Buying churn at full price, and calling it growth.

A note on this audit. Unlike the other eight audits published here, this page is not an anonymised client document. No source audit exists for this niche in publishable form. It is built to the same fifteen-section AME standard and the same depth as the eight real audits, using the corresponding case study as its factual source. Every measured figure on this page comes from that published case study. Where a measure would ordinarily come from the client account, this page reports a rating or states that the figure is not disclosed rather than inventing one.

The programme was an established US pet food subscription channel running at a mid-five-figure monthly revenue level at the point of audit, in what is seasonally the category’s strongest month. Subscriber acquisition volume was healthy and growing. On every report the brand routinely read, the programme looked like it was working.

It was not. Seventy-one per cent of subscriber acquisition came through coupon and cashback partners promoting a heavily discounted first box. Those subscribers cancelled at more than twice the rate of every other channel. Across affiliate-sourced subscribers as a whole, only 41% reached the third billing cycle — and the programme paid full commission on all of them regardless of whether they stayed one cycle or twenty.

This is the defining structural failure in subscription affiliate programmes, and it is almost always invisible from the revenue report. First-order commission pays for an acquisition event, not a customer. In a business where the customer’s value accrues over months, that means the programme pays its largest single amount at the exact moment it knows least about whether the customer is worth acquiring — and it pays the same amount whether they stay for one box or three years.

The consequence was a partner mix the model had selected for. Voucher and deal partners were the largest contributor to acquisition and produced subscribers with the worst retention in the programme, at 29% reaching the third cycle. Cashback and loyalty partners followed at 44%. Content and editorial partners — who produced subscribers retaining at 58%, materially better than any other type in the programme — contributed just 9% of revenue, because a model paying flat first-order commission made careful, slow, trust-building content economically uncompetitive against a discount code.

Veterinary and professional content, trainers, and genuine owner accounts were absent from the programme entirely. Not because they had declined, but because nobody had approached them, and because the commission model would not have supported them if anyone had. A partner who spends three weeks producing a considered feeding-transition guide cannot compete for placement against a partner who lists a discount code in ninety seconds, when both are paid identically.

Measurement was the enabling failure. The programme had no cohort tracking by partner, so nobody could see which partners sent subscribers who stayed. It had no subscription attribution across billing cycles, so recurring revenue was invisible to the affiliate channel entirely. And it had no cancellation reason capture, so the brand could not distinguish a subscriber who left because the product was wrong from one who was only ever there for the introductory discount. Without those three instruments the retention problem was not merely unaddressed — it was unobservable, which is why a competently managed programme had been running this way without anyone noticing.

The commercial case for change. Rebuilding the commission model around retention means paying less at the point of acquisition and materially more when a subscriber proves durable. It reduces cost on churning subscribers and increases it on retained ones, which is the correct direction in a subscription business. It also, unavoidably, reduces revenue in the short term: removing discount-led volume removes that volume immediately, while the partner types replacing it take months to produce. The programme should expect roughly two flat-to-declining months, and that expectation should be agreed before the first change is made rather than explained afterwards.

The good news is that everything required was already present. The audience existed, the product retained well when sold to the right person, and the partner types capable of producing durable subscribers were available and uncontested — because every competitor in the category was making the same commission mistake.

2A. Metric scorecard

Every measured figure here comes from the published case study for this programme. Where a measure would ordinarily come from the client account and is not published, this page states that it is not disclosed rather than substituting an invented value. Benchmarks are AME Reference Library values for subscription retail and are not client data.

2B. Area scorecard

A subscription funnel that converts. The programme was acquiring subscribers at genuine pace. The conversion mechanics worked, the offer was compelling, and the path from click to first box was efficient. That matters because it removes an entire class of problem from the diagnosis: this was never a conversion failure or a demand failure. The machine worked. It was pointed at the wrong outcome.

A product that retains when it reaches the right customer. The retention split by partner type is the most valuable diagnostic in the audit. Subscribers arriving through content and editorial partners reached the third billing cycle at 58%, against 29% for those arriving through voucher and deal partners. Same product, same funnel, same price — twice the retention. The product is not the problem. The customer selection is. That single comparison is the entire commercial case for the rebuild, and it existed in the data before anyone went looking for it.

Genuine seasonal strength. The programme captured the category’s December gifting peak effectively, and the audit month was one of its strongest. That competence is real and worth keeping. It is also, uncomfortably, part of why the problem persisted: a strong seasonal month flatters a programme with poor retention underneath it, because the acquisition number looks excellent precisely when the churn from it has not yet arrived.

An uncontested opportunity in the partner types that matter. Veterinary content, trainers, behaviourists and genuine owner accounts produce the most durable subscribers in this category by a distance. They were absent from this programme — and they were absent from competitor programmes too, for the same reason: nearly every subscription programme in the category runs flat first-order commission, which prices those partners out. That means the opportunity was not merely available; it was uncontested. A brand that fixes the commission model first can recruit a partner type its competitors structurally cannot afford.

An operationally competent team. The programme was being run carefully. Offers were managed, partners were communicated with, volume was tracked. The failure was one of target rather than execution, which is a much easier problem to fix: the same operational discipline redirected at retention produces results quickly, because the habits already exist.

5.1 Overview

The programme’s partner base was heavily concentrated in discount-led acquisition. Seventy-one per cent of subscriber acquisition came through coupon and cashback partners. Content and editorial contributed 9% of revenue. Veterinary, professional, trainer and owner-account partners were absent entirely.

The mix was not a recruitment accident. It was the predictable output of a commission model that paid every partner type identically for a first order, in a business where the value of that order varies by a factor of more than two depending on who sent it.

5.2 Type distribution and retention

Read the second and third columns together and the whole audit is visible in one table. The programme’s revenue share was inversely related to its retention performance. The partner type with the worst retention had the largest share. The type with the best retention had the smallest. Same product, same funnel, same price — the difference is entirely in who was sent.

The three absent partner types matter disproportionately in this category. Pet nutrition is a trust purchase made on behalf of an animal that cannot state a preference, and owners change food reluctantly and after consultation. That makes veterinary professionals, trainers and genuine owner accounts unusually influential — and unusually durable in what they produce. They were absent here for a specific and fixable reason: a partner who spends three weeks producing a considered feeding-transition guide cannot compete for the same commission as a partner who lists a discount code in ninety seconds.

5.3 Concentration

Conventional concentration analysis asks whether too much revenue depends on too few partners. In a subscription programme that question is secondary. The concentration that matters is by durability: what share of acquisition comes from partners whose subscribers stay. On that measure this programme was critically concentrated in its weakest cohort, and no report in the account would have shown it.

5.4 Device and journey

Device-level detail is not disclosed for this programme. The structurally relevant observation concerns the journey rather than the device: in this category the decision to change an animal’s food is typically made over days or weeks, often after consulting a professional or another owner, and completed later. That extended, advice-led journey is exactly the pattern last-click attribution handles worst — and it is why assist commission matters more here than raw device optimisation.

5.5 Approvals and partner quality

Approval detail is not disclosed. The relevant structural point is that a programme rewarding first-order volume will approve and retain discount-led partners efficiently, because they perform well against the metric being measured. The approval process was not failing; it was succeeding at the wrong objective. Changing the objective changes which partners are worth approving, which is why the commission rebuild has to precede the recruitment work rather than run alongside it.

Relationships existed with the largest volume partners, which under the old model were the discount-led ones. No relationships existed with the partner types that produce durable subscribers, and no framework existed to identify which partners those were.

The relationship problem created by the commission change. Rebuilding commission around retention means some partners earn materially less. Those partners will ask why, and they are entitled to an answer. Without partner-level cohort data, there is no answer — only an assertion that their subscribers churn, which no partner will accept and none should. This is why cohort tracking is a precondition of the model change rather than a follow-up to it: the new model cannot be defended to the partners earning less under it without the reporting that justifies it.

What partners need to see. Every partner should receive their own retention figure alongside their volume, on a regular cadence. That single addition changes partner behaviour more than any communication a programme can write, because it makes the metric the programme now pays on visible to the people being paid. A partner who can see that their cohort retains at 30% and that another tier exists above them has a route to earn more. A partner who cannot see it simply experiences a rate cut.

The conversations that follow. Some of these discussions are genuinely difficult, particularly with high-volume discount partners whose economics change substantially. They are worth having properly and with the data present. Some of those partners will improve their targeting once retention is what pays; others will leave. Both outcomes are acceptable, and the second is not a failure — a partner whose entire model depends on being overpaid for churn is a partner the programme cannot afford at any volume.

Building the new relationships. Veterinary professionals, trainers and owner accounts are recruited individually and slowly. They will ask about product sourcing, nutritional formulation and what claims they are permitted to make, and they will decline if those answers are not available. They typically work on commission-only terms, which suits a retention-weighted model well — but they need the partner pack, the claims guidance and the tier to exist before the first conversation, not after it.

Cadence to establish. Monthly retention reporting to every active partner. Individual contact with the professional and creator cohort while it is small. Quarterly reviews with the partners whose cohorts retain best, regardless of their volume — because under the new model those are the most valuable relationships in the programme even when they are not the largest.

Recruitment had followed the path of least resistance. Coupon and cashback partners are the easiest to recruit, the fastest to produce volume and the cheapest to serve, and under a flat first-order commission model they perform excellently against the metric being measured. The programme had accumulated them efficiently.

The partner types capable of producing durable subscribers — veterinary content, trainers, behaviourists, genuine owner accounts — had never been approached.

Why they were absent, and why that is an opportunity. Those partners were not unavailable and had not declined. They were priced out. A veterinary practice producing a considered piece on transitioning an animal to a new food, or a trainer whose recommendation carries real weight with their clients, cannot justify that effort against a commission rate identical to a discount code listing. And because nearly every subscription programme in this category runs the same flat first-order model, those partners are equally priced out everywhere. The opportunity is therefore uncontested — a brand that fixes its commission model first can recruit a partner type its competitors structurally cannot afford to pay properly.

The prerequisites, in order. First, the retention-weighted commission model, so a durable subscriber is worth more to the partner than a discount-chaser. Second, assist commission, so an upper-funnel professional partner is not outbid at the last click by a voucher code. Third, approved claims guidance, because a veterinary professional will not publish nutritional claims they cannot stand behind and a creator will not risk their audience’s trust on vague language. Fourth, the partner pack: ingredient and sourcing detail, nutritional formulation, feeding-transition guidance, and imagery. None of these are optional preliminaries; each one is a question the partner will ask in the first conversation.

Then recruitment, by type. Veterinary practices and veterinary content publishers, approached individually and with the professional materials in hand. Trainers and behaviourists, who reach owners at exactly the moment a feeding decision is being made. Genuine owner accounts and creators with category audiences, on commission-only terms, which is standard for this type and works well under a retention-weighted model. Category content and editorial publishers, to grow the type that already retains best. And community and forum operators, because that is where owners actually ask for recommendations and where trust transfers most directly.

What to stop recruiting. Further discount-led partners, until the mix has rebalanced. The programme never had a shortage of capacity to convert discount-motivated demand; it had a shortage of everything else.

9.1 Current state

The programme paid a flat commission entirely on the first order. There was no component tied to whether the subscriber stayed, no differentiation by partner type, and no cap on the entry discount that drove the majority of acquisition.

In a subscription business this arrangement inverts the programme’s incentives completely. It pays the largest single amount at the moment the programme knows least about the customer’s value. It pays identically for a subscriber who cancels after one box and one who stays for years. And by setting every partner type’s reward equal, it makes the effort required to produce a durable subscriber economically irrational: a considered veterinary feeding guide and a ninety-second discount listing earn the same, so the programme reliably accumulates more of the latter.

The retention data proves the cost. Subscribers arriving through voucher and deal partners reached the third billing cycle at 29%; those arriving through content and editorial partners reached it at 58%. The programme was paying the same rate for customers worth roughly half as much, and the majority of its acquisition sat in the cheaper cohort.

9.2 Recommended architecture

9.3 Budget impact

The headline effect is a reduction in first-order commission cost, offset by a new retention payment released at the third billing cycle. Because a minority of subscribers reached that cycle at audit, the immediate net effect is a reduction in total commission spend — and as retention improves, spend rises in direct proportion to value delivered, which is the correct relationship for a subscription business.

The important consequence is distributional rather than aggregate. Partners sending subscribers who stay earn more than they did before, on a lower first-order rate, because the retention payment is larger than the reduction. Partners sending churn earn considerably less. That is the whole point of the change, and it should be communicated in exactly those terms: this is not a rate cut, it is a redistribution towards partners producing customers who stay.

The short-term cost is real and should be stated before the change. Removing discount-led volume removes that volume immediately, while the partner types replacing it take months to produce their first subscribers. The programme should expect roughly two flat-to-declining months. A plan that does not say so in advance will be abandoned in month two, immediately before the new mix begins to compound.

The entry discount cap contributes separately: capping it improves margin on every acquisition and reduces the programme’s dependence on the discount as its primary promotional mechanism, which is a precondition for recruiting partners who promote on merit rather than price.

The full inventory by area

Because no client source document exists for this niche, the complete task inventory is set out below rather than summarised, so the total shown above can be checked against it.

Days 0–30: make it measurable, then rebuild the model

Measurement comes first and the sequence is not negotiable. The commission change cannot be defended to the partners who earn less under it without partner-level retention data, and it cannot be implemented at all without subscription attribution to trigger the retention payment.

Days 31–60: build the prerequisites and start recruiting

Days 61–90: scale the new mix

On the two flat months. Removing discount-led volume removes that volume immediately, while the partner types replacing it take months to produce their first subscribers. The programme should expect roughly two flat-to-declining months once the model changes. Two months of standing still is what it costs to change what a programme rewards, and it should be agreed in writing before the first partner is re-tiered — a plan that hides this cost gets abandoned in month two, immediately before the new mix begins to compound.

12.1 Standing cadence

The distinguishing feature of this calendar is what is measured monthly. A conventional programme reviews volume, revenue and mix. A subscription programme has to review retention by partner, lifetime value and cancellation reasons, because those are the metrics the commission model now pays on — and because volume, reviewed alone, is exactly what concealed the problem in the first place.

12.2 Category calendar moments

Two entries below carry warnings rather than plans. The category’s discount-heavy peak periods are precisely where a subscription programme can undo a retention rebuild in a single month, by re-acquiring the cohort it just spent two flat months removing.

No — and scaling it in its audit state would have made the underlying problem materially worse. The programme was acquiring subscribers at pace, and every additional unit of that pace was buying a customer who was statistically unlikely to reach the third billing cycle, at full commission. Growth on those terms compounds the cost rather than the value.

What held it back. One structural decision, and three measurement gaps that made it invisible. The decision was paying full commission on the first order in a subscription business, which sets every partner type’s reward equal regardless of the durability of the customer they send. The measurement gaps — no cohort tracking by partner, no subscription attribution across billing cycles, no cancellation reason capture — meant nobody could see the consequence. A partner sending subscribers who churned at 71% and one sending subscribers who retained at 58% appeared identical on every report the brand read.

The finding that reframes everything. Same product, same funnel, same price, and subscribers arriving through content partners retained at roughly twice the rate of those arriving through voucher partners. The product was not the problem and the funnel was not the problem. Customer selection was, and customer selection is determined by what the commission model pays for. The programme was not failing to manage retention; it was successfully buying churn, exactly as its incentives instructed.

The first five tasks, and why that order. One: implement cohort tracking by partner, because until the problem is observable no change to it can be justified or defended. Two: implement subscription attribution across billing cycles, because the retention payment has no event to trigger on without it. Three: rebuild commission as a retention-weighted split, because it is the change everything else exists to enable. Four: cap the entry discount, because the programme was paying twice for its weakest customer. Five: enable assist commission, because without it the improved rates never reach the upper-funnel partners the recruitment plan depends on.

What the brand should not do yet. No recruitment of professional, veterinary or creator partners until the retention model, assist commission, claims guidance and partner pack are all in place — those partners ask about all four in the first conversation, and a partner who declines once is significantly harder to approach again. No re-tiering of existing partners before the cohort data exists to justify it, because a rate reduction without evidence is simply a rate cut and will be received as one. And no deep-discount seasonal activity during the rebuild, which would re-acquire the exact cohort the programme just spent two months removing.

What to review in 30 days. Is partner-level cohort reporting in production and does it reconcile to brand subscriber data? Are billing-cycle events attributing to the acquiring partner? Is the split commission model designed, modelled against lifetime value, and is the cancellation rule written? Is the entry discount cap agreed? Have affected partners been briefed individually with their own data?

What requires external evidence before a final conclusion. Brand subscriber cohort data by acquiring partner, without which the retention analysis cannot be extended beyond the published headline figures. Lifetime value and margin data, to set the split between the first-box rate and the retention release defensibly. A qualified veterinary or nutrition contact to approve the claims language before professional recruitment begins. And the brand’s subscription unit economics, to set the entry discount cap at a level that protects margin without collapsing conversion.

One thing to agree before starting. This rebuild costs roughly two flat-to-declining months. Removing discount-led volume removes it immediately; the partner types replacing it take months to produce. That cost should be written down and signed off before the first partner is re-tiered. A programme that discovers it in month two abandons the change immediately before the new mix begins to compound — and having abandoned it once, is considerably harder to persuade a second time.