Lesson 6 of 8 · 9 min read · intermediate
How traffic quality is scored, and how revenue is clawed back
Search engines and feed providers grade every publisher's traffic. Learn what goes into that grade, how it changes what you are paid, and how clawbacks work.
A publisher in a search feed is being graded all the time. The grade is rarely shown in full, the formula is not published, and it decides three things: whether each click is paid at all, how much each paid click is worth, and whether the feed stays switched on. This lesson explains the logic behind that grading, without pretending to know the private formulas.
A restaurant buys vegetables from a wholesaler. The wholesaler buys from farms. If one farm's tomatoes keep arriving bruised, the restaurant first knocks money off the invoice, then pays that farm less per crate, and finally tells the wholesaler to stop sending that farm's produce. The wholesaler, who wants to keep the restaurant, does its own checks before the lorry is loaded. The restaurant is the search engine, the wholesaler is the feed provider, and you are the farm.
Three layers of judgement
- Filtering at the clickAutomated systems decide in real time whether a click is billable. Google describes its defences as automated detection, manual review and research. Clicks judged invalid here are simply not charged to the advertiser and not paid to the publisher.
- Pricing by qualityClicks that pass can still be discounted. With smart pricing, the search engine lowers what an advertiser pays for clicks from a source that is less likely to lead to a sale. The click is valid. It is just worth less.
- Review after the factLater analysis, complaints or audits can reclassify clicks that were already counted. Advertisers are credited, and the money is deducted from the publisher's earnings.
What the grade is built from
No search engine publishes its weights, and anyone who claims to know them exactly is guessing. What is public, from policy pages and from how the system behaves, is the kind of evidence used.
| Signal | Question it answers | Who sees it |
|---|---|---|
| Advertiser conversion | Do people from this source go on to buy, call or sign up? | Search engine, via advertisers' own conversion tracking |
| Click validity rate | What share of clicks was filtered as invalid? | Search engine, partly shared with provider |
| Behaviour on the page | Does the visit look like a person reading and choosing? | Search engine, provider and publisher |
| Query quality | Are searches natural, relevant and chosen by the user? | Search engine and provider |
| Policy record | Have the pages or upstream ads broken rules before? | Search engine and provider |
| Advertiser reaction | Are advertisers excluding this placement or complaining? | Search engine |
The first row matters most. Advertisers tell Google and Microsoft which clicks led to sales so that automated bidding works. That same data shows, in aggregate, which partner sources deliver customers. Google has said, for instance, that it applies smart pricing and its Smart Bidding systems to search partner network traffic. A publisher never sees an individual advertiser's sales, but the effect arrives in their RPC.
The provider's own score
A feed provider cannot afford to wait for the search engine's verdict, because the verdict lands on the provider's account. So most providers run their own traffic quality score for each publisher, and often for each Channel ID. The names and scales differ from provider to provider, and they are not comparable with one another. Typically the score combines what the search engine reports back with the provider's own IVT detection. A falling score tends to lead to a lower revenue share, a feed cap on daily volume, a request to remove a source, and finally an feed suspension.
From estimate to money in the bank
Through March the arbitrageur buys 100,000 visitors at $0.20 each. The ad platform bills its card every few days: $20,000 of ad spend is gone before any income arrives. (An illustrative month.)
- Money goes out first: Through March the arbitrageur buys 100,000 visitors at $0.20 each. The ad platform bills its card every few days: $20,000 of ad spend is gone before any income arrives. (An illustrative month.)
- The dashboard shows an estimate: Each day the feed reports estimated revenue. By month end it reads $28,000, a paper profit of $8,000. It is a running tally, like a restaurant bill before the manager checks it.
- The engine reviews the clicks: After the month closes, the search engine’s systems finish checking for invalid clicks: bots, accidental taps, repeated clicks, traffic that broke policy. The publisher does not get an itemised list of what was removed.
- Advertisers are credited: Clicks judged invalid are credited back to the advertisers who paid for them. Say that comes to 5% of this publisher’s clicks. Nobody in the chain earns anything on a refunded click.
- The clawback: 5% of $28,000 is $1,400, deducted from the publisher: a Clawback (revenue deduction). The $20,000 spent buying those visitors is not refunded by anyone. A heavy clawback can turn a profitable month into a loss after the fact.
- Revenue is finalised: $28,000 minus $1,400 gives finalised revenue of $26,600. Real profit is $6,600, not $8,000. Google’s own AdSense timeline posts finalised earnings around the 3rd of the following month; feed providers set their own dates.
- Payment arrives on Net terms: The provider pays on Net terms, here Net 30: about 30 days after month end. Money spent on 1 March comes back around 30 April. Some contracts are Net 45 or Net 60.
- Growth eats cash: By the time March is paid, April’s $20,000 has also been spent. This cash-flow float means a growing arbitrageur needs working capital of one to two months’ spend, and a late payout or large clawback can sink a business that looks profitable.
Estimated, finalised and clawed back
The number on today's dashboard is estimated revenue. It becomes finalised revenue only after the search engine has finished its checks, commonly after the month has closed. The difference is the clawback. Google's AdSense help says directly that invalid clicks can cause a gap between estimated and finalised earnings, and that it deducts invalid activity to refund advertisers.
Why this is dangerous is easiest to see with numbers. Take an illustrative month: an operator spends $85,000 on traffic and the dashboard shows $100,000 of estimated revenue, a $15,000 profit and a 15% margin on revenue.
| Deduction | Finalised revenue | Ad spend | Profit or loss |
|---|---|---|---|
| 0% | $100,000 | $85,000 | +$15,000 |
| 5% | $95,000 | $85,000 | +$10,000 |
| 12% | $88,000 | $85,000 | +$3,000 |
| 15% | $85,000 | $85,000 | $0 |
| 20% | $80,000 | $85,000 | -$5,000 |
The spend was paid weeks ago. It cannot be recovered from the traffic source on the grounds that the feed later disagreed about quality. This asymmetry, costs that are certain and revenue that is provisional, is the central financial risk of the business and is covered from the cash side in the economics track.
What a sensible operator does with all this
- Budget for deductions. Work out your margin on a conservative finalised figure, not on the dashboard.
- Ask the provider what feedback exists. Many will share a score or tier per channel if you ask. Review it on a fixed day every week.
- Watch RPC by source over time. A slow decline on one source while others hold steady is the visible shadow of a quality adjustment.
- Keep sources separate. A score applied to a blended channel punishes your good traffic along with the bad.
- Use an independent check. A third-party score gives you a view that does not depend on waiting for month end. ClearTrust's TQI Score™, for example, rates traffic quality using more than 150 filters for bots, fake clicks and fake impressions. Such a score is your early warning. It does not replace the search engine's decision.
Key takeaways
- Traffic is judged in three layers: filtering at the click, pricing by quality, and review after the fact.
- The strongest signal is whether advertisers get real customers from a source; publishers feel it through RPC.
- Feed providers keep their own quality scores and act on them with lower shares, caps and suspensions.
- Estimated revenue becomes finalised only after checks; the gap is the clawback, and it can erase a thin margin.
- Independent scoring is useful as early warning but never overrides the search engine's own verdict.
Questions people ask
What is a clawback in search arbitrage?
A clawback is revenue that appeared in a publisher's estimated earnings and was later removed because the search engine judged the clicks invalid or low quality. The advertiser is credited and the publisher's finalised revenue is lower. Because the publisher already paid for the traffic, a clawback larger than the profit margin turns a winning month into a loss.
How does Google measure traffic quality from search partners?
Google does not publish a formula. Publicly it describes automated filters, manual reviews and research for invalid traffic, and it uses advertisers' conversion data to price partner clicks through smart pricing and Smart Bidding. In practice, partner sources that send visitors who buy are paid more per click, and sources that do not are discounted or removed.
What is smart pricing?
Smart pricing is Google's practice of reducing what an advertiser is charged for a click when its data suggests that click is less likely to lead to a sale than a click on Google's own results. The publisher's share falls accordingly. The click is still valid. It is priced lower because the source has a weaker record of producing customers.
Why is my finalised revenue lower than my estimated revenue?
Estimated revenue is counted before all validity checks are complete. When the search engine finishes its review it removes clicks it considers invalid, credits the advertisers and reduces the publisher's earnings. Small differences are normal. A large or growing gap on one source is a sign that the source has a quality problem that needs investigating.