A practical framework for scoring publishers on traffic quality, content type, conversion trend, and incrementality — and using the tiers to drive commission, priority access, and recruitment decisions.
Quick Answer
How often should a publisher tiering system be re-scored?
Quarterly at minimum, monthly if publisher volume supports it. A tier assigned once at onboarding and never revisited isn't a scoring system — it's a static label that decays in accuracy every month it goes unchecked, especially in categories with strong seasonality.
# How to Build a Publisher Tiering System That Actually Predicts Performance
Most affiliate programs have a tiering system in name only. There's a spreadsheet somewhere labeled T1/T2/T3, publishers get sorted into it once — usually at onboarding, based on gut feel or the size of their following — and then nobody touches it again until someone in a quarterly review asks why the "top tier" list is full of publishers who haven't driven a sale in four months.
That's not a tiering system. That's a filing cabinet.
A tiering system that actually predicts performance is a live scoring model: it ingests real publisher behavior on a defined cadence, weights the dimensions that actually correlate with revenue, and feeds directly into commission structure, priority access, and where you spend recruitment time. Get it right and it becomes the operating layer for the entire program — it tells you who gets the exclusive early-access product drop, who gets a commission bump before renegotiation, and who gets quietly deprioritized without a single awkward conversation. Get it wrong and you end up over-paying dormant "T1" accounts while genuinely high-converting mid-size publishers churn because nobody noticed they'd earned a better deal.
This is a practical build guide: what to score, how to weight it, the mistakes that quietly wreck most tiering models, and how to actually operationalize the tiers once they exist — for commission tiers, partner prioritization, and recruitment targeting inside networks like Impact, Awin, CJ, Amazon Associates, and Levanta.
Focus: Why Most Tiering Systems Fail Before They Start
The most common failure mode in publisher tiering isn't a bad formula — it's the wrong input variable. Programs default to tiering by follower count, domain authority, or historical GMV alone, because those numbers are easy to pull and easy to defend in a meeting. The problem is none of the three reliably predicts future conversion.
A 400K-follower Instagram account and a 40K-subscriber niche newsletter can produce wildly different ROAS on the same product, and the follower count tells you nothing about which one it'll be. Domain authority tells you a site ranks well; it says nothing about whether the traffic it ranks for actually buys. And historical GMV rewards publishers for volume they already have — which means your "T1" list quietly becomes a list of publishers who were good six months ago, not publishers who are good now.
The fix is to stop treating tiering as a one-time classification exercise and start treating it as a live, multi-dimensional scoring model — one that re-scores publishers on a fixed cadence using data the network already gives you, weighted toward the dimensions that actually predict repeat performance rather than the ones that are just easiest to screenshot.
Findings: The Four Dimensions That Actually Belong in the Score
Building a tiering model comes down to picking dimensions, weighting them honestly, and refusing to let vanity metrics sneak back in through the side door. In practice, four dimensions carry almost all of the predictive weight.
1. Traffic Quality, Not Traffic Volume
Raw traffic or follower count is a poor proxy for buying intent. What actually matters is the composition of that traffic: is it organic search intent (someone actively looking to buy a product like yours), owned-audience intent (a subscriber who trusts this specific publisher's recommendations), or low-intent referral traffic (a coupon aggregator or a browser-extension-driven click that would have converted anyway)?
Score this using data you already have from the network: click-to-order rate, average order value relative to your category median, and — critically — new-customer rate. A publisher sending high volume but mostly repeat customers or last-click coupon-hunters is providing far less incremental value than the numbers suggest, even if the topline commission payout looks identical.
2. Content Type and Placement Context
Where and how a publisher features your product matters as much as who they are. A dedicated, evergreen "best air purifiers" review post with a comparison table and a clear affiliate disclosure behaves completely differently, long-term, than a single Instagram Story that disappears in 24 hours or a sitewide banner ad.
Score content type on a durability-and-intent axis: long-form editorial and comparison content (highest durability, high intent) sits above video/shoppable content (strong intent, shorter shelf life unless it's evergreen YouTube), which sits above social organic (variable, platform-dependent), which sits above paid social or display placements run through an affiliate wrapper (lowest signal — this is rented attention, not earned trust). Coupon and deal sites deserve their own separate scoring track entirely, because they optimize for last-click attribution rather than genuine influence, and mixing them into the same scale as editorial publishers distorts the whole model.
3. Conversion History (Trended, Not Static)
This is where most programs get lazy and just pull lifetime GMV. Don't. Score conversion history as a trend, not a snapshot: rolling 90-day conversion rate, rolling 90-day EPC (earnings per click), and the trajectory of both — is this publisher accelerating, flat, or declining over the last two to three scoring cycles?
A publisher with modest lifetime GMV but a conversion rate that's climbed for three consecutive quarters is a better T1 candidate than a publisher with huge lifetime GMV built entirely from one viral post eighteen months ago that's since gone cold. Trended data catches both directions — it promotes publishers on the way up and flags publishers on the way down before they've fully decayed, rather than after a quarter has already been wasted overpaying them.
4. Audience Overlap and Incrementality
The dimension almost every program skips, and the one that most separates a real tiering system from a vanity one: how much of this publisher's converted traffic is actually incremental, versus how much would have converted anyway through another channel (paid search retargeting, email, direct)?
This is hard to measure precisely without proper multi-touch attribution, but you can proxy it. Compare a publisher's new-vs-returning customer split against your blended program average. Check whether their conversions cluster suspiciously around your own promo/discount code windows — a signature of pure last-click capture rather than genuine influence. And where you run multiple publishers in overlapping niches (three "best home gadgets" bloggers all pushing the same product), watch for cannibalization: are they collectively driving more total conversions than any one of them alone, or just splitting the same finite intent pool three ways and reducing your blended margin for no incremental gain?
Evidence: A Practical Scoring Framework
Here is a workable structure — adapt the weights to your category and platform mix, but keep the shape.
| Dimension | Weight | What to measure | Data source |
|---|---|---|---|
| Traffic quality | 30% | New-customer rate, click-to-order rate, AOV vs. category median | Network reporting (Impact, Awin, CJ dashboards) |
| Content type/placement | 20% | Content durability score, disclosure compliance, placement context | Manual audit + URL/creative review |
| Conversion trend | 30% | Rolling 90-day conversion rate and EPC trajectory (3 cycles) | Network performance reports |
| Audience overlap/incrementality | 20% | New-vs-returning split vs. program average; promo-window clustering | Network reporting + your own promo calendar |
Run this quarterly at minimum, monthly if you have volume to support it. A publisher's score should be able to move both up and down every cycle — if your model only ever promotes and never demotes, it isn't scoring, it's grandfathering.
Set explicit tier thresholds and stick to them: for example, top-decile composite score into T1, next 25% into T2, everyone else into T3, with a standing rule that no publisher stays in T1 for more than two consecutive cycles without re-qualifying on current data, not historical reputation.
Delivery: Common Scoring Mistakes to Avoid
A few mistakes show up repeatedly across programs, regardless of vertical:
- ◆Scoring on follower count or reach alone. Reach is an input to opportunity, not a predictor of conversion. Keep it out of the composite score entirely, or weight it near zero.
- ◆Never re-scoring. A tier assigned at onboarding and never revisited is a guess, not a system. Build the re-scoring cadence into the process before you build the model itself.
- ◆Treating all commission-driven traffic as equal. A coupon-code publisher and an editorial reviewer both show up as "conversions" in a flat network report, but they represent completely different acquisition value. Segment them before scoring, not after.
- ◆Letting relationship history override the data. Publishers who've been in the program a long time, or who have a personal relationship with the affiliate manager, tend to get grandfathered into T1 regardless of current performance. This is the single fastest way to erode trust in the tiering system internally — once one manager overrides the score, everyone starts asking why they can't.
- ◆Scoring in a vacuum, ignoring seasonality. A home goods publisher's Q4 numbers will always dwarf their Q2 numbers. Normalize against category or seasonal baselines, or you'll systematically over-tier publishers who happen to specialize in Q4-heavy categories and under-tier everyone else.
- ◆Using one network's data to score cross-network publishers unevenly. If a publisher runs your program through both Impact and Awin, or is present in Amazon Associates and a direct network deal simultaneously, aggregate their performance before scoring — don't let them show up as two mediocre partial-view publishers when they're actually one strong one.
Operationalizing the Tiers
A tiering model that doesn't change how you operate the program is just an analytics exercise. Here's where the tiers should actually plug in:
Commission structure. T1 publishers get the strongest base commission and are the first candidates for negotiated flat-fee or hybrid deals. T2 stays on standard published rates. T3 either stays at entry-level rates or gets moved to a probationary structure with a defined review point. The tier should directly gate what your team is authorized to offer in a rate negotiation — without that link, the tiering exercise has no teeth.
Priority access. T1 gets first access to new product launches, exclusive discount codes, and early sample units — the things that make a publisher's content genuinely differentiated from competitors covering the same product. This is often more valuable to a serious publisher than a marginal commission bump, and it costs you far less.
Recruitment targeting. Once you know which dimensions actually predict a T1 outcome, use that profile to target recruitment. Instead of prospecting broadly by follower count or vertical, prospect for the traffic-quality and content-durability signals your model has already proven correlate with performance — this is what separates efficient publisher recruitment from mass outreach that mostly adds noise to your T3 tier.
Deprioritization, quietly. Publishers who consistently score T3 don't need to be cut off — but they shouldn't be consuming account management time, personalized creative assets, or promotional priority that a rising T2 publisher would use more productively. Let the tier dictate where your limited operational attention goes.
What the Networks Actually Charge (So Your Tiering Economics Are Grounded in Reality)
Tiering decisions — especially commission-rate decisions — need to account for what the network itself takes off the top, since that shapes your real margin per publisher relationship. Fee structures vary by network and are frequently negotiated per advertiser, but the public structures are informative:
- ◆Impact.com publishes tiered plans on its own pricing materials starting with a Starter plan (a low monthly minimum or a percentage-of-revenue fee, whichever is greater), scaling through mid-tier plans into custom Enterprise contracts for high-volume advertisers, with a network fee applied on top of commissions paid through the platform.
- ◆Awin publishes two advertiser plan tracks on its own pricing page — Awin Access (aimed at smaller businesses and agencies getting started) and Awin Accelerate (aimed at businesses scaling an existing program) — and its own help documentation confirms advertiser invoices include a tracking/network fee calculated against commissions, in addition to the underlying plan fee.
- ◆CJ Affiliate does not publish advertiser pricing publicly; onboarding is quote-based and typically includes a setup component plus an ongoing network fee, negotiated per advertiser.
- ◆Amazon Associates doesn't charge advertisers a network fee in the traditional sense — instead, it operates on a published category-based commission-rate structure that Amazon sets and updates via its Associates Program Operating Agreement (rates have been revised more than once in recent years, including a significant restructuring in 2026, so any specific percentage cited outside of Amazon's own current agreement should be treated as a snapshot, not a stable figure) — advertisers effectively "pay" through the fixed rate card rather than a separate platform fee.
- ◆Levanta publishes its own advertiser pricing tiers directly, structured as monthly plans for Amazon sellers (creators join and use the platform for free), with paid plans layering in additional service fees on top of the base commission the seller sets.
The practical implication for tiering: your commission-tier economics aren't just "what do I pay the publisher" — they're "what do I pay the publisher, plus what does the network take on that same transaction." A T1 commission bump on a network with a meaningful transaction fee costs more in real margin terms than the same nominal bump on Amazon Associates' fixed rate-card model. Build that into the commission-tier math before you finalize thresholds, not after.
Bringing It Together
A publisher tiering system is only as good as its inputs and its refresh cadence. Score traffic quality over raw reach, weight conversion trend over lifetime totals, account for content durability and audience overlap, and rebuild the tiers on a fixed schedule rather than letting them calcify. Then actually use the tiers — for commission negotiation, for priority access, for where recruitment time goes — or the whole exercise is just a label with no operational consequence.
Programs that treat tiering as a living scoring model, re-run quarterly against real network data rather than reputation, consistently get better at predicting which publishers to invest in before the rest of the market notices. That's the entire point of building the system in the first place.
Frequently Asked Questions
Should follower count or audience size factor into publisher tier scoring?
Only as context, not as a scored dimension. Follower count indicates potential reach but has no reliable correlation with conversion quality. Weighting it directly into the composite score tends to inflate large-audience publishers who convert poorly and undervalue smaller, high-intent publishers who convert well.
How should coupon and deal sites be scored relative to editorial publishers?
Separately. Coupon and deal sites optimize for last-click attribution and often capture credit for sales that would have happened anyway, while editorial and comparison content tends to drive genuinely incremental purchases. Scoring both on the same scale distorts the model — segment them into distinct evaluation tracks before applying weights.
What's the biggest mistake programs make when building a tier system?
Letting relationship history or tenure override current performance data. Once an account manager grandfathers one long-tenured publisher into T1 despite declining numbers, the entire tiering system loses credibility internally, because everyone else starts asking why the rules don't apply evenly.
Do network fees actually affect which commission tier a publisher should get?
Yes. The network's own fee structure — a percentage-of-commission tracking fee, a platform subscription, or in Amazon's case a fixed category rate card that Amazon revises periodically — changes the real margin behind any commission increase. A T1 commission bump should be evaluated against what it actually nets after the network's cut, not just against the published commission rate.
Can a publisher tiering model work across multiple affiliate networks at once?
Yes, but only if you aggregate a publisher's performance across every network and program they're active in before scoring. Evaluating the same publisher separately per network — say, partial data from Impact and partial data from Awin — creates two incomplete pictures instead of one accurate one, and typically under-tiers otherwise strong partners.
