to know, then use it only for profiling and targeting.
• If you are tempted to segment by demographics or personas, then switch to direct, individual-level data, because those approaches suffer from the streetlight effect.
• If a customer segment is not valuable enough to justify the attention it receives, then reduce investment in it, but do not fire or ignore it wholesale.
• If you are tempted to believe training or marketing can turn most of the base into top customers, then assume most of their value is predetermined and expect only limited development.

• Treating customers as monolithic and spending broadly (EA's spray-and-pray TV spend). The tell: no per-customer value differences in reports and no clear measure of marketing returns.
• Measuring only payments and ignoring referrals, volunteering, sharing, and endorsements. The tell: your best advocates look average in your value metrics.
• Profiling by demographics and personas. The tell: segments look tidy, but customers within them behave very differently.
• Believing development can make everyone a top customer (the Phelps fallacy). The tell: heavy loyalty or upsell spend on low-propensity customers with little change in their behavior.
• Judging goodness from early outcomes alone, not a lifetime projection. The tell: the best customer is whoever bought most recently, and long-run retention and total value are not modeled.

• ▸Preference indicator: how you will observe fit with needs and choice of you over competitors.
• ▸Propensity behaviors tracked: loyalty/retention, referral, upgrade purchases, cost to serve, with the data source for each.
• ▸Potential estimate: expected future value and the method used (CLV model, period, discount factor).
• ▸Candidate engagement proxy: the behavior you test as the best predictor of future purchasing (EA used in-game engagement).
• ▸Proxy validation result: how well the proxy predicted later purchases, and the decision to keep or drop it.
• ▸Nonfinancial behaviors: list each, then mark it as inside CLV or profiling-only, with the reason.
• ▸Customer ID coverage: share of purchases and activities linked to a unique customer in the CRM.
• ▸Top-customer profile: shared inherent traits of your highest-value customers, from individual-level data.
• ▸Resource shift decision: where you will increase or reduce attention based on goodness, and the review date.

1. 1Define engagement as a set of customer-level behavioral propensities, not one metric. Name the ones your business can observe: repeat purchase, likelihood to leave (retention), higher-value purchases, and referral or evangelizing. The source treats these as linked on average ("If a customer stays, the customer spends") but not guaranteed to be linked in any one customer.
2. 2Score each propensity separately per customer. Do not collapse them into one loyalty score. The source's example: one customer prefers you over competitors (low propensity to leave) but has little need for your offering (infrequent repeat cycle). A second has the same preference and a greater need (frequent repeat cycle). Keep the two propensities in separate columns.
3. 3Find your own engagement proxy for future value. EA's analytics group found that in-game engagement was a good proxy for a player's future value, and the single best predictor of whether a customer would buy other games. Test which observable engagement behavior predicts later purchasing in your data, and standardize its definition before you use it.
4. 4Answer the question the source poses: given two customers with different propensity profiles, which has the higher CLV? Project each customer's expected lifetime (how long they stay) and expected spend over that lifetime, then compare projected lifetime value rather than recent spend.
5. 5Track nonfinancial value signals alongside payments. Referrals, social sharing or promoting of your content, volunteering (for member organizations), and wellness activities (for pharmacy-style programs) are all in the source. They are harder to track than payments, so assign an owner and a capture method for each one.
6. 6Decide where each nonpurchase behavior lives. The source cautions against building nonfinancial activities directly into the CLV formula in most cases. Use them for profiling and targeting outside CLV, unless you can show the behavior improves long-run value predictions and you know how to include it. Ask of each: does it improve prediction, or is it just "nice to know"?
7. 7Invest in propensities that can still move. Baseline propensities are not fixed. Premium services, loyalty programs, and a stronger brand reputation can raise propensities to stay and spend. A terrible service experience (the escalating "I demand to speak to your manager!" threat) makes leaving much more likely. Target development spend at customers who already have good baselines, because the source says you cannot turn most of the base into your best customers.
8. 8Keep accounting rules consistent. For returns, choose one treatment: (a) erase the purchase, (b) keep it at zero or reduced value, or (c) count both purchase and return as transactions since both were engagement. Whichever you pick, apply it uniformly, and have marketing and accounting agree on it and revisit it often.

Electronic Arts (EA) was named the worst company in America in 2013 for the second year running. Five years before its stock began to recover, analysts on the sales planning team found that EA spent 22% of revenue on marketing, much of it in pre-launch TV and broadcast spending with no clear way to measure returns. A marketing-mix algorithm showed the spending was inefficient. A team then standardized metrics so hundreds of data points could be analyzed consistently. When Andrew Wilson became CEO in 2013, he used the correlations between engagement and customer value to champion a "Player First" mantra. The Global Analytics & Insights group showed that in-game engagement was a good proxy for the future value of individual players and the single best predictor of whether a customer would buy other games. Game designers now use engagement data as guideposts for what is working and what needs to improve. Over six years EA's stock value rebounded 1,000%, and in 2017 it cleared a billion dollars in profit. The lesson: find the engagement behavior that predicts future purchasing, standardize it, and build product decisions around it.

• If a customer scores high on one propensity (e.g., prefers you to competitors), then do not assume high CLV. Check need and purchase frequency before ranking them.
• If a nonfinancial behavior (referral, sharing, volunteering) measurably improves long-run value predictions and you know how to model it, then include it in CLV. If it is only "nice to know," then use it for profiling and targeting outside the formula.
• If a customer has a good baseline and responds to premium services or loyalty programs, then invest in developing them. If the baseline is weak, then do not expect spend to transform them.
• If a customer has a bad service experience and threatens to leave, then treat them as high churn risk, because they are much more likely to follow through.
• If you adopt a treatment for returns or overhead allocation, then apply it consistently and review it with accounting regularly. Consistency matters more than which option you choose.

• Measuring loyalty only through payments. The tell: referrals, volunteering, and sharing are absent from your data, and customers who spend less are written off even though they bring in others.
• Treating revenue-generating behavior as goodness. The tell: customers who frequently incur overdraft fees rank highest on financial CLV, which the source says misses the point of customer centricity.
• Collapsing propensities into one loyalty score. The tell: a loyal but low-need customer is ranked the same as a loyal, high-need customer, and repeat-purchase forecasts miss.
• Folding every engagement signal directly into the CLV formula. The tell: the model contains behaviors nobody can show improve prediction, and you cannot explain which inputs drive the number.
• Inconsistent accounting for returns or costs. The tell: marketing and accounting report different customer values, because one erases returned purchases and the other counts them as transactions.

• ▸Customer or segment: identify who is being scored.
• ▸Propensity to stay: observed retention evidence and expected lifetime.
• ▸Propensity to purchase: repeat-purchase cycle and frequency.
• ▸Propensity to spend higher: evidence of higher-value purchases.
• ▸Propensity to refer or evangelize: referrals, social sharing, or promotion observed.
• ▸Other nonfinancial engagement: volunteering, wellness actions, or similar behaviors, with who captures them.
• ▸Engagement proxy tested: the behavior and whether it predicts future purchasing in your data.
• ▸Treatment of nonpurchase signals: in CLV, or profiling/targeting only, with the reason.
• ▸Development lever and risk flags: premium service, loyalty program, or brand investment planned, plus any service-failure or churn-threat signals.

1. 1Fix the unit of analysis as the individual customer relationship, not 'the average customer.' The source calls averaging an outdated way of thinking that treats customers as a monolithic entity. Every later step produces one value per customer.
2. 2Standardize your metrics before you model anything. EA's analytics team created 'a unified, reliable way' to analyze hundreds of data points by standardizing metrics. Agree on one definition each for customer, purchase, active, and engagement, so each customer's history reads the same across systems.
3. 3Record each customer's status and contractual context. Mark whether the customer is currently active, and whether the relationship is one where ending it is visible or one where the customer can go quiet without notice. This determines how you read silence when you project retention.
4. 4Map heterogeneity explicitly. List the qualities the source names as varying: propensity to buy, tendency to be loyal, ways of communicating with one another, and categorically different responses to the same offers. Capture at least one measurable field for each, so two customers who look alike on the surface can still separate.
5. 5Find a leading indicator of future value and test it. EA found in-game engagement was a good proxy for the future value of individual players and the single best predictor of whether a customer would buy other games. Pick your candidate behavior, then check whether it predicts later purchases before you rely on it.
6. 6Project the whole lifetime, not the next transaction. Forecast each customer's future cashflows with variable retention, since customers do not leave at a uniform rate. Discount them to present value, and add relevant nonfinancial value as a separate, labeled line.
7. 7Rank customers by projected value and accept that much of it is predetermined. The Phelps analogy holds that much of a customer's potential value exists before the first purchase, and you cannot turn most of the base into your best class. Use the ranking to decide whom to acquire, retain, and develop, and do not assume everyone can be developed.
8. 8Redirect spend against the ranking, with a measurable goal for each dollar. EA cut marketing spend, which was 22% of revenue, after its marketing-mix analysis showed inefficiency. The cut produced no negative impact on efficacy. Track returns on any reallocation the same way.

Electronic Arts shows the logic of CLV-style thinking even though the passages give no CLV formula. Before the turnaround, EA spent 22% of revenue on marketing, mostly right before a game launch, on TV ads and other broadcast 'spray-and-pray' approaches. No one had clear goals for measuring the returns. A few analysts on the sales planning team optimized a marketing-mix algorithm. It showed the spend was inefficient, and the resulting decrease in marketing dollars came with no negative impact on efficacy. After EA bought Playfish in 2009, Zachery Anderson led analytics for EA Digital. His team standardized metrics across hundreds of data points. Later, under CEO Andrew Wilson's 'Player First' mantra, Global Analytics & Insights showed that in-game engagement was a good proxy for the future value of individual players. It was also the single best predictor of whether a customer would buy other games. EA could then treat players as individuals with different expected futures rather than as one monolithic base. The company went on to clear a billion dollars in profit in 2017.

• If customers look the same on the surface but behave differently, then stop using an average and value each customer separately. That is the principle EA acted on.
• If you must estimate future value for customers with little history, then use a tested behavioral proxy. EA used engagement. Do not use a proxy you have not checked against later purchases.
• If a marketing spend has no stated goal for measuring its return, then treat it as unproven and test it. EA's pre-launch broadcast spend had none, and cutting it cost nothing in efficacy.
• If you are tempted to spend heavily to lift every customer to top-tier value, then stop. Much of a customer's potential is predetermined, so direct development effort at customers with the potential to repay it.
• If teams worry that data will override creative or product judgment, then position the value insights as guideposts for that judgment. EA's studios use engagement insights to scaffold ideas rather than let data decide everything.

• Valuing the 'average customer.' The tell is a single CLV number, or one retention rate, applied to the whole base.
• Product-centric chasing of sales to anyone and everyone at any cost. The tell is that acquisition and marketing budgets are organized around product lines, with no ranking of customers by value.
• Judging value on a short race instead of the long game. The tell is that customers are ranked by last-period spend or the first purchase, not by projected lifetime.
• Inconsistent metrics. The tell is that teams cite different counts of active customers or engagement, and you cannot compare customers across data sources until someone standardizes the definitions.
• Trusting a proxy because it sounds plausible. The tell is that no one has ever compared the proxy to subsequent purchases. The EA finding looks obvious in hindsight but came from analysis.

• ▸Customer ID and status: record the unique ID and whether the customer is currently active.
• ▸Contractual context: note whether the relationship has a visible end or whether the customer can lapse silently.
• ▸Heterogeneity markers: record the measurable fields you hold for propensity to buy, loyalty tendency, communication behavior, and response to offers.
• ▸Leading indicator: name the behavior (for example, engagement) and its current level for this customer.
• ▸Indicator validation: note whether you have tested the indicator against later purchases, and the result.
• ▸Projected cashflows by period: forecast each period's cashflow, reflecting your retention assumption.
• ▸Discount rate and present value: state the rate you used and the resulting present value.
• ▸Nonfinancial value: list any relevant nonfinancial value separately, and say why it counts.
• ▸Action tier: decide whether to acquire, retain, develop, or deprioritize, with the measurable goal for the spend.

1. 1Fix the definition before you build anything: customer equity (CE) is the sum of all present and, ideally, future customer lifetime values (CLVs) across the customer base. It is a sum of individual customer values, not an average and not a brand score.
2. 2Build the CLV for each customer as a forward-looking projection of their entire lifetime with you. The authors call this 'playing a long game.' Past spend alone is not CLV.
3. 3Pick the discount factor(s) deliberately and write them down. The source says a CRM should be set up to 'choose appropriate discount factor(s),' so future value is not counted at face value. Record the choice so the CE figure can be reproduced.
4. 4Where direct value data is thin, find a behavioral proxy and test it. EA found that in-game engagement was a good proxy for a player's future value and the single best predictor of whether they would buy other games. Look for the equivalent signal in your own data before you sum.
5. 5Sum the CLVs, but keep the distribution visible. Customers who look alike on the surface are not alike (the EA lesson), and much of a customer's potential value exists before the first purchase (the Phelps analogy). Report CE alongside the share held by your best customers, not as one blended number.
6. 6Put CE next to your acquisition, retention and development spend. Firms spend an estimated 8% of GDP on these activities, yet investors rarely see the return. Ask what CE each dollar of spend adds or protects.
7. 7Track CE as a time series and use the direction as the signal. The source frames CRM data as evidence of whether the company is 'moving in a good direction' or needs to retool. A CE that shrinks while acquisition spend rises points to an unhealthy dependence on acquisition, the pattern the authors flag at Wayfair and Blue Apron.
8. 8Treat brand as a driver, not a separate ledger. A strong brand shows up as customers who pay more, stay loyal under competitive threat, and wait for your improved product. Those effects are already inside CLV, so do not add a second 'brand value' line on top.

EA shows the CE logic in practice, though the passages do not report an EA CE total. Marketing took 22% of revenue, mostly in pre-launch TV and broadcast spending with no clear way to measure returns. A small analytics team built a marketing mix algorithm that proved the spend was inefficient. EA cut marketing dollars universally with no loss of efficacy. The analytics group then found that in-game engagement was a good proxy for a player's future value and the best predictor of future game purchases. In CE terms, engagement lets you estimate each player's CLV, and summing those estimates gives the value of the player base. That gave CEO Andrew Wilson's 'Player First' mantra something measurable to aim at. The authors report that the stock rebounded 1,000% over six years and that 2017 profit passed a billion dollars. They stress that the turnaround took more than a decade to become visible. Contrast this with brand valuation: Interbrand put Apple at $184,154 million and Google at $141,703 million. The authors doubt six-digit precision and note that such figures cannot guide acquisition, retention or development decisions. A CE built from customer CLVs can.

• If a stakeholder asks for the 'value of the brand,' then answer with CE. Brand value cannot be measured precisely or used to direct tactics, while CE is quantifiable and tied to specific customers.
• If you cannot observe lifetime value directly, then find and validate a behavioral proxy (EA used engagement) before estimating CLV, and sum CLVs only once the proxy predicts future purchases.
• If CE is reported as one blended figure, then break it out by customer value tier, because customers are heterogeneous and a few best customers carry disproportionate value.
• If marketing spend rises but CE does not, then treat it as a warning of acquisition dependence and review the spend, as EA did when it cut marketing with no loss of efficacy.
• If leadership debates the discount rate, then pick a defensible one, document it, and apply it consistently, so that changes in CE reflect customers and not changed assumptions.

• Substituting a brand-valuation figure for CE. The tell: a precise-looking number (like Interbrand's six digits) that no one can link to an acquisition, retention or development decision.
• Treating CE as current revenue or past purchases. The tell: the model has no future component, so it cannot be 'present and future CLV.'
• Averaging across the base. The tell: the same offer goes to everyone, which is the product-centric habit of chasing sales 'to anyone and everyone, at any cost.'
• Ignoring discounting, or changing it silently. The tell: CE moves between reports with no change in customer behavior.
• Keeping CE inside marketing only. The tell: finance and investors never see it, even though the authors say finance can build broader buy-in than marketing and investors increasingly want customer-value data.

• ▸Customer unit and time horizon: define who counts as a customer and how far forward 'future' CLV is projected.
• ▸CLV method and data source: state how each customer's lifetime value is projected and which data feeds it.
• ▸Discount factor(s) chosen: record the rate(s) and the reason for choosing them.
• ▸Behavioral proxy (if used): name the signal (e.g., engagement) and the evidence that it predicts future purchases.
• ▸Total CE (sum of CLVs): present plus future value across all customers.
• ▸CE by value tier: show how much of the total sits with your best customers versus the rest.
• ▸Spend mapped to CE: list acquisition, retention and development spend and the CE each is meant to add or protect.
• ▸Prior-period CE and direction: note the last reported CE and whether it is rising or falling.
• ▸Brand effects observed: list any price premium or loyalty under competitive threat, and confirm they are already reflected in CLV.

1. 1Start with the gap. List what your current valuation covers: financial and operational assets. Then list what it omits: brand equity and customer value, which the source says are missing entirely from traditional approaches. Write that gap down as the scope of the customer-based valuation.
2. 2Pick the sponsor. Marketing usually pushes customer centricity first, but the authors have seen finance start it and win broader buy-in. If finance owns the valuation, put the customer-based view in its hands.
3. 3Fix the CRM foundation before modeling. Define customers consistently and give each a unique ID. Link as many purchases and activities as possible to specific customers. Be judicious about customer costs. Decide up front whether nonpurchase activities are baked into CLV. Have finance and accounting choose the discount factor(s).
4. 4Segment customers by status before computing CLV: future prospects, just-acquired customers with no history, active customers, and already-churned customers. Use different formulas for each. Only churned customers give a certain CLV, so use them to train predictions for active and future customers, and quantify the uncertainty on the rest.
5. 5Separate contractual customers (subscriptions, memberships) from noncontractual ones (online retail, coffee shops, hotels). For noncontractual customers you cannot be sure who is still active, so model that uncertainty explicitly.
6. 6Model retention by acquisition cohort, not as one aggregate rate. Aggregating new, low-retention customers with mature ones hides the rising retention that a single cohort shows. The source reports that aggregate-rate valuations understate the true customer-base value by roughly 25%–50% in standard settings.
7. 7Sum the cohort-level CLVs into a customer-equity figure and place it beside the asset-based valuation. Report both. The customer-based view supplements the traditional one and does not replace it.
8. 8Test the acquisition dependence. Check whether growth rests on continually buying new customers rather than on cohort retention and development. The source flags this 'addiction to acquisition' (Wayfair, Blue Apron) as a warning for valuation. Use cohort curves to judge whether growth periods are behind you.

The source's own illustration is the 10-period cohort in Figure 1.2. A group of customers is acquired in the same period. A constant retention rate of 76%, the average of projected retention over 10 periods, is drawn as a dotted line. The realistic cohort trend, where retention rises as weaker customers drop out, is drawn as a solid line. The constant-rate line counts far fewer surviving customers by period 10, and the gap keeps widening beyond that. Valuing a customer base this way therefore understates future value. The authors (Peter and Bruce, Marketing Science, 2010) put the understatement at roughly 25%–50% in standard settings. The cause is aggregation: new customers with low retention are averaged in with seasoned customers, so reported company-wide retention looks flat. A traditional valuation built on that flat rate captures financial and operational assets but misses this customer value. The chapter's other cases, Wayfair and Blue Apron, are cited as companies whose dependence on acquisition does not bode well for valuation. The passages give the conclusion but no figures, so use cohort curves like this one to test the same dependence in your own firm.

• If your valuation lists only financial and operational assets, then add a customer-equity line built from cohort-level CLV before you rely on the total.
• If retention is reported as one company-wide rate, then split it by acquisition cohort before using it. A flat aggregate rate probably masks rising cohort retention and understates value.
• If most growth comes from acquiring new customers, then test whether cohorts are retained and developed. Heavy acquisition dependence is a negative valuation signal.
• If a customer is noncontractual or newly acquired, then treat their CLV as an uncertain prediction. Calibrate it on churned customers, the only group whose CLV is known with certainty.
• If marketing alone cannot win buy-in, then bring finance in as co-sponsor, since finance-led efforts have created broader organizational support.

• Valuing on the 'average customer'. Tell: a single retention rate and a single CLV for everyone, with no cohort or segment breakdown.
• Using a constant aggregate retention rate. Tell: flat reported retention while individual cohorts improve, and customer-base value that looks 25%–50% too low.
• Mixing customer statuses in one formula. Tell: churned, active and prospective customers share one CLV method, and the certainty of ex-customers is applied to current ones.
• Treating contractual and noncontractual customers alike. Tell: inactive noncontractual customers are assumed either gone or still active with no uncertainty modeled.
• Mistaking acquisition-driven growth for health. Tell: revenue rises while the business depends on continually buying new customers, the Wayfair and Blue Apron pattern.

• ▸Traditional valuation (financial + operational assets): record the current figure and method.
• ▸Omitted value: note whether brand equity and customer value are included, and where.
• ▸Customer ID and definition check: confirm customers are consistently defined and uniquely identified in CRM.
• ▸Cohort list: list acquisition cohorts by period, with the number of customers in each.
• ▸Cohort retention by period: record actual retention per cohort and compare it with the aggregate rate.
• ▸Customer status split: count prospects, new customers, active customers and churned customers, and note the CLV formula used for each.
• ▸Contractual vs noncontractual: tag each segment and note how activity uncertainty is handled.
• ▸Discount factor(s): record the rate agreed with finance and accounting.
• ▸Acquisition dependence flag: state what share of growth comes from new customers versus retained and developed ones, and whether growth looks to be behind you.

1. 1Drop the 'average customer' question. Replace 'what is the average customer worth?' and 'how will the customer respond to this offer?' with 'which customers respond how, and what is each worth?' The source calls the average view outdated because it misses distinct attributes at a granular level.
2. 2Score every customer on the three forms of customer goodness: preference (how well your offering fits their needs, and how often they choose you over a competitor), propensity (likelihood to be loyal, refer others, buy higher-valued offerings), and potential (future value, and what you could maintain, enhance, or extract). The source says these are among other forms, so add others if your business needs them.
3. 3Treat these traits as largely fixed at the moment of acquisition. Customers arrive with their characteristics, and marketing and sales can develop value only 'a bit further.' Set uplift targets accordingly, and do not plan to turn most of the base into your best customers.
4. 4Use past, already-churned customers as your evidence base. Their CLV is the only one you can calculate with certainty. Mine them to build models that predict current and future customers, and keep them in segments rather than one bucket, because they are not representative of the active base.
5. 5Split analysis by acquisition cohort before computing any retention or CLV figure. Aggregate retention blends new customers (low retention) with older ones, hiding the rising retention pattern within each cohort.
6. 6Separate contractual from noncontractual customers. With subscriptions, retention is observable. With online retailers, coffee shops, or hotels, you cannot know who is still alive, so 'retention rate' is meaningless and you need a different CLV formula.
7. 7Profile the characteristics of your highest-valued customers, usually with CRM data, then find and acquire more customers who share them.
8. 8Allocate time and resources by value tier. Do not fire or ignore low-value customers. Find where you spend more attention than a customer's value justifies and trim it.

Electronic Arts shows the principle in practice. By 2013 it was named the worst company in America for the second year running. It spent 22% of revenue on marketing, mostly pre-launch TV and broadcast 'spray-and-pray' with no clear way to measure returns, and it had spent most of its history thinking about customers monolithically. A small analytics team optimized a marketing mix algorithm and showed the spend was inefficient. A standardized metrics effort followed. Under CEO Andrew Wilson's 'Player First' mantra, Global Analytics & Insights found that in-game engagement was a good proxy for the future value of individual players. It was the single best predictor of whether a customer would buy other games. In practice, that means customers who look the same on the surface are separated by an observable behavior that carries different future value. EA then designed games to attract and retain its best customers. The source reports that the stock rebounded 1,000% over six years and that profit passed a billion dollars in 2017. The turnaround took more than a decade to become visible.

• If a metric is reported as one company-wide average (retention, CLV, response rate), then break it out by cohort and segment before acting on it. The average masks the pattern.
• If the business is noncontractual, then do not use a retention rate or the subscription CLV formula. Use a model suited to uncertain customer status.
• If a segment gets the same offer and attention as every other segment, then check whether its value justifies that level. Reduce resources on low-value customers without firing them.
• If you need a leading indicator of future value, then look for a behavioral proxy such as EA's in-game engagement. Validate it against future purchasing before you rely on it.
• If you want to grow the base, then profile your best customers and acquire lookalikes. Do not expect to convert average customers into top ones.

• Calculating CLV only from aggregate retention. Tell: a flat, constant retention rate that appears across cohorts. The source reports that this underestimates customer-base value by 25%–50% in standard settings.
• Lumping all past customers into one bucket. Tell: churned-customer data is used as a single benchmark with no segment cuts, even though those customers are unrepresentative of the active base.
• Applying one CLV formula to contractual and noncontractual customers. Tell: a 'retention rate' reported for a coffee shop or online retailer.
• Treating customer centricity as better service for everyone. Tell: the strategy aims at 'the needs of our overall customer base' and the same offers go to all.
• Believing marketing can make most customers into top customers. Tell: development plans assume large uplift across the whole base, despite the Phelps analogy that much of the value is predetermined.

• ▸Segment/cohort: name the acquisition cohort or segment being scored (for example, all customers acquired in one period).
• ▸Business model: mark contractual or noncontractual, which determines whether retention is observable and which CLV formula applies.
• ▸Preference evidence: record how well your offering fits their needs and how often they choose you over a competitor.
• ▸Propensity evidence: record observed loyalty, referrals, and purchases of higher-valued offerings.
• ▸Potential estimate: state the future value, and what you could do to maintain, enhance, or extract it.
• ▸Behavioral proxy: note the leading indicator you use for future value (for example, engagement) and how you validated it.
• ▸Cohort retention pattern: record retention by period for this cohort, and compare it with the aggregate rate to see what the aggregate hides.
• ▸Current resource level vs. value tier: state the attention and spend this segment receives and whether its value justifies it.
• ▸Action: choose acquire lookalikes, develop, maintain, or reduce spend (do not fire).

1. 1Reframe the acquisition goal in writing. Replace 'maximize new customers at the lowest cost per acquisition' with 'bring in customers with the highest expected CLV.' The source calls the volume-and-cost view product-centric and the CLV view customer-centric. Acquisition is the one lever where you have the most control over who enters, because you can rarely change customers after the fact.
2. 2Drop the 'it costs X times more to acquire than to retain' rule as a planning input. The authors call it misguided advice that comes from volume and cost-minimizing thinking. Judge each acquisition channel by the value of the customers it brings in, not by its unit cost.
3. 3Profile your best customers through their goodness: preference (how well your offer fits their needs and how often they choose you over a competitor), propensity (likelihood of loyalty, referral, buying higher-valued offerings), and potential (future value). Pull this from your CRM. Do not profile them by persona or demographics, which the authors call archaic acquisition approaches to be retired.
4. 4Look at the CLV distribution before you set targets. Expect heavy concentration: roughly 80:20 in many businesses, 60:20 in grocery, 90:5 in freemium games. Your modal customer sits at the low end of CLV and a long right tail carries the value. Acquisition aimed at the average customer therefore mostly buys low-value customers.
5. 5Find a proxy for CLV that you can observe at the point of acquisition. Amazon used the affluence of book buyers as a reasonable proxy for CLV. Define yours explicitly, and keep it a proxy until real CLV data confirms it.
6. 6Choose acquisition channels and tactics that find more customers like your best ones, and cut broadcast-style spending you cannot tie to a return. Test the mix with data. EA's marketing-mix analysis showed that its pre-launch broadcast spending could be cut with no loss of efficacy.
7. 7Set a measurement goal for every acquisition dollar before you spend it. EA spent 22% of revenue on marketing with no clear goals for measuring returns. Define the expected outcome and who counts as a 'good' customer, then compare cohorts acquired through each channel on preference, propensity, and potential.
8. 8Check non-financial value too. If CLV in your business includes non-financial activity, include it in the acquisition profile, since most CLV calculations are purely financial and that is an omission, not a rule.
9. 9Feed results back into the profile. Re-run the best-customer profile as the CRM fills with data, then adjust targeting.

Amazon in 1994 is the clearest case of acquisition strategy built on customer value. Books are heavy, costly to ship, and low-margin, and Americans are not known for being bookish en masse, so the product made little sense on its own economics. Bezos's logic concerned the buyers. People who buy books are relatively affluent, which makes them a reasonable proxy for high CLV, and selling to them offered a rich source of data on the behavior and untapped potential of a highly valuable online segment. With that insight Amazon found and acquired throngs of other online customers with similarly high lifetime values. The contrast is EA, which spent 22% of revenue on marketing, mostly just before launches on TV and other 'spray-and-pray' broadcast, with no clear goals for measuring returns. A marketing-mix algorithm showed the spending was inefficient, and the firm cut marketing dollars with no negative impact on efficacy. The pattern is the same in both: Amazon chose acquisition targets by value, and EA replaced undirected volume spending with measured spending.

• If a proposed channel is justified mainly by low cost per acquired customer, then require evidence on the expected CLV of customers it brings in before you fund it.
• If your best-customer profile is built on personas or demographics, then rebuild it from preference, propensity, and potential observed in your CRM.
• If you cannot state how an acquisition dollar's return will be measured, then treat the spend as unproven and test cutting it, as EA did with broadcast spending.
• If value is highly concentrated (80:20, 60:20, 90:5), then aim acquisition at the characteristics of the right tail, not the modal customer, and do not 'fire' low-value customers on a guess, because choosing accurately is hard and mistakes are costly.
• If a retention-first argument ('keep the ones you have, since acquiring is costlier') is used to cut acquisition, then reject it as a cost-minimizing view and ask what kind of customers each path produces.

• Optimizing for volume and cost per acquisition. The tell: success is reported as customer counts and CPA, with no CLV by acquisition source.
• Targeting by persona or demographic. The tell: the acquisition brief describes 'who they are' (age, segment) rather than how they behave and what they are worth.
• Treating a proxy as truth. The tell: an affluence-style proxy is never validated against realized CLV, as the Amazon-style proxy was only 'reasonable'.
• Spray-and-pray spending with no return goals. The tell: budget is concentrated around launch moments and no one can say what the spend was supposed to return.
• Overcorrecting by pruning low-value customers. The tell: plans to cut customers appear before the CLV distribution has been modeled, though the modal customer sits at the low end and the right tail is long.

• ▸Channel/tactic: name the acquisition route being evaluated.
• ▸Current success metric: record how it is judged today (volume, cost per acquisition, other).
• ▸CLV proxy used: state the observable proxy for value at acquisition and its evidence level (assumed or validated).
• ▸Preference signal: note how well the offer matches the needs of customers this channel brings in and how often they choose you over competitors.
• ▸Propensity signal: note loyalty, referral, and upgrade behavior of cohorts from this channel.
• ▸Potential signal: note estimated future value of this channel's cohort, including any non-financial value.
• ▸Similarity to best customers: rate how closely this cohort matches your top-tail customer profile from the CRM.
• ▸Return measurement: write how and when the return on this spend will be measured.
• ▸Decision: scale, test a reduction, or stop, with the reason.

1. 1Segment customers by CLV before choosing any tactic. Tier them (the source's metal tiers run from lead and iron up through the middle tier to platinum) using current value, potential, and the other 'forms of goodness'. A tactic assigned before the tier exists is a shotgun tactic.
2. 2Build the Figure 3.2 2×2 with two axes: offense vs. defense, and high-value vs. low-value customers. Place each tactic in a cell: premium offerings = offense/high value; loyalty programs = offense/lower value; strategic account management = defense/high value (white-glove support); customer service = defense/lower value (low-touch assistance).
3. 3Inventory every live retention and development program (loyalty, premium tiers, account teams, service channels, cross-sell and up-sell campaigns). Write down for each one which customer tier it actually reaches and whether its stated goal is to grow value or to protect it.
4. 4Map each program onto its cell and compare against the tier it serves. Flag mismatches, such as premium offers pushed at lead-tier customers or white-glove account coverage spent on low-value accounts.
5. 5Check your investment mix by quadrant. The source's test: lay out the strategy and see whether you are over- or underinvested in a given quadrant, and whether each tactic is the right 'card to play' at that moment.
6. 6Set each program's goal to match the CLV of its target customers. For lead-tier customers, aim for modest value extraction (a loyalty program that might nudge a lead customer up to iron). Do not expect platinum, and do not ignore them, per the Paradox of Customer Centricity.
7. 7Note spillovers deliberately. A loyalty program can also lock in doubtful customers, which is defense. Record its primary role and its secondary role so it is judged against the right goal.
8. 8Start simple and assign an owner to each quadrant to create alignment and accountability. Refine toward 'shades of gray' tactics only after the basic 2×2 is working. Test CRM-supplied CLV and retention predictions for each segment before relying on them to place customers.

The source's own illustration is the lead-tier customer. A lead customer cannot be 'magically alchemized' into platinum, but the Paradox of Customer Centricity says you still retain and develop some lower-value customers to balance the business. Suppose a firm runs a classic 'buy 9, get 1 free' program and a premium offering, and finds the premium offer is mostly pitched at its lowest tier. Placing the tactics on the 2×2 exposes the mismatch. The premium offering is an offensive tactic that works best on already-high-value customers, so it moves to the top tier. The loyalty program stays aimed at lead customers, where it squeezes out a little more value and could lift some to iron. The loyalty program also keeps wavering customers through points and extras, so it is credited as defense too, though that lock-in is not enhanced value. Platinum accounts get strategic account managers for white-glove defense, and lower tiers get low-touch customer service. The firm then checks which quadrant it is overinvested or underinvested in and assigns an owner to each quadrant.

• If a customer is high value and the aim is to grow profit, then use an offensive premium offering. If the aim is to protect the relationship, use a strategic account manager with white-glove support.
• If a customer is lower value, then default to low-touch customer service for defense and a loyalty program for modest offensive value extraction. Do not assign account managers or premium pushes.
• If a customer sits in the lead tier, then do not ignore them and do not overinvest. Use inexpensive tactics like loyalty programs, aiming at iron rather than platinum.
• If a program serves two roles (a loyalty program is both offense and defense), then name its primary goal first and judge it against that, treating lock-in as a secondary benefit rather than enhanced value.
• If your quadrant mix shows heavy spend in one cell and little in another, then rebalance toward alignment with CLV before adding new tactics.

• Treating cross-selling and up-selling as one-size-fits-all development. The tell: the same offer goes to every tier, regardless of CLV.
• Over-obsessing on acquiring top customers and neglecting lower-value ones the business needs for balance (the Paradox). The tell: lower tiers get no defined tactic at all.
• Reading loyalty programs only as value creation and ignoring their defensive lock-in role. The tell: the program's ROI is judged on uplift alone while it is really retaining wavering customers.
• Running a disorganized jumble of shotgun tactics with no owner. The tell: nobody can say which quadrant a program belongs to or who is accountable for it.
• Trusting CRM tier assignments without testing them. The tell: predicted CLV or retention rates for a segment are far off actuals, yet tactics are still allocated from them.

• ▸Program/tactic name: list each live retention or development program.
• ▸Primary role (offense or defense): state whether it mainly grows value or protects it.
• ▸Secondary role: note any spillover, such as loyalty lock-in.
• ▸Target CLV tier: lead, iron, middle, or platinum, as actually reached today.
• ▸Intended CLV tier: the tier this tactic should serve under the 2×2.
• ▸2×2 quadrant: premium offering, loyalty program, account manager, or customer service.
• ▸Program goal vs. tier: write the value goal and check it fits the tier (e.g., lead → modest uplift).
• ▸Over/under-invested? Mark the quadrant's current investment level against the others.
• ▸Owner and accountability metric: name the person and the KPI for this quadrant.

1. 1Start from the strategy, not the software. Write down the customer-centric questions the CRM must answer: the CLV of a customer, retention rates, and acquisition performance for a particular segment. The source says a customer-centric strategy only works if it is closely aligned with your CRM solution, so each question becomes a stated requirement.
2. 2Own the scorecard. The CRM will not supply all the insight a customer-centric strategy needs, so you must develop the scorecard and KPIs that measure how well you are developing your firm's customer value. Assign a named owner on the business side, not the vendor side.
3. 3Run a gap audit. For each scorecard metric, mark the data your CRM can supply and the data it cannot. The source expects both 'tiny and yawning gaps.' Log each gap with the decision it blocks. Cover acquisition too: check which direct tactics (known prospects) and indirect tactics (unknown prospects) you can connect to a customer record.
4. 4Take the gap list to the vendor. Business executives, not IT alone, challenge CRM providers to close the gaps. Confirm in writing that the vendor understands the goals of your customer-centric plan and your expectations for supporting them. If either is unclear, write a practical plan that closes the gap.
5. 5Back-test the CRM's assertions. Take its current predictions of CLV, retention rate, and acquisition for a particular customer segment and compare them with what customers actually did. If the comparison is way off, call the vendor and find out why. Do not accept CRM analysis on blind faith.
6. 6Build the supporting processes and the leadership role. Gene Marks's review of 11 'terrible' CRM systems found that the systems were not the problem. The organizations had not set up processes to support the CRM initiative, and leadership took a passive role. Name an executive sponsor and define who reviews CRM outputs and acts on them.
7. 7Run the work in short iterative cycles. Use sprint-style planning, build, and review with stakeholders. Each review checks the CLV models and scorecard against what the CRM delivers, and the next cycle's discovery is informed by it. Budget for a sequence of cycles, because a Monday rollout does not produce answers by Tuesday.
8. 8Keep CLV cross-functional. Feed CRM-derived CLV into acquisition, retention, and development decisions together, in line with the manifesto's 'cross-functional uses of CLV over siloed applications.' Avoid one team's private version.

The source's real case is Gene Marks's Forbes article, '11 Terrible CRM Systems for Your Company.' He looked at 11 systems and concluded that none of them was the terrible part. The organizations had failed to establish supporting processes, and leadership stayed passive. Applied here: a firm has spent five or more figures on a CRM and finds that it falls short. The first move is not to replace the tool. Run the audit above. (1) List the CLV, retention, and segment-acquisition questions the strategy needs answered. (2) Mark which the CRM answers and which it cannot. (3) Back-test its segment predictions against actual customer behavior. (4) If the predictions are way off, call the vendor with the specific gaps and your expectations. (5) Assign an executive owner and a review process, so the fix does not depend on a passive sponsor. The figures stay as placeholders until you run the test on your own data. The source gives no numbers for this case.

• If the CRM's CLV, retention, or segment-acquisition predictions are way off actual outcomes, then call the vendor and diagnose the cause before anyone acts on the numbers.
• If a scorecard metric cannot be produced from the CRM, then log it as a gap and have the executive owner escalate it to the vendor. Do not drop the metric.
• If the CRM underperforms but no one owns its processes or reviews its output, then fix leadership involvement and process first. The Marks finding says the system is probably not the problem.
• If the vendor cannot restate your customer-centric goals and expectations, then write a practical plan to close that gap before expanding the rollout.
• If you expect full answers right after go-live, then reset the plan to iterative cycles with reviews. Answers accumulate over sprints.

• Treating CRM rollout as the finish line. The tell: no scorecard or KPIs exist after launch, and people ask why the system 'doesn't tell us anything yet.'
• Accepting CRM analytics on faith. The tell: no one has compared predicted CLV or retention with actual segment results.
• Blaming the tool for an organizational problem. The tell: the team wants to switch vendors, yet no processes, owners, or sponsor involvement have been defined. This is the pattern Marks found.
• Leaving the vendor unbriefed. The tell: the vendor cannot explain how its outputs support your CLV-based strategy, and gap requests are informal or absent.
• Siloed use of CRM data. The tell: marketing, sales, and service each run their own customer metrics, and no shared CLV view feeds acquisition, retention, and development decisions.

• ▸Strategy question: the customer-centric question the CRM must answer (CLV, retention rate, segment acquisition).
• ▸Scorecard metric / KPI: the measure of customer-value development tied to that question.
• ▸Metric owner: the named business-side person accountable for it.
• ▸CRM can supply? (Yes / Partly / No): record what the system provides today.
• ▸Gap description: what data or analysis is missing, and the decision it blocks.
• ▸Vendor request and date: the specific ask sent to the CRM provider, with the expectation stated.
• ▸Back-test result: CRM prediction vs actual for a named segment, and whether the difference is acceptable.
• ▸Supporting process and executive sponsor: the review routine and leader who keep this metric acted on.
• ▸Next sprint review date: when the metric and CLV model are next checked.

1. 1Write the destination as a SMART plan before any sprint starts. Each goal must be specific, measurable, achievable, relevant and time-limited, with incremental milestones. 'Become customer centric' fails the test. A goal tied to the customer-centric concepts in the book, such as CLV-based decisions in acquisition, retention and development, can pass it.
2. 2Baseline the five scaffolding dimensions: mission, culture, leadership, structure and processes. For each, record how aligned it is with customer centricity and how mature it is. This gives you the backlog of change work, and it shows you which dimension is the weakest.
3. 3Choose the cycle framework your team already knows: Deming Cycle, Agile, Lean Startup or ITIL's Continuous Service Improvement. The source says they overlap heavily, so familiarity matters more than the brand. The one non-negotiable is cycle length: weeks, not months or years.
4. 4Open each cycle with discovery and planning. Define a concise set of deliverables, each aimed at one of the five dimensions and tied to a SMART milestone. Keep the list short enough to finish inside the cycle.
5. 5Action the deliverables, then quality-check them inside the same cycle. A deliverable is not done until it has been checked. Keep the work lean and cut busywork, in the spirit of the Agile manifesto's preference for individuals and interactions over processes and tools.
6. 6Close each cycle with a reflection session that includes the project stakeholders. Ask what happened, what moved the milestone and what the next cycle should change. Use the answers to adjust priorities and refine deliverables.
7. 7Leave breathing room in every plan for deliverables nobody anticipated. The source calls new deliverables 'inevitable'. A cycle packed to 100% forces you to drop either the new item or the quality check.
8. 8Recruit cross-functional participants from the start, not only the C-suite. Transformation does not always begin at the top, as EA's story shows. Put people from the layers the change touches into planning and reflection, so buy-in forms during the cycles rather than after a launch.

The source shows the outcome of a customer-centric change, not its sprint log, so this example separates the two. The real outcome is Best Buy. It looked headed for the fate of CompUSA, Circuit City and Radio Shack, then pivoted to a customer-centric strategy. Its stock rose from $10 a share in December 2013 to more than $70 at the close of 2017, including a 50% gain in 2017 alone. The passages do not describe Best Buy's internal cycles, so what follows is an illustration of the method, not a claim about what Best Buy did. Suppose the baseline shows processes are the least mature dimension. Cycle 1 (a few weeks): discovery defines two deliverables, a draft definition of 'best customer' and a list of the decisions that currently ignore it. Both are built and checked, and in the reflection stakeholders add a third item nobody planned, which goes into Cycle 2's reserved slack. Each pass repeats the plan, action, check, reflect loop, so the organization corrects course in weeks rather than discovering a wrong turn after a multi-year program.

• If a goal cannot pass the SMART test, rewrite it before planning a cycle, because an unmeasurable goal cannot be corrected between cycles.
• If a cycle plan runs longer than a few weeks, split it. The framework's value comes from short cycles and frequent course correction.
• If reflection surfaces a new deliverable, add it to the next cycle's reserved slack and re-prioritize. Do not squeeze it into the current cycle.
• If the weakest of the five dimensions (mission, culture, leadership, structure, processes) is also the least represented in the room, add its stakeholders before the next cycle.
• If your team already runs a change method it knows, adapt that one rather than adopting Agile because the book spotlights it.

• Treating change as a C-suite event. Tell: the only people in planning and reflection are executives, and middle layers learn the plan from an announcement.
• Cycles stretched into months. Tell: a 'sprint' that needs a long steering-committee review, so course corrections arrive after the damage.
• Skipping the reflection phase. Tell: the next cycle's plan looks identical to the last one, and no priority or deliverable ever changes.
• Planning with no slack. Tell: new requests are either refused or silently dropped, and quality checks get cut when anything unexpected appears.
• Changing processes while leaving mission, culture and leadership alone. Tell: new workflows exist, but decisions and incentives still follow product-centric logic.

• ▸Cycle dates and length: state start and end; it must be weeks, not months.
• ▸SMART goal for this cycle: specific, measurable, achievable, relevant and time-limited, linked to a milestone in the larger plan.
• ▸Dimension targeted: mission, culture, leadership, structure or processes; add its current alignment and maturity.
• ▸Deliverables (concise list): each one small enough to be actioned and quality-checked within the cycle.
• ▸Quality check per deliverable: say how and by whom each is verified as done.
• ▸Reserved slack: capacity held for unplanned deliverables.
• ▸Stakeholders and cross-functional participants: name who is involved from outside the C-suite.
• ▸Reflection notes: record what worked, what did not and what stakeholders said.
• ▸Next-cycle adjustments: list changes to priorities, refined deliverables and newly added items.