A programme that looked like growth, and was actually a discount desk.

The programme ran across two markets on a single network under a premium apparel brand with four decades of heritage behind it. At the point of audit it was turning over approximately $214,000 per month, with revenue and transaction volume both growing sharply month over month. Those headline figures suggested strong momentum. Structural weaknesses underneath them were actively undermining programme health and economics.

The most pressing technical issue was a defective server-to-server tracking implementation via the commerce platform plugin. The click reference was not being passed correctly, meaning the attribution chain between click and sale was incomplete — and unreliable attribution makes every downstream decision, from commission rates to partner disposition, a guess. Compounding it, the customer-acquisition parameter always returned RETURNING regardless of the customer’s actual status. The programme therefore could not distinguish new customer acquisition from repeat purchase at all. For a brand running a 15% welcome discount aimed squarely at first-time buyers, that is a critical blind spot: the single most important thing the discount was supposed to buy was the one thing nobody could measure.

Conversion sat within the sector benchmark range, but the figure was inflated by the partner mix rather than earned by it. One sub-network generated a large share of revenue at a conversion rate close to zero, pumping enormous click volume to produce very few conversions, from referrer domains that did not survive inspection. Another top partner showed device mismatches and was using SMS and new-customer prefixed voucher codes that were not registered in the offers list at all. Seven unverified codes were identified in live transaction data, indicating code leakage, unauthorised marketing activity, or fraudulent attribution — and the programme was paying commission on every one of them.

The Awin Index had declined to well under the 70% healthy threshold. The approval rate meant that roughly 43% of tracked transactions were being declined — catastrophically below the 90% benchmark. For cashback and loyalty partners, who front-fund customer rewards out of their own pocket before they are paid, that means losing money on nearly half the sales they generate. This was the single most destructive metric in the programme, and it is the kind of number that quietly ends partnerships without anyone sending an email about it.

Operationally, several hundred commissions awaited validation, representing close to a full month’s revenue held in limbo. The auto-validation period had been changed from thirty days to one hour, with no documented justification — raising the serious question of whether fraudulent transactions were being auto-approved before any human review could occur. Payment status showed the credit limit exceeded, with average payment time running at nearly double the 30-day benchmark.

The commission architecture spanned seven groups and the programme carried seven active offers, with terms revised three times since launch. Against that: the profile contact section was empty, zero documents had been uploaded, the product feed contained zero products, and Conversion Protection had no rules configured at all — leaving the programme entirely unprotected against attribution fraud. The partner mix was heavily skewed towards sub-networks, with the top eight publishers exceeding 85% of revenue and five of those eight showing active fraud or compliance flags.

The good news, and the basis of the recovery plan: the brand had genuine equity in premium apparel, with four decades of heritage, real design credentials and a target demographic that aligns well with high-value editorial and creator partnerships. The commission architecture already differentiated by publisher type, which is more sophisticated than most programmes at this maturity. Seven triggered communications were live, providing basic lifecycle engagement. Terms had been revised three times in ten weeks, showing active governance attention. With the tracking defects resolved, the unregistered codes investigated, the partner mix cleaned and rebalanced, and the approval rate addressed, this programme had a clear route to becoming a strong performer — which is what subsequently happened.

2A. Metric scorecard

Figures shown as a rating rather than a number are commercially private to the client. Benchmarks are AME Reference Library values for the fashion and apparel sub-sector and are not client data.

2B. Area scorecard

Deliberate commission differentiation from day one. Unlike the majority of programmes at this maturity, the brand had established seven distinct commission groups, differentiating by publisher type from coupon sites at the bottom through content creators and sub-networks in the middle to creators and VIP publishers at the top, with performance bonus tiers above that. This architecture demonstrates commercial intent and positions the programme to attract higher-value partner types through visible incentive differentiation. Commercially it means the programme can recruit against competitors by offering meaningful rate uplift to strategic partner types. To protect the advantage, the bonus tiers need margin justification on paper, and the 3% coupon default needs raising far enough not to deter legitimate voucher publishers who do contribute incremental traffic.

Comprehensive triggered communication coverage. Seven automated lifecycle communications were active: welcome, application accepted, first click, first sale, declined transactions, re-activation and dormancy. That covers the critical publisher journey touchpoints from onboarding through activation to re-engagement without manual overhead. For a programme at this maturity the level of automation is above average and confirms the initial setup was thorough. The build should now extend to manual newsletters, seasonal briefings and segmented outreach.

Active programme terms with a revision cadence. Terms had been revised three times since launch, covering general policy, PPC, transactions, branding, commission, notice periods, publishers and de-duplication across eight tabs. The PPC tab was fully populated with twelve or more policy definitions. The general tab covered trademark bidding prohibitions, direct linking permissions, approved code usage, email marketing rules and incentivised traffic policy. Three revisions in ten weeks indicates genuine attention to governance — which is rarer than it should be, and worth saying plainly. The cadence should continue quarterly and extend explicitly to cover the unregistered codes found in transaction data.

Strong brand narrative in the programme profile. The profile carried a comprehensive brand description communicating founding story, design heritage, product range, target demographic, international shipping reach, commission structure and the active volume bonus. That gives publishers the context needed to craft authentic promotional content, which matters disproportionately in premium apparel where brand storytelling directly drives conversion. It should be supplemented with a downloadable welcome pack, an imagery guide and a bestseller list so publishers have something actionable beyond profile text.

Rapid revenue growth trajectory. Revenue in the audit month had roughly doubled against the prior month, with transactions growing faster still. That growth required scrutiny given the partner-mix quality concerns — growth bought through low-incrementality partners is not the same asset as growth earned through demand creation — but it did confirm real underlying demand and genuine product-market fit translating into affiliate-driven sales. The right response was to audit the quality of the growth before scaling it, which is precisely what the task list does.

5.1 Overview

The programme had accumulated several hundred publishers in a matter of months, which points to some combination of aggressive inbound growth, auto-approval, and recruitment through sub-networks that bring large publisher pools with them. The quality of that base was the concern rather than its size: revenue was concentrated among a small number of publishers, several of which exhibited clear fraud indicators.

In the audit month the programme recorded strong click, transaction and revenue growth. Conversion and earnings per click were both declining even as revenue rose — the signature of volume being bought rather than demand being created. Combined with the approval rate, these metrics describe a programme generating significant gross volume with substantial quality problems underneath it.

5.2 Type distribution

Actual mix against the sector optimum

The gap is severe and it is systematic. The programme’s partner mix was close to the exact inverse of the sector optimum, with sub-networks and traffic-arbitrage publishers dominating where editorial, content and creator partners should lead. This is the finding the whole rebalance turned on: it is not that the programme had bad partners, it is that it had almost none of the partner types that generate demand rather than intercept it.

5.3 Concentration heat map

5.4 Device performance

Mobile carried 85% of traffic and converted at roughly a quarter of the desktop rate — a 3.6× conversion gap. For a premium apparel brand where the overwhelming majority of traffic arrives on mobile, that gap represents enormous revenue leakage. If mobile converted at even half the desktop rate, programme revenue would rise by roughly 80% with no additional traffic and no new partners at all.

The cause is likely to be some combination of checkout friction on mobile, page load speed, payment method availability, and the nature of the traffic itself — sub-network arbitrage typically delivers low-intent mobile clicks, so part of this gap is a partner-mix symptom rather than a site problem. Separating the two requires analytics funnel data, and that separation should happen before any spend is committed to fixing either.

5.5 Pending approvals

Three publisher applications were sitting unprocessed at the time of audit: a lifestyle content blog, an international cashback platform, and a fitting-tool comparison application. Two of the three were exactly the partner types the programme most needed — the content blog and the comparison tool — and both had been left waiting.

The recommendation was to approve the content and comparison applicants with priority, review the cashback applicant for geographic fit against the programme’s markets, and process all three within forty-eight hours. A programme that leaves its most-needed partner types waiting in a queue while paying nine tenths of its commission to traffic-arbitrage sub-networks has its priorities inverted at the level of daily admin, not just strategy.

Partner names are replaced with type-and-rank labels. Every analytical column is the client document’s own assessment, unchanged. Note that the matrix includes partner types the programme did not yet have — recruitment targets are dispositions too.

The programme showed no evidence of active publisher relationship management beyond its automated triggered communications. No manual newsletters were identified, no publisher segmentation was in use, no review cadence existed, and no exclusive relationships had been established beyond a single exclusive code holder.

Top-20 plan. The programme should identify and segment its top twenty publishers by verified value contribution — excluding those under fraud investigation — into three tiers. Tier one, the top five by verified revenue, receives monthly check-ins, early access to seasonal promotions and bespoke commission negotiation. Tier two, ranks six to fifteen, receives bi-monthly communication and promotional updates. Tier three, ranks sixteen to twenty, receives quarterly review and standard newsletter inclusion. Critically, this exercise has to wait until the compliance review is complete, because several of the current top publishers will not survive it.

Segmentation. Publisher tags should categorise partners on three axes: type (editorial, creator, content, cashback, coupon, sub-network, comparison), tier (VIP, standard, probation) and compliance status (verified, under review, flagged). Those tags should then drive communication targeting, commission group assignment and reporting segmentation. Without them the Communication Centre cannot address a segment, only an individual or everyone.

Early access. For seasonal peaks the programme should brief tier-one publishers four weeks ahead with exclusive previews of upcoming offers, early access to creative and priority placement opportunities. This is standard practice in apparel, where editorial lead times for gift guides and buying guides routinely run six to eight weeks — meaning a brand that briefs at four weeks has already missed the print and scheduling deadline for its most valuable placements.

Content freshness. Publisher-side content quality could not be assessed from within the network account; it requires external review of each publisher’s promotional material. This should form part of the compliance review, particularly for the partners whose promotional methods are currently unverified.

Competitor risk. In premium apparel the brand competes for publisher attention with a well-defined set of established competitors, most of whom run their own affiliate programmes with competitive rates and existing publisher relationships. Without active relationship management the programme risks losing quality publishers to competitors who offer better support, faster payment and reliable approval rates. At the audit’s network index and approval rate, the programme was measurably less attractive than any competitor running a healthy one — and publishers can see both numbers before they apply.

Discovery usage. No evidence of Partner Discovery usage was found anywhere in the account. The publisher base had accumulated through inbound applications and sub-network partnerships rather than targeted outreach — which is precisely why the mix looked the way it did. Given the severity of the imbalance, strategic recruitment was the primary lever available to rebalance the programme. Rate changes alone would not do it: you cannot rebalance towards partner types that are not in the programme.

Invitation pipeline. No targeted invitations had been sent. The programme needed a recruitment pipeline built around three priority segments: national fashion and lifestyle editorial titles; category-specialist content creators and reviewers; and inclusive fashion creators aligned with the brand’s positioning and demographic.

Gap analysis

Recruitment order. The programme should prioritise in this sequence. First, fashion editorial publishers capable of producing seasonal content — buying guides, seasonal roundups, gifting features — because their lead times are longest and they anchor the rest of the mix. Second, creators with engaged followings in the brand’s demographic. Third, quality cashback platforms in both programme markets, for legitimate last-click volume. Fourth, comparison and CSS publishers, which cannot begin until the product feed exists.

That ordering matters. Recruiting cashback before editorial would simply replace one form of last-click dependence with another, and the programme would arrive at the same problem by a different route.

9.1 Current state

The programme operated a seven-group commission architecture that was more differentiated than most programmes at this maturity: a bottom tier for coupon and voucher sites, a mid tier shared between content creators and sub-networks, a higher tier for creators, a VIP tier for proven partners, and three separate bonus levels above that. A flat volume bonus sat alongside them.

The architecture demonstrated genuine commercial intent. The problems were in the calibration rather than the design. The default coupon rate sat far below the sector benchmark for a default commission — and while a low coupon rate may be intentional as a way of discouraging coupon-only publishers, this one was low enough to deter legitimate voucher partners who do deliver incremental traffic. The content rate aligned with the lower end of the sector benchmark. The creator rate was reasonable.

The three bonus levels carried no naming convention and no qualification criteria, which made it impossible for a publisher to understand how to qualify or what any of them represented. A bonus tier nobody can qualify for on purpose is not an incentive; it is a discretionary payment with a percentage attached.

The most consequential structural flaw was that content creators and sub-networks shared a single rate. A direct content partner who creates demand and an aggregating sub-network who intercepts it were being paid identically — which, given the mix, meant the programme was systematically overpaying the partner type it wanted less of.

9.2 Recommended architecture

9.3 Budget impact

The direct budget impact of the rate changes is modest: raising the coupon rate two points, creating a standard tier at the sector floor, and separating sub-networks from content. The coupon increase costs a small amount per sale but makes the programme viable for legitimate voucher publishers. The new standard tier provides a catch-all above the floor rather than below it. The editorial tier creates a real recruitment incentive for the most strategically important partner type in the programme.

The larger budget effect comes from the compliance review rather than the rate card. If the flagged sub-networks are removed, commission spend falls substantially — those publishers accounted for a large share of current revenue, and their removal frees budget for investment in editorial, creator and content partnerships. The net effect of cleaning the mix and adjusting rates together is likely to be budget-neutral or better, because fraudulent and low-incrementality commission spend is eliminated and replaced with lower-volume, higher-quality conversions.

This is the arithmetic the client was asked to accept up front: revenue falls first, then recovers on a different foundation. It did fall, for two months, by design.

The client document carries each task with fifteen columns. The fields are laid out here as a card because fifteen columns is unreadable on any screen. No field has been dropped in the transform.

Day 0–30: stabilise

The first thirty days repair the tracking chain, remove the immediate financial and fraud exposure, and make the programme safe to measure. Nothing downstream can be trusted until attribution works.

Day 31–60: tune

Day 61–90: scale

A note on sequencing. The rebalance was expected to reduce revenue before it increased it, and the client was told so before the work began. Removing low-incrementality partners removes their revenue immediately, while the editorial and creator partners replacing them take weeks to produce. Two months of planned decline is the honest cost of the change, and a plan that hides that cost is not a plan.

12.1 Standing cadence

12.2 Retail calendar moments

Fashion and apparel carries one of the densest retail calendars of any sector, and editorial lead times are the binding constraint. A brand briefing at four weeks has already missed the scheduling deadline for its most valuable placements, which is why the editorial briefing dates below sit six to eight weeks ahead.

Not yet, and scaling it in its audit state would have made the underlying problem worse. The programme was growing, and that growth was real in the sense that money was arriving. But two thirds of it came from partners bidding on the last click for demand the brand had already created. Scaling that structure would have bought more of the same: higher gross revenue, lower margin, and a channel increasingly indistinguishable from a discount desk.

What held it back. Three things, in order. First, the attribution chain was broken — the click reference was missing and the customer-acquisition parameter never reported a new customer, so the programme could not measure what it was buying. Second, the partner mix was the inverse of the sector optimum, with sub-networks and traffic arbitrage occupying the position editorial and creator partners should hold. Third, the trust metrics — approval rate, payment time, credit status — were bad enough to repel exactly the quality partners the rebalance depended on recruiting. That third point is what made the sequencing non-negotiable: you cannot recruit editorial partners into a programme that declines half their sales and pays late on the rest.

The first five tasks, and why that order. One: fix the click reference, because until attribution works every other decision is a guess. Two: fix the customer-acquisition parameter, because the entire acquisition case for the channel is unmeasurable without it. Three: trace the seven unregistered codes, because the programme was paying commission on transactions it could not account for, concentrated on its largest partner. Four: correct the one-hour auto-validation window, because it was auto-approving those transactions before anyone could review them. Five: resolve the credit status and diagnose the decline rate together, because they are the two numbers a prospective quality partner sees before deciding whether to apply.

What the brand should not do yet. No recruitment spend, no seasonal campaign investment and no rate increases until attribution is repaired and the compliance review is complete. Recruiting quality partners into a programme with a 43% decline rate and late payments burns the relationship on first contact, and those partners do not come back for a second attempt. Equally, do not remove the flagged sub-networks before the replacement pipeline is contracted — sequence the removal against the recruitment, or the revenue gap is larger and longer than it needs to be.

What to review in 30 days. Is the click reference present in every server-to-server call? Are both new and returning values appearing in transaction data? Have all seven codes been traced and resolved? Is the validation window corrected? Is the payment warning cleared? What has the decline-rate analysis found, and how much of it is systemic?

What requires external evidence before a final conclusion. Analytics funnel data is needed to diagnose the mobile conversion gap and to separate genuine site friction from low-intent arbitrage traffic. Traffic source documentation is needed from four sub-networks before their disposition can be finalised. The brand’s own marketing team must confirm which, if any, of the seven codes were issued through a non-affiliate channel. And the brand’s genuine return and cancellation rate by channel is needed to establish how much of the decline rate is real.