Why the generic STP does not transfer to retention
The framework page treats segmentation as a way of carving up an outside market, and its example splits prospective clients by need before a website rebuild. Retention changes the population and the purpose. The population is known individually, often down to every order, which removes the guesswork about who is who but adds a duty to use that data lawfully. The purpose is no longer to choose a market but to decide where a limited budget for discounts, service time and messages should go.
That shift changes what a good segment looks like. A retention segment does not need to be large or reachable through media; it needs to behave differently from its neighbours when treated differently. And the targeting step gains an outcome the market-facing version barely mentions: deliberately spending nothing on a group.
Decisions that change for retention
| STP step | In the generic sequence | For retention | Decision it forces |
|---|---|---|---|
| Segmentation | Needs, demographics or firmographics across a market | Recency, frequency, value, category mix and trend from first-party records | Which held variables predict staying |
| Targeting | Choose segments to enter and a coverage pattern | Rank tiers by value at risk; name the tiers to let go | Where retention spend stops |
| Positioning | Why this offer beats alternatives for a newcomer | The reason to stay, for someone who knows the product | What each tier values enough to remain |
| Reach | Media and channels that find the segment | Owned channels and audiences limited by consent | Which tier may be contacted on which channel |
| Maintenance | Re-run at a planning cycle | Tiers recalculated on a fixed rhythm | How often customers are re-sorted |
Take segmentation variables from behaviour and value already held
The business knows what each customer bought, when, how often, through which channel and at what margin. Those fields predict retention far better than age or postcode. A practical starting split combines how recently someone bought, how often, and the gross margin they bring, then adds one variable specific to the category, such as subscription status or the breadth of categories purchased. The email segmentation guide covers the lifecycle and engagement fields worth collecting; here the point is choosing which few of them define the strategic tiers.
The trade-off is precision against usability. Dozens of micro-segments can be computed, but nobody can write distinct messages for them. Choose the smallest number of tiers that would each receive a genuinely different treatment, and add the trend in spending as a flag so that a valuable customer who is quietly fading is not filed with stable ones.
Decide which customers are worth a retention effort and which to let go
Targeting on the base is an economic ranking. Estimate the value at risk in each tier, the expected lifetime value (LTVTotal value a customer brings over time.) that leaves if those customers leave, and set it against what keeping them would cost. Some tiers will come out negative: customers who buy only at clearance prices, who return a large share of orders, or whose service needs exceed their margin. For those, the decision is to stop active retention spend, not to stop serving them.
The choice is between coverage and return. Spreading save offers across everyone feels safe but pays discounts to people who would have stayed anyway. Concentrate effort where value at risk is high and the customer is still reachable; accept that letting go of a low-value tier will lower the headline retention figure. The customer retention strategy guide explains the diagnosis that tells a fixable departure from an inevitable one.
Position the reason to stay
A retained customer already knows what the product does, so acquisition messaging is wasted on them. The positioning question becomes what each tier would lose by leaving. For a high-value, broad-basket tier it may be convenience and the effort of finding a substitute supplier; for a single-category tier it may be a specific product they cannot get elsewhere; for a lapsing tier it may be a reminder of a problem the product solved. The jobs-to-be-done framework helps here, because customers stay for the job, not the brand.
The trade-off is between one renewal message and several. A single "thank you for being a customer" line is easy to govern and says nothing to anyone. Write a short reason-to-stay statement per targeted tier, and keep the number of tiers small enough that each statement is actually used.
Respect consent and contact rules per segment and channel
Not every customer can be reached on every channel. Marketing email and text messages generally need the right permission, uploading customer lists to ad platforms for RemarketingAdvertising to previous visitors or engagers. raises its own notice and lawful-basis questions, and people who have objected must be suppressed everywhere. The UK regulator's direct marketing guidance cited below is the reference for the rules. In practice, each tier splits again into contactable and non-contactable members by channel, and the strategy has to work for both.
The decision is how to treat a valuable customer who cannot be messaged. Options include service-led retention at the next order, packaging inserts, account-page content and, where permitted, the custom audiences approach on Meta. Choose the channel the customer has permitted, even when a stronger channel is technically available.
Refresh segments on a schedule
Customers move. A loyal buyer lapses, a lapsed one returns, a single-category customer starts buying across the range. Tiers that are calculated once and left alone send win-back offers to people who ordered last week. Set two rhythms: a frequent recalculation that re-sorts customers, matched to the purchase cycle, and a slower review of the tier definitions themselves. The e-commerce email and SMS flows guide shows how triggered flows can respond to tier changes between scheduled runs.
Decision points in a worked STP for retention
A pet-supplies retailer sorts its base into retention tiers
Hound & Harbour, an invented online pet-supplies retailer, has 42,000 customers who bought in the past 18 months. The team segments on recency, order frequency and gross margin, with a flag for subscription to repeat food deliveries.
| Tier | Definition | Customers | Margin per year | Targeting decision | Reason to stay | Contactable by email |
|---|---|---|---|---|---|---|
| Pack leaders | Subscribed, 6+ orders a year, food and accessories | 5,200 | £96 | Protect: priority service, early access | Never running out, one supplier for everything | 81% |
| Steady feeders | Subscribed, food only | 11,800 | £41 | Develop: cross-category prompts | Vet-recommended diets in stock reliably | 74% |
| Drifting | Bought 3+ times, none in the past 90 days | 9,400 | £38 at last rate | Win back once, then review | The specific diet they used to order | 62% |
| Deal-only | Bought only during sales, high returns | 15,600 | £4 | Let go: no save offers | None written | 55% |
Before STP, the retailer sent the same 15% discount to everyone who had not ordered for 60 days, costing £31,000 a quarter. After the split, deal-only customers receive no retention offer, drifting customers get a single message naming their usual product, and pack leaders get service perks rather than discounts. Tiers are recalculated weekly; the definitions are reviewed each quarter.
The table shows the decisions at work. Value-based variables separated the deal-only group, which looked active by order count but contributed almost nothing. The let-go decision removed the largest cost line. The reason-to-stay column forced different messages, and the consent column showed that a fifth or more of every tier needed a non-email route.
Measuring a retention STP: KPIs and cadence
The primary KPIA metric tied directly to an objective. is retained value in the targeted tiers, not overall retention, because the strategy deliberately accepts losses in the let-go tier. Beside it sit tier migration (how many customers moved up, held or slipped down each period), the cost of retention per tier, and the share of each tier reachable on at least one permitted channel. A holdout group within each targeted tier, receiving no special treatment, shows whether the effort changed behaviour or merely coincided with it. Migration can be read on the recalculation rhythm; retained value and cost need a full purchase cycle and belong in a monthly or quarterly review. Audience features in analytics tools, such as the Google Analytics audiences cited below, can mirror the tiers for on-site personalisation where consent allows.
Where a retention STP misleads
Behavioural tiers describe what customers did, not why, so a drifting tier may contain people who moved house, lost a pet or found a cheaper rival, each needing a different response. Value estimates lean on past margin and can undervalue young customers who would have grown. The let-go decision can become an excuse to ignore service failures that pushed customers into low tiers. And holdouts are often skipped because they feel like lost revenue, which leaves the strategy unable to prove anything.
Common mistakes
- Segmenting on demographics because they are in the CRM. Behaviour already held predicts staying far better than age bands.
- Saving everyone. Blanket discounts reward customers who would have stayed and keep unprofitable ones.
- Reusing acquisition copy. Existing customers need a reason to stay, not a pitch for a product they own.
- Ignoring consent by channel. A tier is not one audience if half of it cannot be emailed.
- Calculating tiers once. Static segments send the wrong message within weeks.
