Conversion Protection

A bot that converts costs you more than a bot that clicks. The fake conversion trains your automated bidding to read the source that produced it as a success and buy more of it. This page is what ClickLens does about that: how it grades each reported conversion, what it can and cannot write back to each ad platform, and how it measures whether the protection helped.

If you have not reported a conversion yet, start with the conversion tracking guide. For the full mechanism behind each verdict, see the Conversion Protection feature page.

How a conversion is judged

Each conversion resolves to a graded verdict from three stored inputs:

1. The session's humanness score. The same 0–100 score the click was given, built from five weighted detection categories. See detection methodology.
2. The submit provenance. Seven meta-timing and structure signals captured when the conversion fired, never any field value. The conversion tracking guide lists all seven.
3. The coherence of the journey. Whether the path to the conversion is one a human produces, or a teleport straight to checkout and a looped identical path.

The verdict is one of three, and the bar rises with the cost of getting it wrong.

The verdict is a pure function of those stored signals, so every conversion on your dashboard lists the contradictions that produced its grade. The feature page sets out what counts as an independent contradiction class and why two classes are needed to retract.

What each platform allows

A verdict is only useful if the platform will act on it. ClickLens reports honestly on what each ad platform's API can and cannot do.

Google and Microsoft expose a documented advertiser-side retract and restate keyed by click-id. Meta's Conversions API and TikTok's Events API are forward-only ingestion with no per-click retraction, so a retraction there shows on your dashboard but cannot be synced back. That asymmetry is permanent until those platforms ship a retraction endpoint.

Measured rollout: what the dashboard shows today

Every gate ships in shadow mode first. A conversion is scored and its verdict is visible on your dashboard now, but write-back — the restate or retract on Google and Microsoft Ads — turns on only once its false-positive rate has been measured on live traffic.

So today the dashboard shows the adjustment ClickLens would make, before any gate changes platform data. A verdict we are not yet acting on is shown to you in full; we do not act on one we have not yet proven.

Holdout and CPA divergence

Conversion protection makes a claim about your spend, so it has to be measured against a control rather than asserted. One in ten conversions is held out of any adjustment as that control, untouched. Comparing the protected cohort against the 10% holdout is how the effect is measured on your own account.

The CPA divergence readout puts a number on it. A headline CPA counts every conversion equally. ClickLens divides it by the share of leads that turn out genuine, to get a cost per genuine lead:

cost per genuine lead = cost per conversion / genuine rate
genuine rate          = closed-won / leads reconciled by your CRM

The cost basis is the per-conversion click cost. When your real cost per click is not available, ClickLens falls back to a $2.00 default and flags the figure as an estimate, so a cost-per-genuine-lead built on the fallback is never presented as measured. The difference between the protected arm and the holdout is reported with a 95% confidence interval. At typical small-account volume that interval is wide and the difference is often not statistically significant yet — the readout says so rather than overclaiming a win.

Junk leads and the CRM loop

The submit-time score predicts whether a human filled the form. It cannot know whether the lead was worth money: a spam lead and a real customer can fill a form identically. Your CRM has that answer once the lead is worked.

When your CRM marks a lead, post the outcome to the ClickLens webhook and it reconciles against the conversion it already scored. A disqualified lead becomes a retraction candidate, so the bidder stops chasing junk. A closed-won lead cancels any pending suppression, so a real lead is never removed. That is ground truth the score could only predict.

FAQ

Does ClickLens change my Google conversion data right now?

No. Write-back to Google and Microsoft Ads is in measured rollout. The verdict and the adjustment ClickLens would make are visible on your dashboard today; the adjustment changes platform data only after its false-positive rate has been measured on real traffic against the 10% holdout.

Why can ClickLens retract on Google but not on Meta?

Google and Microsoft document an advertiser-side adjustment keyed by click-id, so a conversion can be restated or retracted after the fact. Meta's Conversions API and TikTok's Events API only accept new events; there is no endpoint to retract a specific past conversion. On those two, a retraction is reported to you but cannot be synced.

Will protecting conversions cut my conversion count?

Down-weight keeps the conversion and lowers its reported value; only retract removes one, and retract needs two independent contradiction classes plus a confidence above 0.7. The 10% holdout and the CPA divergence readout exist so you can see the effect on your own account rather than take the trade on faith.