A YouTube audit should not begin with view rate. It should begin with the business decision the campaign was designed to influence. A reach campaign, a Video views campaign and a conversion-focused Demand Gen campaign buy different outcomes and count interactions differently. Combining them into one "YouTube performance" table produces confident but unusable conclusions.
One platform change is essential in 2026: Google states that all Video Action campaigns were upgraded to Demand Gen by May 2026. If an audit still recommends creating a new Video Action campaign, it is working from an obsolete playbook. The Google Ads migration documentation confirms the change.
1. Write the campaign contract
For each campaign, record its job, audience stage, intended inventory, primary KPI, guardrail and evaluation window. For example: "Introduce the new range to UK category buyers; maximise qualified views at or below the agreed CPV; judge brand-search and site-engagement movement after six weeks." Do not retrofit the objective after seeing the results.
Map campaign type to job:
| Job | Likely campaign | Primary evidence |
|---|---|---|
| Efficient reach | Video reach | Unique reach, frequency, CPM, on-target delivery |
| Consideration and views | Video views | TrueView views, format-specific view rate, CPV, engaged sessions |
| Sales or leads | Demand Gen | Conversions, value, CPA or ROAS, qualified backend outcomes |
| Channel engagement | Demand Gen YouTube engagements | Subscriptions or follow-on views, cost per action |
2. Verify goals and event quality
Open Goals > Conversions > Summary. Check which actions are Primary, their values, counting method, attribution model and windows. Then open each campaign's goal settings. A conversion campaign optimising to newsletter signups and purchases together is not necessarily optimising to profitable customers.
Segment the campaign by Conversions > Conversion action. For video, also use Conversions > Ad event type to separate click-associated, engaged-view and impression-associated conversions. Google's engaged-view conversion guide explains that the event is designed for behaviour where a viewer watches but does not click before converting later.
An engaged-view conversion is an observed attribution event under Google's rules. It is not proof that the ad caused the conversion. Reconcile purchases with orders and leads with CRM stages.
3. Audit campaign and channel settings
Check locations, location options, languages, devices, schedule, excluded content, inventory type and conversion goals. For Demand Gen, inspect channel controls at ad-group level. Google now allows manual selection among YouTube in-stream, in-feed, Shorts, Discover, Gmail, Maps and the Google Display Network where available.
Use Segment > Network at campaign, ad-group and ad level. For YouTube traffic, segment by ad format to distinguish skippable in-stream, in-feed and Shorts. Google's Demand Gen channel-control guide documents both the settings and the reporting breakdown.
If the campaign brief says "YouTube only" but spend appears on Discover or Gmail, that is a setup issue. If all channels were intentionally enabled and YouTube received little spend, it is an allocation outcome, not evidence that ads failed to serve.
4. Audit audiences as rules and signals
List every segment in each ad group: customer lists, site visitors, YouTube users, custom segments, in-market, affinity, demographics and lookalikes. Note whether optimised targeting or audience expansion is active. Distinguish strict controls such as location and exclusions from audience suggestions the system may expand beyond.
Review audience overlap and creative fit. One ad group combining previous purchasers, cart abandoners and broad category prospects cannot produce a clean remarketing result. A custom segment built from competitor URLs does not mean the people reached visited those exact sites; it helps Google estimate an audience with related intent.
Google's Demand Gen audience overview recommends separating distinct audience strategies into ad groups when creative also differs. Do this only when budget and volume can support the split.
5. Review creative frame by frame
Use the YouTube creative-testing framework to turn the findings into controlled hypotheses. Watch every video on mute, on a phone and at normal speed. Score evidence rather than taste:
- First two seconds: can the intended viewer identify relevance before the skip decision?
- Product clarity: is the offer understood without prior brand knowledge?
- Proof: is the claim demonstrated, quantified or supported?
- Pacing: does each shot add information?
- Branding: can viewers identify the advertiser without waiting for the end card?
- Action: is the next step and destination clear?
- Format: are vertical, square and landscape edits composed for their surfaces?
Then open the Ads or Assets report. Add impressions, cost, clicks, conversions, value, TrueView views and the format-specific view-rate columns supported by the campaign. Since October 2025 Google has renamed the general Views metric to TrueView views, and format-specific rates matter because a Shorts view and an in-stream view are counted differently. See Google's Video views documentation.
6. Inspect placements and suitability
Review Content > Where ads showed for campaign types that expose placement reporting. Sort by cost and impressions, then look for brand-safety concerns, children-oriented inventory, parked or low-context placements and obvious geographic or language mismatch. Use placement exclusions for suitability, not as a ritual deletion of every placement without a direct conversion.
Conversion campaigns can find value across many placements, and low-volume placement rows are noisy. Excluding aggressively from a handful of impressions narrows inventory without demonstrating improved business performance.
7. Follow the click to the page
Test every final URL on mobile. Confirm the page repeats the ad's product, promise, price and proof. Check load behaviour, consent banners, variant selection, shipping clarity, stock status, form errors and analytics parameters. Compare landing pages by cost and conversion outcome.
A strong video can create interest that a generic homepage then wastes. The YouTube landing-page guide provides the full mobile and message-match diagnostic. Conversely, a low click-through rate does not automatically implicate the page because most exposed users never arrived there. Diagnose the stage where the loss occurs.
8. Read metrics in sequence
Start with delivery: spend, impressions, reach, frequency and format. Then attention: TrueView views, format-specific view rate and quartile metrics where available. Then traffic: clicks, click-through rate and engaged sessions. Finally outcomes: conversion actions, CPA, value, ROAS and backend quality.
Use the sequence diagnostically:
- Low delivery: check eligibility, bids, budget, targeting and policy.
- Delivery but weak attention: investigate hook, relevance and format composition.
- Attention but little traffic: assess proposition and call to action, while remembering that not every awareness ad should maximise clicks.
- Traffic but weak conversion: check message match, page experience, offer and tracking.
- Reported conversion but weak backend value: fix goals, values, attribution expectations or lead quality.
9. Test bidding against the objective
A Video views campaign uses target CPV to seek TrueView views. Reach campaigns commonly use CPM-based bidding. Demand Gen can use conversion- or value-focused strategies depending on eligibility and setup. Judge each against its own optimisation job.
Inspect the bid strategy report and conversion lag before changing targets. If a conversion campaign receives few outcomes, splitting into many audiences and ads may leave every cell underpowered. Consolidate until there is enough data, or accept that the test is exploratory rather than statistically decisive.
10. Reconstruct what changed
Filter Change history by campaign and date. Plot material edits against spend, channel mix, conversion lag and site events. Include creative launches, audience changes, target changes, product-feed connections and automatic recommendations. A performance shift that predates the alleged cause is not explained by it.
The audit deliverable
A useful audit ends with a prioritised decision log, not 60 screenshots. For each issue include evidence, commercial consequence, proposed change, owner, review date and what would falsify the hypothesis.
Example: "Shorts received 62% of spend after vertical creative launched; landing-page conversion was lower than in-stream. This is observed. We infer the page or offer may not suit the faster context. Test a Shorts-specific page and hold audience and offer constant." That language is more honest than "Shorts traffic is low quality."
What the audit cannot settle
Platform attribution cannot establish incrementality on its own. View and click metrics describe interaction, not profit. Audience and asset reports are not randomised tests. If the budget decision is large, use video or Demand Gen experiments, Brand Lift when eligible, or a properly designed geographic holdout.