Customer Centricity Playbook for Executives: Building Value in the Age of Personalization
Deciding to become customer-centric is straightforward. Doing it involves a specific sequence, and the order matters more than any individual step, because most programmes fail by starting in the wrong place.
This is the implementation companion to what customer centricity actually means, which covers the definition and the test of whether the approach applies to a given business. What follows assumes it does.
The framing owes a good deal to Peter Fader and Sarah Toms in The Customer Centricity Playbook, with the sequencing drawn from where these programmes tend to stall in practice.
Start with the value model, not the platform
The most common opening move is selecting technology. A year later the data is integrated, the platform is configured, and nobody has decided which customers matter.
Start instead with a per-customer view of contribution. It requires three things.
Identity resolution. Knowing that these transactions belong to one customer. Harder than it sounds where records are split across systems, and the ordinary answer is a manual matching exercise on the largest accounts rather than a technology project.
Margin, not revenue. Direct cost allocated to each customer, so the view is contribution rather than turnover.
Cost to serve. Discussed below, and the step that most often changes the answer.
A first version can be built in a spreadsheet with existing transaction data in four to eight weeks. It will be imprecise and it will still be more informative than anything currently in the monthly pack.
Cost to serve
Ranking customers on revenue is the default and it is frequently wrong, because the heaviest revenue often arrives with the heaviest cost: more support tickets, more custom work, more account management time, deeper discounts, slower payment.
Allocating that cost does not require a full activity-based costing exercise. A workable approximation:
- Support cost per customer, from ticket volume multiplied by an average handling cost.
- Account management time, from a rough allocation of who spends time where.
- Custom or implementation work delivered without separate charge.
- Discount, treated as a cost rather than as absent revenue.
- Working capital cost where payment terms differ materially.
Doing this typically moves a meaningful share of the apparent top tier downwards and lifts quieter accounts up. Any programme skipping this step is optimising against a ranking nobody has verified.
Segment on value, not on demographics
Two axes: current value and potential value. Four groups, each needing different treatment.
High current, high potential. Growth investment. Named ownership, roadmap influence, first access.
High current, low potential. Defend and serve efficiently. These customers are already at their ceiling, and the goal is to keep them at a cost that reflects that.
Low current, high potential. The most commonly missed group, and the one where analysis earns its cost. Customers whose spend is small relative to what their profile suggests they could spend.
Low on both. Serve at a cost matching contribution, which usually means self-service rather than exit.
Potential is the harder axis and is estimated rather than measured: from the size of the customer, their spend with comparable suppliers, the share of their category you currently hold. Approximation is acceptable. The distinction between someone at their ceiling and someone with room is more useful than precision about either.
Acquisition
Set cost ceilings per segment. A single blended target either overpays for customers who will never be worth it or underpays for those who would.
Judge channels on value delivered, not volume. It is common for the channel with the lowest cost per acquisition to produce the lowest lifetime value. In aggregate reporting that channel looks like the best performer, and budget flows to it for years.
Target on resemblance to good customers. Once the top segment is identified, the useful question is what its members had in common at the point of acquisition: which channel, which offer, which initial purchase, which industry. Those characteristics are the targeting brief.
Track the value of each cohort as it matures. The judgement on an acquisition campaign is not available at the point of conversion, and most organisations close the file there.
Retention
A note on the number everyone quotes. The claim that acquiring a customer costs five times more than retaining one appears constantly and has no reliable source; the ratio varies enormously by business and is not a constant. The underlying logic holds without it: in a repeat-purchase business an existing customer usually costs less to sell to, and retention improvements compound while acquisition improvements do not. Use the reasoning, and calculate your own ratio if a number is needed.
What actually moves retention:
Early life. The largest single lever in most businesses. A customer who reaches first meaningful value quickly retains at a materially different rate from one who does not, and this is a design problem in onboarding rather than a marketing one.
Early risk detection, with a caveat. Predictive churn models identify risk reasonably well and explain causes badly. A model flags that an account resembles others that left and cannot say whether the cause was price, a service failure or a change of contact, so the intervention is guesswork. They earn their place in high-volume consumer settings where responses are cheap and testable. In smaller portfolios a conversation with the account manager outperforms the model.
Serving the top segment properly, which means the loss of a major account should never come as a surprise, and it frequently does.
What does not work is discounting as the standard response to churn risk. It buys a renewal without addressing the cause, teaches customers that threatening to leave produces a discount, and permanently lowers the value of an account retained precisely because it was valuable. The cost appears in next year's numbers.
Organisation
One owner. A single executive accountable for the value model with a place in resource allocation. Owned by marketing it becomes a campaign input; by finance, a reporting exercise; by nobody, analysis that changes nothing. The reporting line matters less than whether the person can influence where money goes.
Customer-based reporting. Alongside product and region, report by cohort: acquired value, retention by cohort, contribution development over time. The question of what a cohort acquired two years ago is now worth is answerable and more informative than most of the monthly pack.
Incentives that match. Where sales is paid on new revenue alone, the acquisition of low-value customers continues regardless of what the analysis says. Some part of variable pay has to reflect the value of what was brought in.
Technology, last
A customer data platform becomes worthwhile once the value model exists, segmentation is being acted upon, and the manual effort of unifying data has become the binding constraint.
Bought earlier, it is an expensive way to assemble data nobody has decided how to use.
The order is model, decide, act, automate. Reversed, the usual result is a well-integrated dataset and an unchanged allocation of resources.
A first ninety days
- Weeks one to four. Assemble per-customer contribution from existing data. Resolve identity on the largest accounts manually. Produce the distribution.
- Weeks three to six. Allocate cost to serve, approximately. Re-rank. Note which customers moved and by how much.
- Weeks five to eight. Estimate potential and build the four-quadrant view. Test it with the people who know the accounts, who will disagree usefully.
- Weeks seven to ten. Choose three decisions to change: an acquisition ceiling, a service level, a roadmap weighting. Three, not a transformation programme.
- Weeks nine to twelve. Establish cohort reporting and agree who owns the model and refreshes it.
At the end of that, the organisation has a value model, a segmentation, three changed decisions and a reporting line. That is a foundation. It is not a transformation, and calling it one at this stage is how these programmes acquire the reputation they have.
How it fails
- Technology first. Covered above, and it remains the most expensive version of failure.
- Analysis with no decision attached. A deck everyone admires and no budget moved.
- Refusal to differentiate. The organisation accepts that customers differ in value and will not act on it, which is a decision to decline the strategy.
- Revenue-based ranking. Value calculated without cost to serve directs investment at expensive customers with full confidence.
- Built once. Customer value shifts. A model never refreshed directs this year's spending using last year's customers.
- Called a transformation. Setting expectations at transformation scale for what is initially a reporting change guarantees disappointment at the first review.
The point
The sequence is: build the value model, allocate cost to serve, segment on current and potential value, change a small number of real decisions, report by cohort, and automate only once all of that is running.
Most of the difficulty is not analytical. It is the willingness to act on the conclusion that customers are not equally valuable, which requires someone with authority over budget to decide that some accounts will receive less than they do now.
