How to analyse your traffic source data to cut losers fast
How to analyse your traffic source data to cut losers fast
Most affiliate campaigns don't die from bad offers — they die from slow decisions. You buy traffic, watch clicks pile up, feel vaguely optimistic because the CPC is cheap, and three days later you're down a few hundred dollars on a placement that never converted once. To analyse traffic source affiliate data properly you need two things: a conversion signal that isn't faked by the network, and a cut-off rule you apply without emotion. This guide gives you both, plus the exact breakdowns to run when a campaign is bleeding.
Everything below assumes you're reading conversions from a tracker via postback rather than trusting the traffic source's own dashboard. If you're still comparing network-reported conversions to affiliate-network-reported conversions, you're not analysing data — you're reconciling two versions of a story that both parties have an incentive to flatter.
Why traffic source dashboards alone will make you keep losers alive
The traffic source dashboard tells you what it sold you: impressions, clicks, spend, and its own conversion count. Three problems with using that as your decision layer:
- Its conversion count is usually a black box. Some networks fire a pixel on the landing page and call that a conversion. That's a page load, not a sale.
- It has no idea what happened after the click left. The affiliate network knows about the sale. The traffic source doesn't, unless you feed it back.
- It's optimised to show you volume, not efficiency. A placement with 4,000 clicks and zero conversions looks busy in the dashboard. Busy is not the same as good.
The fix is a server-to-server postback: the affiliate network fires a conversion to your tracker, your tracker attributes it to the click (and the sub-ID that produced it), and only then do you have a number worth acting on. In Adtraxo this is a Pro feature — postback conversion tracking and sub-IDs sub1–sub5 are not available on the Free plan, which is worth knowing before you build a workflow around them.
Set your cut-off rule before you launch, not after
This is the single highest-leverage habit in performance marketing. Decide the kill threshold while you're calm, then obey it while you're emotionally invested.
A workable rule has three parts:
- A test budget per placement or campaign. The amount you're willing to lose to learn whether this segment converts. Express it in money, not clicks, because CPC varies by geo and device.
- A minimum click floor. Don't kill on 30 clicks with zero conversions — that's noise. Pick a floor where a single conversion would still be plausible at your expected conversion rate.
- A decision point. At the budget ceiling, the segment is either profitable, close enough to optimise, or dead. No extensions "just to see".
Here's the mechanism, not a made-up benchmark: if you expect roughly one conversion per X clicks, then X clicks with zero conversions is weak evidence and 3X clicks with zero conversions is strong evidence. Set your floor from your own historical conversion rate on similar offers, not from a blog post. If you have no history, be more generous with the floor and stricter with the budget.
Write the rule down per campaign. Something like: "Kill any placement at $15 spend or 600 clicks with zero postback conversions, whichever comes first." Then the only analysis you do mid-flight is checking whether the rule has fired.
The four numbers that decide whether a segment lives
When you open a campaign's report, most columns are decoration. These four are the decision:
- Cost — what the traffic source actually charged you.
- Postback conversions — verified by the affiliate network, attributed by your tracker.
- Revenue — payout × conversions, using the real payout per geo and per offer.
- ROI or profit — revenue minus cost. Profit in currency is easier to act on than a percentage, because it tells you how much budget is actually at stake.
Clicks and CTR are diagnostic, not decisive. They explain why a segment is failing (bad creative, bad placement, bot traffic) but they should never be the reason you keep one alive. A 6% CTR on a placement with no conversions is a creative that gets clicked by people who were never going to buy.
Break down by sub-ID, not by campaign
Campaign-level ROI hides everything useful. A campaign at −20% might contain one placement at +80% and fifteen at −60%. If you kill the campaign you kill your winner. If you leave it alone you fund the losers.
This is what sub-IDs are for. In Adtraxo, sub1–sub5 pass through to the click record, so you can tag the traffic source's placement ID, zone ID, creative ID, and anything else the network exposes. Then you break the campaign report down by sub-ID and sort by profit.
A practical mapping for most sources:
- sub1 — placement or zone ID (the traffic source's own identifier)
- sub2 — creative or ad ID
- sub3 — landing page variant (pair this with split testing)
- sub4 — geo or device bucket, if the network doesn't already pass it
- sub5 — anything campaign-specific, like a day-part or a bid strategy label
Keep the mapping identical across campaigns. The moment sub3 means "lander" on one campaign and "device" on another, your cross-campaign reports become unreadable. There's a fuller walkthrough in How to use sub-IDs to find your best-performing ad placements.
Run the kill-list review in this order
When you sit down to analyse traffic source data, work top-down. Each step removes noise before you look at the next level.
- Campaign level. Is the campaign as a whole above or below your cut-off? If it's deeply negative and every sub-segment is negative, kill the campaign and move on. Don't hunt for a winner that isn't there.
- Placement level (sub1). Sort by spend descending. The top spenders are where your money is going — check those first. Kill any that have passed your rule.
- Geo and device. Look for a geo that's eating budget with zero conversions, or a device split where one side never converts. This is often a targeting error, not a traffic quality problem — a mobile-only offer bought on desktop inventory will look like a loser and isn't.
- Creative level (sub2). Only after placement and geo are clean. A creative that fails everywhere is dead; a creative that fails on one placement but works elsewhere is a placement problem.
- Time of day. Only worth checking once you have enough conversions to see a pattern. Day-parting on 3 conversions is astrology.
Adtraxo's per-campaign analytics include geo and device breakdowns alongside the sub-ID report, which means steps 2–3 happen in one screen rather than an export and a pivot table. Country analytics is a Pro feature.
Separate "no conversions" from "no real clicks"
Two segments can both show zero conversions and need opposite responses.
Zero conversions with normal engagement — real users, wrong audience. The offer doesn't match the placement. Kill the placement or re-target.
Zero conversions with suspicious engagement — this is where fraud detection earns its keep. The classic signatures: an IP range hitting the same link repeatedly (IP velocity), clicks from datacenter IPs that shouldn't be seeing consumer inventory, bot user agents, an invalid or missing referer, and uniqueness conflicts where the same user appears to convert or click in impossible ways. Adtraxo's fraud detection flags these patterns and lets you build rules around them — also Pro-only.
If you're buying pop or push traffic, this distinction matters enormously, because the cheap end of that inventory is where junk clicks concentrate. Push traffic vs pop traffic — which converts better for affiliates? covers the quality differences in detail.
Use a one-line arrival snippet to catch the clicks that never loaded
Here's a failure mode that silently corrupts your analysis: a click is recorded by the tracker, but the landing page never renders. Slow host, geo-blocked CDN, mobile redirect loop. The tracker counts a click, the affiliate network counts nothing, and you conclude the placement is bad when actually your page was down for that segment.
Any Adtraxo plan — Free or Pro — lets you add a one-line arrival snippet to your landing page. It records which clicks actually loaded the page in a real browser. Compare recorded clicks to arrivals: a segment with a large gap between the two is a technical problem, not a media problem. Fix the page before you cut the placement, or you'll cut a winner for the wrong reason.
Compare traffic sources on the same scorecard
Analysing one source in isolation tells you if it's working. Analysing sources against each other tells you where the next dollar goes. Build one scorecard and apply it to every source — PropellerAds, EvaDav, RichAds, Taboola, MGID, solo ads.
Columns that matter:
- Spend and revenue in the same currency
- Postback conversions
- Profit and ROI
- Cost per conversion (spend ÷ conversions) — the number you compare against payout
- Conversion rate on arrivals, not on clicks
- Share of spend that went to placements later killed (a quality signal for the source)
That last one is underrated. A source where 70% of your initial spend went to placements you eventually killed is telling you something about its targeting controls or its inventory quality — even if the surviving 30% is profitable. It means your testing budget on that source is structurally higher.
For source-specific setup, the mechanics differ enough to warrant their own guides: How to track PropellerAds campaigns with a postback URL, EvaDav tracking setup — the complete guide, and RichAds vs PropellerAds — which is better for affiliate campaigns?. Solo ads need a different lens entirely because you're judging a seller, not a placement — see Solo ads tracking: how to measure seller quality and ROI.
When a campaign is borderline, test the variable instead of guessing
A campaign at −10% with a few conversions isn't a clear kill. It's a question: is the lander the problem, the offer, or the traffic? Guessing costs you a week. Testing answers it in a day.
Split the landing page first — it's the cheapest variable to change and usually the biggest lever. Run two variants against the same traffic and the same offer, and read the result on postback conversions, not clicks. Adtraxo's A/B split testing (Pro) handles the split and attributes conversions to the correct variant via sub3, so the comparison is clean. The method is in How to run A/B split tests on affiliate landing pages.
If the lander isn't the issue, test the offer against the same traffic before you blame the source. And if you're running a multi-step funnel, track each step — a drop-off at step two looks identical to bad traffic in a single-conversion report. How to set up a lander funnel and track every step shows the structure.
What to do with the survivors
Cutting losers is half the job. The other half is not strangling the winners. Once a placement or campaign clears your rule with positive profit across enough conversions to trust, the move is to increase budget in steps and watch whether cost per conversion holds. Scale too fast and you'll bid yourself into worse inventory, which shows up as rising CPC and falling conversion rate on the same placement.
Adtraxo's AI Campaign Optimizer can flag which segments are worth scaling and which are drifting, but the discipline is yours: raise budget in increments, re-check the scorecard after each increment, and stop when marginal profit stops growing. The full method is in How to scale a winning traffic source without killing your ROI, and the broader framework for managing multiple sources sits in How to track and optimise every traffic source for affiliate marketing.
The weekly routine that keeps losers from accumulating
- Pull the scorecard for every active campaign. One screen, same columns.
- Apply the cut-off rule mechanically. Anything past the threshold gets paused — no debate.
- Check the arrivals gap on anything with zero conversions before pausing it permanently.
- Break the top three spenders by sub1 and kill bad placements inside good campaigns.
- Promote one winner to a scale test with a defined increment.
- Log what you killed and why. Two weeks later, check whether any killed placement would have converted — if the pattern repeats, your floor was too low.
That's the whole system: a verified conversion signal, a pre-committed rule, sub-ID granularity, and a fraud check to separate bad traffic from bad targeting. Do it weekly and your losers never get more than seven days of budget.
Frequently asked questions
How many clicks should I wait before cutting a placement with no conversions?
Derive it from your own expected conversion rate rather than a fixed number. If you expect roughly one conversion per X clicks on similar offers, treat X clicks with zero conversions as weak evidence and around 3X as strong evidence. Pair the click floor with a money ceiling — whichever fires first triggers the cut. If you have no historical rate for the offer, set a more generous click floor and a stricter budget so the cost of being wrong stays small.
Should I trust the traffic source's conversion count or my tracker's?
Your tracker's, provided it's fed by a server-to-server postback from the affiliate network. The traffic source only knows what happened up to the click (or at best a landing-page pixel), while the affiliate network knows about the actual sale. Postback conversions attributed by your tracker are the only number that reflects real revenue. In Adtraxo, S2S postback conversions are a Pro feature.
What's the difference between a placement that doesn't convert and bot traffic?
Engagement quality. A non-converting placement with real users shows normal session behaviour but no sales — the audience doesn't match the offer. Bot traffic shows fraud signatures: repeated clicks from the same IP range, datacenter IPs on consumer inventory, bot user agents, invalid referers, and uniqueness conflicts. Adtraxo's fraud detection flags these patterns so you can tell a targeting problem from a traffic-quality problem before you decide what to cut.
Why break down by sub-ID instead of just looking at campaign ROI?
Because campaign ROI averages winners and losers together. A campaign at −20% can contain one profitable placement and fifteen unprofitable ones; killing the campaign loses the winner, and leaving it alone funds the losers. Sub-IDs sub1–sub5 let you tag placement, creative, lander and geo, then sort the report by profit so you cut at the right level. Sub-IDs are available on Adtraxo Pro.
How do I know if zero conversions are my landing page's fault?
Check the arrivals gap. Add the one-line arrival snippet to your landing page — available on any Adtraxo plan — and compare recorded clicks to actual page loads in a real browser. A large gap on a segment means the page didn't render for those users: slow host, geo-blocked asset, redirect loop. Fix that before you cut the placement, or you'll kill a segment for a technical reason and never know.
Ready to stop guessing which placements are draining
Track your affiliate campaigns for free
Clicks, conversions, fraud detection — all in one tracker.
Get started free →Not ready for an account? Get the Tracking Domination Playbook — free, just your email.