Updated: 10 hours ago
Is Your Retention Programme Improving Customer Lifetime Value?
By Trayan Hristov · Published 27 October 2026

Your retention report can look healthy while the economics underneath it weaken.
More customers place another order. More revenue gets attributed to email. More campaigns go out. Yet discounts absorb the margin, customers need another offer to return, and the business earns less from each relationship.
That is how a retention programme can quietly damage customer lifetime value. The problem sits in the decisions the reporting encourages.
The title is deliberately sharp. A disappointing result alone does not establish that an agency caused harm. The useful question is whether its choices are improving customer value after the relevant costs, and whether you have evidence strong enough to tell.
Check the gross margin on the second order
Take two first-order cohorts acquired through the same offer. For each, count customers making a second purchase within the same window, deduct returns, fulfilment and discounts, then compare contribution per original customer. Ask the account lead whether the next planned promotion still makes sense. A higher repeat rate can coexist with lower profit when repeat orders depend on a deeper discount.
Start by defining the value you want to grow
Revenue LTV describes revenue per customer over a stated horizon. For an operating decision, I would also examine contribution per customer: what remains after the variable costs of generating and fulfilling those sales.
For a 90-day review, use:
90-day contribution per acquired customer = total contribution from qualifying orders in each customer’s first 90 days, less the agreed programme costs, divided by eligible first-time customers with a complete 90-day window.
Define contribution consistently. Start with sales after discounts and refunds, then account for product costs, fulfilment, shipping subsidies, payment fees and other relevant variable costs. If discounts or refunds are already included in net sales, do not subtract them again.
Allocate messaging and intervention costs consistently. Show agency and software costs separately or include an agreed allocation. State which approach you used. To assess acquisition payback, also subtract customer acquisition cost without counting it twice.
A 90-day measure is an observed window. It is not a forecast of a customer’s entire lifetime. It is useful because you can compare it without waiting years.
1. Discounts can buy repeat orders that contribute less
A promotion may bring back customers profitably. It may also subsidise purchases that would have happened at full price or shift demand into the sale window.
The concern is a programme that reaches for another discount without testing who needs it, what it costs and what happens afterwards.
Here is a hypothetical example with two comparable groups of 1,000 first-time customers. Assume first-order contribution is identical, every returner makes one second order within 90 days, and there are no refunds:
Full-price follow-up: 240 second orders at $120 generate $28,800 in sales. At $64 variable cost per order, contribution is $56 per order, or $13,440.
Discount-led follow-up: 300 second orders at $96 generate the same $28,800 in sales. At $63 variable cost per order, contribution is $33 per order, or $9,900.
The second-purchase rate rises from 24% to 30%, but second-order contribution falls by $3,540, before programme costs or later purchases.
These are illustrative assumptions, not a benchmark or a client result. The discount-led approach could still win over a longer horizon if it creates enough valuable later purchases. Measure that possibility rather than assuming it.
A better brief: compare an offer with a relevant alternative, define eligibility, and evaluate contribution and later behaviour alongside conversions.
2. Attribution can reward claiming a sale without showing what changed
A customer can receive several marketing messages and still purchase for reasons that the report cannot observe.
Klaviyo’s documentation explains how it assigns credit using eligible interactions within configured attribution windows. A higher attributed total does not, on its own, establish how many additional purchases the programme created.
This becomes a problem when the team optimises entirely around the credited number. More contact with people already likely to buy can make the channel report look stronger without improving the underlying customer relationship.
Keep attribution as one view. Read it beside store outcomes, equal-age customer cohorts and appropriately designed experiments.
A better brief: “Which customer behaviour should change, and what comparison would help us judge whether the intervention caused it?”
When feasible, randomly assign eligible customers to the current experience and the proposed change. Keep assignment stable, define the outcome in advance and allow enough time and customers for a useful result. An underpowered holdout can remain inconclusive.
3. Generic post-purchase selling can miss the reason customers fail to return
The period after the first order is where the customer discovers whether the product delivers.
If they are confused about setup, unsure about fit or disappointed with the routine, another bestseller email may leave the actual obstacle untouched.
Map the experience for a specific first purchase:
What must happen for the customer to get value?
What concern might stop them using or keeping the product?
What support, education or proof would help?
What is a relevant next purchase, and when could it make sense?
For apparel, this might mean care, fit confidence and styling before a complementary item. For a consumable, it might mean usage guidance and a reminder based on realistic depletion. For durable equipment, support and suitable accessories may matter more than another main-unit purchase.
The agency cannot fix every product or service problem. It should identify what the messages depend on and bring the right issue back to your team.
A better brief: build follow-up around the first product’s success, then test the next useful step. Our Progressive Personalisation framework explains how treatment can change as the relationship develops.
4. Averages can hide deterioration in the customers you are acquiring now
A growing pool of long-standing customers can support returning revenue while recent first-time buyers become less likely to return. Rapid acquisition can also depress a blended repeat rate simply by adding customers who have had little time to buy again.
Both patterns make the same headline difficult to interpret.
Shopify’s customer cohort report groups customers by first-order date by default. That provides a starting point for examining how acquired groups develop over time.
For a fair 90-day comparison, include customers who have each completed 90 days of follow-up. Define qualifying orders and customer identity consistently. Compare second-purchase rate, order timing, net revenue and contribution per original customer.
Then investigate meaningful differences such as first product, acquisition context or offer exposure. Those comparisons provide hypotheses; observational differences do not prove that a product or channel caused higher LTV.
A better brief: show the customer count, definition and observation window beside every retention rate. Our first-to-second purchase guide gives a worked calculation.
5. Contact decisions can spend the relationship faster than they develop it
Frequency should follow the audience, the reason to send and the evidence.
A programme can become expensive when campaigns and automations repeatedly compete for the same person’s attention. A customer may receive overlapping offers, an irrelevant product prompt or another message after already completing the intended action.
Review the combined experience across channels. Look at actual profiles, the timing of messages and the rules that make people eligible.
Use clicks, purchases, unsubscribes, complaints, delivery trends and customer progression together. Opens alone are an incomplete view of customer interest. Revenue per delivery can help diagnose sending efficiency, but it does not capture every message’s purpose or establish incremental profit.
A better brief: agree who should receive each message, what makes it useful and what causes someone to exit the journey. Test meaningful changes with commercial outcomes and customer-experience measures in view.
The commercial model can reinforce the wrong decisions
An agency paid on attributed revenue has a reason to improve that number. A fixed production package creates a reason to complete the contracted output. Neither arrangement guarantees a poor result.
The important work is aligning the incentives with the decisions you want.
If a fee depends on performance, agree the baseline, attribution settings, excluded revenue, treatment of discounts and refunds, and what happens when acquisition or reporting changes. A performance fee is a commercial arrangement; it is not evidence of incremental impact.
For any pricing model, ask which decisions the team would make differently if it were accountable for contribution from comparable customer groups.
Give your agency one focused customer-value problem
Choose a commercially meaningful customer group and a specific transition. Avoid rebuilding every journey before learning whether the first hypothesis helps.
For example: eligible first-time buyers of a particular product category who need support reaching a second purchase.
Agree:
The baseline: customer count, full observation window, qualifying orders and contribution definition.
The intervention: a specific change in education, product recommendation, timing or offer eligibility.
The comparison: a randomised control where practical, or a clearly qualified observational comparison.
The outcome: contribution per eligible customer, supported by second-purchase rate, refunds, contact costs and delivery health.
The decision date: long enough to observe the buying behaviour and relevant returns.
Check implementation and early indicators while the test runs. Read the mature outcome at the agreed point. Where the evidence remains uncertain, say so and decide what information is worth collecting next.
This makes the agency’s contribution easier to discuss: the customer problem, the proposed change, the work completed and the strength of the result.
Questions to put to your retention agency
Can LTV fall while repeat purchase rate rises?
Yes. The additional orders may carry lower margins, higher returns or higher service and intervention costs. Review contribution per original customer over the same horizon, as well as the share who return.
Is a lower email revenue share a sign of failure?
It needs context. Changes in acquisition, other channels and attribution can alter the share. Review total business performance and customer outcomes before drawing a conclusion about the email programme.
Should every retention test run for 90 days?
Choose the window around the product, buying cycle and decision. Ninety days can be a useful early operating horizon, but some outcomes need longer. State the window in advance and compare customers at the same age.
Make customer value the brief
A stronger retention programme gives customers useful reasons to return and measures the economics of that return.
If you are reviewing your current relationship, use 7 Signs You’ve Outgrown Your Email Agency alongside this article.

.png)

Comments