Skip to content
Primores AI wiki

Engaged-View Conversions (EVC): Why YouTube ROAS Reads Too High

Engaged-View Conversions (EVC)

TL;DR: An engaged-view conversion is credited when someone watches a YouTube ad — ~10 seconds of a skippable ad, or the whole thing if it’s shorter — without clicking, and then converts within an attribution window (Google default ~3 days). Unlike classic view-through conversions, EVCs are counted inside the main “Conversions” column for Video-action and Demand Gen campaigns. That single design choice is why a YouTube report can show more conversions than clicks and an implausible ROAS: most of the credited “conversions” had no click at all — the ad was seen near a purchase, not proven to have caused it. On a high-traffic site with strong organic/branded demand, EVC inflation is maximal, because there’s a large pool of people who would have bought anyway. The fix is causal: strip to click-through only as a floor, reconcile against the merchant’s own order backend, and ultimately run glossary/incrementality-testing.

What it means

Google Ads counts conversions from YouTube video ads in more than one way:

Conversion typeTriggerClick required?Counted in main “Conversions” column?
Click-through conversion (CTC)User clicks the ad, then convertsYesYes
Engaged-view conversion (EVC)User watches ~10s+ of a skippable ad (or the full ad if shorter), doesn’t click, converts within the window (default ~3 days)NoYes — this is the key point
View-through conversion (VTC)An impression was served, no engagement, later conversionNoUsually a separate column

(Windows and inclusion are Google-configurable and Google changes defaults over time — treat the ~3-day EVC window and column placement as the standard defaults, not fixed law. Verify in the account.)

The load-bearing fact: EVCs live inside the headline Conversions number for video-action campaigns. CTCs are bounded by clicks; EVCs are not. So the moment engaged-view credit is switched on, the reported conversion count is no longer tied to how many people actually clicked.

The decisive tell: conversions > clicks

For a single purchase event on a normal-priced product, one click can produce at most roughly one purchase. So if a YouTube report shows credited conversions greater than clicks, it is arithmetically certain that the majority of the credited conversions had no click — they are engaged-view (or view-through) credit. This is the single most decisive diagnostic for EVC inflation: it doesn’t require trusting anyone’s model, only counting.

Two secondary tells travel with it:

  • Fractional / decimal conversion counts — the signature of modeled conversions: Google statistically estimates conversions it can’t directly observe (consent-mode and cookie gaps) and fills them in. Part of the number isn’t measured; it’s inferred.
  • An implausible ROAS or CPA. Restate the ROAS as a cost-per-acquisition — a 60× ROAS at €60 AOV implies a €1 CPA on cold-ish video traffic. Stated as CPA, the impossibility is obvious in a way the ROAS headline hides.

Why it inflates most on high-demand sites

Attribution inflation scales with the organic baseline. A site doing millions of visits a month of branded/organic demand has a large population of people who were going to buy regardless. Point an engaged-view campaign at that audience and Google credits every watch-then-buy as a conversion the ad “drove” — when the ad merely stood in front of a purchase river and claimed the water.

This is the same mechanism the wiki flags for glossary/performance-max (branded-search cannibalization) and names generally in glossary/incrementality-testing: “when attribution looks too good… the dominant failure mode is that attribution is crediting the channel for conversions that would have happened anyway.” EVC is that failure mode in its purest, most YouTube-specific form.

Two inflation sources to separate

Practitioners conflate these; they compound and have different fixes:

  1. Attribution inflation (the big one) — EVC/VTC crediting sales the ad didn’t cause. Fix: strip to click-through; ultimately a holdout.
  2. Double counting (a config bug) — the same order counted twice: a Google Ads conversion tag and a GA4 conversion import both firing, or a purchase action set to count “Every” instead of “One.” Fix: audit the tag setup and dedupe.

Modeled conversions (the fractional counts) are a third layer — not “fixable,” just a reminder that some of the total is an estimate, not an observation.

What to actually believe — in order of trust

  1. Reconcile against the merchant’s own order backend. The store knows exactly how many orders shipped in the window. Compare UTM/coupon-tagged orders in the merchant’s system to Google’s claimed count. If Google claims ~1,000 and the backend shows ~40, that gap is the inflation — measured against ground truth Google can’t touch. Fastest convincing check; an afternoon’s work.
  2. Segment the report by conversion type. Add CTC / EVC / VTC columns in Google Ads. EVC+VTC dwarfing CTC confirms the diagnosis in minutes, from Google’s own data.
  3. Rebuild as click-through-only, short (e.g. 1-day-click) window, on a verified purchase event, EVC/VTC off. This is the believable floor — but note it still over-credits, because last-click on a high-demand site hands credit to an ad that merely sat on the path (see glossary/multi-touch-attribution).
  4. Run an incrementality test (glossary/incrementality-testing) — a geo-holdout you run yourself is the un-fakeable answer and, unlike a platform lift study, isn’t Google grading its own homework. A Google Conversion Lift study uses the same RCT logic done formally; more credible than attribution, but carries a self-measurement discount.

Honest limits

  • EVC is not fraud or a bug — it’s a legitimate measurement option that captures real view-driven influence for genuinely upper-funnel video. The problem is defaulting to it for a direct-response campaign on a high-demand site and reading the result as caused sales.
  • CTC-only is a ceiling on the believable, not truth. It removes view credit but not the last-click over-crediting of pre-existing demand.
  • The exact window and column defaults change. Google adjusts EVC behavior over time; always verify the account’s actual attribution settings rather than assuming.
  • View-through influence can be real for brand campaigns. The critique here is about counting it as direct-response ROAS, not about denying that awareness ever converts later.

Key Takeaways

  • EVC credits a watch-then-buy with no click — and sits inside the main Conversions column for YouTube video-action / Demand Gen campaigns.
  • Conversions > clicks is the decisive tell — arithmetically proves most credited conversions never clicked.
  • Fractional counts = modeled conversions; absurd ROAS is clearer as an absurd CPA.
  • Inflation scales with the organic baseline — worst on high-traffic, branded-demand sites.
  • Believe, in order: backend order reconciliation → conversion-type segmentation → click-through-only floor → geo-holdout incrementality.
  • Separate attribution inflation from double-counting — different problems, different fixes.

Sources

Do-not-cite as fixed: exact EVC window (defaults change; verify in-account), and the “counted in main column” behavior for every campaign type (true for Video-action / Demand Gen; verify per campaign).

By Andrej Ruckij