Audience and Remarketing Audit: Who You're Really Paying to Reach Again
Remarketing is the campaign everyone loves in the reports.
The return on ad spend is spectacular. The cost per conversion is tiny. On every dashboard, the retargeting line glows. And so it keeps its budget, year after year, never seriously questioned, because why would you question the best-performing thing in the account?
Here is why. The people you are remarketing to are, by definition, people who already visited your site. Many of them were coming back anyway. When they convert, remarketing takes the credit, but a large share of those sales would have happened without the ad. The spectacular return is partly real and partly an accounting illusion, and the only way to tell how much is which is to audit it properly.
Audiences and remarketing reward scrutiny more than almost any other part of an account, precisely because they look so good that nobody gives them any.
The Incrementality Problem at the Heart of It
The core question for any remarketing activity is not "what return did it report?" It is "how much of this would have happened anyway?"
Remarketing advertises to warm audiences: past visitors, cart abandoners, previous customers. These people already know you. Some meaningful proportion of them intended to return regardless of whether they saw your ad. When they do return and buy, the platform attributes the sale to the remarketing campaign, which makes it look like the campaign created the sale. Often it merely intercepted a sale that was coming.
This does not make remarketing worthless. Some of it drives genuinely incremental sales, people who would have drifted away without the nudge. The point of the audit is to find out how much, rather than taking the flattering headline figure at face value.
Audit for: whether any attempt has ever been made to measure the incremental value of remarketing, through holdout tests or similar, or whether it is funded purely on attributed return. An account that has never questioned its remarketing is almost certainly overvaluing it.
Observation vs Targeting: The Setting That Changes Everything
Before the strategy, there is a mechanical setting that quietly changes what audiences do, and it is frequently set wrong or misunderstood.
Audiences can be applied in two modes.
Targeting (or "audience segment with narrow targeting"): restricts the campaign to only show to people in that audience. It narrows reach.
Observation (or "monitoring"): does not restrict anything. It lets you watch how an audience performs and apply bid adjustments, while the campaign still reaches everyone.
Getting these confused causes real damage. An audience set to "targeting" when you meant "observation" silently strangles a campaign's reach, limiting it to a small audience when you thought it was running broadly. The reverse, meaning to restrict but leaving it on observation, means a campaign you thought was tightly targeted is actually showing to everyone.
Audit for: whether each audience is in the mode you intended, whether any campaign has been accidentally restricted by a targeting-mode audience, and whether observation audiences are actually being used to inform bids or just sitting there unused.
Auditing the Remarketing Lists Themselves
The lists underneath the campaigns need checking, because they decay and misconfigure quietly.
Audit for:
Membership duration: whether the lookback window suits the buying cycle, too short and you miss returning buyers, too long and you chase people who have lost interest List size: whether lists are large enough to actually serve, or too small to function Whether lists are segmented meaningfully (cart abandoners, past purchasers, specific page visitors) or lumped into one undifferentiated "all visitors" pool Whether the tags and data feeding the lists are still firing correctly, as tracking changes often break list population silently Whether customer match lists (uploaded first-party data) are current, or last updated long ago
A remarketing programme running on stale, broken or undifferentiated lists is spending money on the wrong people while looking like it works.
Audience Exclusions: The Overlooked Half
Almost everyone thinks about who to include. Far fewer think about who to exclude, and the exclusions are often where the value is.
Audit for:
Whether existing customers are excluded from acquisition campaigns, or whether you are paying to acquire people you already have Whether converters are excluded from prospecting, so you stop advertising to people who just bought Whether remarketing and prospecting audiences overlap, competing for the same people Whether a new-customer acquisition goal or similar is used to focus spend on genuinely new customers Whether audiences that never perform are excluded rather than left running
For lead generation and ecommerce alike, advertising heavily to existing customers while reporting it as acquisition is one of the quietest and most common sources of waste. Exclusions are how you stop it, and most accounts barely use them.
Audience Signals in Automated Campaigns
In Performance Max and similar automated campaigns, audiences work differently, as signals that guide the system rather than hard targeting, and they still deserve audit.
Audit for:
Whether audience signals reflect the actual target customer or just a generic remarketing list Whether first-party data is being used as a signal where available, which tends to help Whether the signals are specific enough to usefully guide the system Whether the automated campaign is leaning on existing-customer audiences and reporting the resulting sales as acquisition
A weak or lazy audience signal, just dropping in an all-visitors list, gives the automation little useful direction and can steer it toward harvesting existing demand.
Audience Overlap and Cannibalisation
When multiple campaigns target overlapping audiences, they can compete with each other and muddy the data.
Audit for:
Whether prospecting and remarketing campaigns are bidding on the same people Whether multiple remarketing campaigns overlap in membership Whether the overlap is splitting data and inflating apparent remarketing performance Whether a clear audience hierarchy exists, so each person is reached by the intended campaign, not several at once
Overlap is doubly damaging: it wastes spend through internal competition and it makes the remarketing numbers look better than they are, because the same warm users get counted across campaigns.
Scoring and Fixing
Rank the findings rather than listing them.
The highest-value fixes are usually the exclusions, excluding existing customers from acquisition and converters from prospecting, because they stop clear waste immediately and are low-risk. Then the observation-versus-targeting settings, which can be silently throttling or over-exposing campaigns. Then the incrementality question, which is more involved but reframes how much the whole remarketing programme is really worth. Be careful not to slash remarketing to zero on suspicion alone; the right response to "this looks non-incremental" is to test it, then reallocate what proves to be waste.
Common Audience and Remarketing Mistakes
Funding remarketing on attributed return alone It takes credit for returning customers who were coming anyway. Measure incrementality, do not trust the flattering headline.
Confusing observation and targeting mode Targeting mode silently restricts reach; observation mode does not restrict at all. Mixing them up throttles or over-exposes campaigns.
Never excluding existing customers Paying to acquire people you already have, and reporting it as acquisition, is quiet, common waste.
Stale or undifferentiated lists Broken tags, wrong lookback windows, or one big "all visitors" pool mean you remarket to the wrong people.
Overlapping audiences competing Prospecting and remarketing bidding on the same people wastes spend and inflates remarketing's apparent performance.
Lazy audience signals in PMax A generic all-visitors signal gives automation no useful direction and steers it toward existing demand.
How AI Can Help
Cross-referencing audience settings, list configuration, exclusions and overlap across an account, and questioning where remarketing credit is really incremental, is detailed work easy to skip.
AI can process:
Which audiences are in targeting versus observation mode, and whether that fits Whether existing customers and converters are excluded where they should be List configuration, membership windows and population health Audience overlap between prospecting and remarketing Where remarketing's attributed return is likely non-incremental Ranked audience fixes by impact and confidence
GoogleAdsAudits.com reviews account structure, tracking, bidding, creative, feed and channel mix, then produces a prioritised action plan positioning each recommendation by impact, effort and confidence.
The judgement about which customers genuinely need reaching again still rests with the business.
Final Thoughts
Remarketing and audiences get the least scrutiny because they look the best, and that is exactly backwards. The spectacular remarketing return is partly real and partly credit for sales that were already coming, and the only way to tell the two apart is to audit it: measure the incremental value rather than trusting the headline.
Alongside that, check the mechanics that quietly go wrong: the observation-versus-targeting setting that throttles or over-exposes campaigns, the stale and undifferentiated lists, and above all the exclusions, because paying to re-reach people who were always coming back, or to acquire customers you already have, is the most flattering waste in the account.
Scrutinise the thing that looks too good to question. That is usually where the money is.
Build a personalised Google Ads action plan that identifies wasted spend, scalable opportunities and the fixes your team should prioritise next.
Frequently Asked Questions
Why is remarketing often less valuable than it looks?
Because you advertise to people who already visited your site, many of whom were returning anyway. When they convert, remarketing takes the credit, but a large share of those sales would have happened without the ad. The reported return is partly real and partly an attribution illusion.
What is the difference between observation and targeting mode for audiences?
Targeting mode restricts a campaign to only show to that audience, narrowing reach. Observation mode restricts nothing; it lets you monitor performance and adjust bids while the campaign still reaches everyone. Confusing the two can silently throttle or over-expose a campaign.
Should I exclude existing customers from my campaigns?
Usually, from acquisition campaigns, yes, so you are not paying to acquire people you already have and reporting it as new business. Converters should also typically be excluded from prospecting. Exclusions are one of the highest-value and least-used parts of audience management.
How do I know if my remarketing is incremental?
By testing, usually a holdout where some of the audience does not see the ads, and comparing. Funding remarketing purely on attributed return overvalues it. If no incrementality test has ever been run, the programme is almost certainly overvalued.
What should I check on my remarketing lists?
Membership duration against the buying cycle, whether lists are large enough to serve, whether they are segmented meaningfully rather than one big pool, and whether the tags feeding them still fire, since tracking changes often break list population silently.
How do audiences work in Performance Max?
As signals that guide the system rather than hard targeting. Audit whether the signals reflect the real target customer rather than a generic all-visitors list, whether first-party data is used, and whether the campaign is leaning on existing-customer audiences and reporting those sales as acquisition.
What is audience overlap and why does it matter?
When multiple campaigns target the same people, such as prospecting and remarketing bidding on overlapping audiences. It wastes spend through internal competition and inflates remarketing's apparent performance, because the same warm users get counted across campaigns.