Every conversion rate has a denominator, and every denominator has holes. What makes this genuinely difficult rather than merely annoying is that the holes do not all push the same way. Two of them make your numbers look better than reality and two make them look worse, and the net effect is unknowable without measuring each separately.
The four holes
- Ad blockersflatters you
- Declined consentflatters you
- Cross-deviceunderstates paid
- Bot traffichurts you
Direction of distortion, not magnitude. Blocked and unconsented visitors are missing from both numerator and denominator, which raises apparent conversion rate. Bots inflate the denominator only.
Which direction each gap distorts your reporting. Ad blockers: flatters you. Declined consent: flatters you. Cross-device: understates paid. Bot traffic: hurts you.
Take them one at a time, because the mechanisms are different and so are the responses.
Ad blockers
Somewhere between 15 and 30 percent of web tracking is blocked, with technical and younger audiences at the top of that range. A blocked visitor is invisible in both the numerator and the denominator, so your conversion rate is calculated on a sample that excludes them entirely.
The commonly sold answer is first party proxying, and it is oversold. Blockers match on request paths, payload shapes and global variable names, not only on domains. Serving your tracker from your own domain helps and does not solve it, and the vendors who imply otherwise are describing a temporary state.
The alternative is obfuscation, which is an ethical choice rather than a technical one. Whatever position you take, the blind spot needs stating in the report. A silent undercount is the failure that destroys trust in every number you produce.
Declined consent
In the EU and increasingly elsewhere, a visitor who declines is not measurable in the ordinary way. The share varies enormously by geography and by how the banner is written, and any vendor quoting you a single global figure is guessing.
Note also that cookieless does not automatically mean consent free. The EDPB's guidance from October 2024 makes clear that consent obligations attach to storing or accessing information on a device by various means, not only to cookies specifically. Anyone selling you a cookieless tracker with an unqualified no consent needed claim is making a legal assertion they are not qualified to make.
Cross-device
A phone clicks the ad and a laptop buys two days later. Most setups record two unrelated visitors, credit the conversion to whatever the laptop session looked like, and quietly understate the paid channel that started it.
This one has a partial technical answer and a documented competitor failure worth knowing about. GA4's funnel explorations attribute a user to the first value recorded for a breakdown dimension, so a mobile to desktop journey counts as mobile at every step. Getting this right means resolving the visitor once at funnel entry and holding that resolution across every step, rather than re-deciding per step.
Bot traffic
The one that cuts the other way. Bots inflate visit counts with sessions that never convert, so they depress conversion rate while making traffic look healthy. Worse, they cost money twice over, in storage and in wrong answers.
- 01Filter bots before storage rather than in the report, so the raw numbers are already clean
- 02Count what you filtered and show the ratio, so a number smaller than the server logs has its explanation attached
- 03Never bill for filtered traffic, and say so
- 04Print the unattributable visitor count next to every cost figure rather than in a footnote
The second of those is the one almost nobody does. Filtering bots quietly produces a number that disagrees with the customer's server logs, and an unexplained discrepancy is indistinguishable from a bug.
The position worth taking
None of this is fully solvable and pretending otherwise is the actual failure. The workable posture is to measure the size of each gap, state it beside the number it affects (the rules we hold ourselves to are on how we measure), and pick one internally consistent source of truth per decision rather than reconciling three tools that were never going to agree.
Why your ad platform and your analytics disagree covers the reconciliation problem in more depth, and what a conversion rate hides covers the analysis errors that sit on top of these measurement ones.