Audience Overlap
by
Audience hygiene for ecommerce teams: find the audiences bidding against each other, price what the competition costs, then recommend which one to cut.
Ad accounts accumulate audiences the way a desktop accumulates files. A lookalike built last year, an interest stack from a campaign nobody remembers, three retargeting windows that were meant to be sequential and are now simultaneous. Many of them contain the same people, which means you are bidding against yourself: two of your own ad sets competing in the same auction, driving up the price you pay for a customer you were going to get anyway. This skill finds the overlapping audiences, prices what the competition costs, and recommends which one to cut.
What It Does
- Measures actual overlap between audiences rather than assuming from definitions.
- Prices the cost of self-competition in raised auction prices.
- Identifies which audience in an overlapping pair is the more efficient one.
- Finds the audiences that have degraded to near-duplicates of broad.
- Checks how much reach is genuinely lost by cutting each one.
- Recommends which audiences to cut, with the saving behind each.
Use Cases
- Stop two of your own ad sets bidding for the same person.
- Find the lookalike that has drifted into being broad targeting.
- Price what audience sprawl is costing before the next budget cycle.
- Cut audiences without losing reach you actually needed.
Example Session
Which of our audiences are fighting each other?
Three pairs overlap enough to matter. Fixing all three saves $9,400 a month.
- Audience pair | Overlap | Cost of competition | Better performer | Action
- Lookalike 1%, Lookalike 3% | 74% | $4,100 per month | 1% | Cut the 3%
- Retarget 7 day, Retarget 30 day | 61% | $3,200 per month | 7 day | Exclude 7 from 30
- Interest stack, Broad | 58% | $2,100 per month | Broad | Cut the stack
- Cart abandoners, Site visitors | 22% | Negligible | Both | Leave alone
The two lookalikes share three quarters of their members, so most of that spend is your own ad sets raising each other's prices. The interest stack has drifted to the point where it is 58% the same people as broad targeting and performs worse, which means it is broad with extra steps. The last pair overlaps only 22% and should be left alone.
Why This Beats Prompting It Yourself
Audience Definitions vs Who Is Actually In Them: The Hidden Risk
A generic prompt reads audience definitions and reasons about whether they sound distinct. A 1% and a 3% lookalike sound like different audiences and share three quarters of their members in practice. Definitions describe intent; only the measured overlap tells you whether two of your ad sets are bidding for the same person.
When You Still Need The Reach You Would Lose
Cutting an audience always saves money and sometimes costs reach you needed. This skill separates the two: the 3% lookalike is 74% duplicated so cutting it loses almost nothing, while an audience with 22% overlap is genuinely additive and stays. Without that check, audience cleanup turns into an unplanned budget reduction.
Known Limitations
Overlap measurement depends on what each platform exposes, which is precise within a platform and approximate across platforms, and the output distinguishes the two. The cost of self-competition is estimated from auction price movements when audiences run concurrently, which is directional rather than exact. It recommends the cuts and does not modify anything in the ad platforms.
Install It
- Download the ZIP.
It is free and there is no account to create. - Unzip it into your agent's skills folder.
Claude Code reads ~/.claude/skills/, which is hidden by default: the command in the folder block opens it. Other agents scan their own directory, so drop the same folder there instead. - Ask your agent to use it.
Restart the agent if it was already running, then it picks the skill up with no config.
skills folder
Copy
~/.claude/skills/audience-overlap/
SKILL.md
references/overlap-method.md
audits/
# one file per audit, with the saving per cut
# macOS: create the folder and open it in Finder
mkdir -p ~/.claude/skills && open ~/.claude/skills
# Windows: paste in the Explorer address bar
%USERPROFILE%\.claude\skills\Frequently asked questions
How is this different from the platform's audience overlap tool?
Platform tools show the overlap percentage. This prices what the overlap costs in raised auction prices, says which audience in each pair performs better, and checks how much reach a cut actually loses. The percentage alone does not tell you what to do.
Which agents does the skill run in?
Any agent that supports the open SKILL.md format: Claude Code, Cursor, Codex CLI, GitHub Copilot, Gemini CLI, Manus, Grok and others load it unmodified. The format is portable, the location is not. Each agent scans its own skills directory, so you drop the same folder into whichever one yours uses. There is no config file to edit and no API key to provision.
What does it need connected to work?
Audience definitions and sizes, spend and performance per ad set, and overlap data from the platforms. Where a platform does not expose overlap the skill estimates it from performance patterns and labels it as an estimate.
How does overlapping cost money?
Two of your ad sets entering the same auction raise the clearing price you pay, so you win the same impression for more. It is a real cost and an invisible one, because each ad set looks separately reasonable in its own report.
Is it safe to cut a lookalike that overlaps 74%?
Yes, and that is the point of the reach check: three quarters of its members are already reachable through the audience you keep. The skill flags any cut where the unique reach lost is large enough to matter, which is rare above about 60% overlap.
What is included with the skill?
The skill itself, the overlap method reference, the audit format, and the competition cost model. It is free and security scanned. We re-publish the ZIP when the model changes, so download it again if your audits start shifting.
Two audiences look separate until the revenue is deduplicated.
Polar reconciles 45+ sources behind one customer record, so the overlap you cut is real overlap and not the same buyer counted twice.
<script type="application/ld+json">{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"How is this different from the platform's audience overlap tool?","acceptedAnswer":{"@type":"Answer","text":"Platform tools show the overlap percentage. This prices what the overlap costs in raised auction prices, says which audience in each pair performs better, and checks how much reach a cut actually loses. The percentage alone does not tell you what to do."}},{"@type":"Question","name":"Which agents does the skill run in?","acceptedAnswer":{"@type":"Answer","text":"Any agent that supports the open SKILL.md format: Claude Code, Cursor, Codex CLI, GitHub Copilot, Gemini CLI, Manus, Grok and others load it unmodified. The format is portable, the location is not. Each agent scans its own skills directory, so you drop the same folder into whichever one yours uses. There is no config file to edit and no API key to provision."}},{"@type":"Question","name":"What does it need connected to work?","acceptedAnswer":{"@type":"Answer","text":"Audience definitions and sizes, spend and performance per ad set, and overlap data from the platforms. Where a platform does not expose overlap the skill estimates it from performance patterns and labels it as an estimate."}},{"@type":"Question","name":"How does overlapping cost money?","acceptedAnswer":{"@type":"Answer","text":"Two of your ad sets entering the same auction raise the clearing price you pay, so you win the same impression for more. It is a real cost and an invisible one, because each ad set looks separately reasonable in its own report."}},{"@type":"Question","name":"Is it safe to cut a lookalike that overlaps 74%?","acceptedAnswer":{"@type":"Answer","text":"Yes, and that is the point of the reach check: three quarters of its members are already reachable through the audience you keep. The skill flags any cut where the unique reach lost is large enough to matter, which is rare above about 60% overlap."}},{"@type":"Question","name":"What is included with the skill?","acceptedAnswer":{"@type":"Answer","text":"The skill itself, the overlap method reference, the audit format, and the competition cost model. It is free and security scanned. We re-publish the ZIP when the model changes, so download it again if your audits start shifting."}}]}</script>