How to Audit a LinkedIn Ads Agency Before You Sign

Checking your own setup? Run the free LinkedIn Insight Tag Checker to inspect browser requests before reviewing account-side attribution.

Choosing a LinkedIn Ads agency is a decision you make on a call and find out about in month four. Every pitch deck contains the same nouns — attribution, full-funnel, CRM integration — and none of them are falsifiable in an hour. So stop trying to evaluate the agency. Audit the account instead. Five checks you can run yourself, in Campaign Manager and your CRM, produce evidence that makes the sales conversation answerable.

You cannot tell a good agency from a bad one by listening to them. You can tell by handing them findings and watching what they do with them.

Why the usual LinkedIn Ads agency due diligence tells you nothing

The standard checks are case studies, client logos, a Clutch rating and a reference call. Each has a structural problem. Case studies are written by the agency, name the metric that moved, and omit the ones that did not. Logos prove somebody signed, not that anybody renewed. References are chosen by the vendor. None of it tells you whether the people who would touch your account can wire a server-side event to a CRM stage.

Certifications are weaker than they look. LinkedIn Marketing Partner status is real, and it is also a commercial relationship with revenue thresholds attached. It says an agency spends at scale. It does not say who does the implementation, or whether that person is on your account.

The alternative is cheap. Everything below runs on access you already have, in about 40 minutes, and every finding is a fact rather than a claim.

Run the five checks before you take the call

Do them in order. Each one narrows what the next one can mean, which is why a checklist beats a list of questions: by step four you are not asking whether the account has a problem, you are asking which one.

Step one is the one people skip, and it decides the value of every number after it. A conversion rule sourced from the Insight Tag fires in the browser, so ad blockers, cookie expiry and iOS privacy remove a share of events before the platform ever sees them. A rule sourced from the Conversions API is sent server-side and survives all three. Reading the data-source column is a thirty-second check that tells you whether the report you are about to argue over is a measurement of the market or a measurement of browsers. The full implementation sequence is a separate job; here you only need to know which one you have.

What each finding actually means

A finding is not a verdict. The same symptom has two very different causes, and the pairing below is what stops you buying the wrong engagement.

What you found	What it means	What to ask the agency
Every rule says Insight Tag	Browser-only signal; blocked events never reach the platform	Who writes the server-side events, and against which CRM object?
No campaign cleared 50 events	Structure problem, not a creative problem	Which campaigns would you merge before touching the ads?
CRM has no LinkedIn field on won deals	The loop was never closed at the CRM end	Which CRM stages get pushed back, and who owns that field?
Nobody can state cost per closed-won	Reporting stops at the lead	What will the monthly report show on line one?
All four are clean	Measurement is not your problem	Then the engagement is campaign work, and it should be priced as that

The last row matters as much as the first four. If the account is clean, measurement work is not what you need, and an agency that sells you a tracking rebuild anyway is doing to you what this post is trying to prevent.

Turn the findings into questions a LinkedIn Ads agency cannot deflect

Hand over what you found and ask for a 30-day plan against it. You are not testing knowledge; you are testing whether the answer names systems, objects and owners. “We would tighten up your tracking” is a sentence anyone can produce. “We would send the sales-accepted stage from HubSpot as a server-side event, deduplicated against the existing pixel rule, then move the campaign target off form fill once it has thirty days of history” is a sentence only somebody who has done it can produce.

Three follow-ups separate the two. Ask who writes the integration — the person on the call, a partner, or your own developer. Ask what happens to the campaign’s learning when the target event changes, because changing the target event discards the model and an agency that does not volunteer that is going to surprise you in week three. Ask what the first monthly report shows on line one. If the answer is cost per lead, you have just been told what the engagement will optimize toward, whatever the deck said.

These are diagnostic questions, not the five questions that separate attribution access from attribution implementation — that set is about capability in general, this one is about your account in particular.

Price the LinkedIn Ads agency engagement against what the audit found

The findings should change the shape of the deal, not just the vendor. An account that fails step one has a fixed-scope implementation problem, and implementation problems are badly served by a monthly retainer that starts before the plumbing works. An account that passes all four has a campaign problem, and campaign problems are badly served by an audit you have already run yourself.

Spend is the other half of the arithmetic. LinkedIn needs roughly 50 conversion events per campaign per month for delivery to stabilise, which is about $5,000 per campaign at a $100 cost per lead, and budget is set per campaign rather than per account. Below roughly $8,000 a month of cold acquisition spend the mechanism cannot run, and an agency quoting a management fee on top of that is selling you a service that cannot mathematically succeed. Ask any agency what their floor is and what it is derived from. A floor with arithmetic behind it is a disclosure. A floor without one is a sales qualification.

What a good 30-day answer contains

Plans are easy to write and hard to fake at the level of the object being changed. A serious first month is sequenced, because each step invalidates the measurement of the one before it, and a plan that runs them in parallel is a plan nobody has executed. Five things should appear in order:

1. A server-side event, named. Not “we’ll fix tracking” but which event, sent from which system, against which CRM object, deduplicated against the existing pixel rule so the same conversion is not counted twice.
2. A baseline captured before anything changes. Thirty days of conversion volume on the current setup, written down. Without it there is no before-and-after, and every improvement claimed afterwards is unfalsifiable.
3. A structure decision, with campaigns named. Which campaigns merge, which stop, and what each survivor’s monthly budget becomes. If the audit found no campaign clearing 50 events, this step comes before any creative work and a good plan says so.
4. A stated waiting period. Changing a campaign’s target event restarts learning, so the honest plan books roughly 30 days of not judging the result. An agency promising a lift in week two has either not changed the target event or is going to report on noise.
5. The first report’s line one. Agreed in advance, in writing, before any money moves. Whatever sits on line one is what the engagement will optimise toward for the next year.

Notice what is absent: audience refinement, creative testing, a content calendar. Those are real work and they are month two. Any of them appearing in week one on an account that failed step one means the plan was written before anybody looked at the account.