The hard part of hiring a measurement consultant is that you usually can't evaluate the work. If you could audit a GA4 implementation yourself, you wouldn't be hiring someone to do it.
So the vetting has to be indirect: ask questions whose answers you can judge even when the underlying work is opaque to you. These are the twelve I'd use, and what separates a good answer from a plausible one.
Before you talk to anyone
Write down the decisions the data is supposed to support. Not the metrics — the decisions. "Which channels get next quarter's budget." "Whether the free trial length should change." "Which landing pages to kill."
This takes twenty minutes and does more to improve the outcome than any question below, because it converts a vague request ("fix our analytics") into a scope. It also immediately exposes anyone who wants to skip straight to implementation.
The twelve questions
1. "What would you look at first?"
Listen for: a diagnostic sequence that starts with comparing platforms against each other and against your backend or CRM.
Warning sign: an answer that starts with what they'd build. Anyone proposing a solution before seeing your data is selling a template.
2. "How will you know what's currently broken?"
Listen for: reconciliation. Specifically, comparing GA4, the ad platforms, and your source of truth, and quantifying the gaps between them.
Warning sign: "I'll audit the GA4 property." That's where the symptoms are, not the causes.
3. "What if the answer is that we don't need this project?"
Listen for: a straightforward yes, ideally with an example of when they've said it.
Warning sign: any version of "that never really happens." It happens constantly. The most common finding in measurement work is that the existing setup is misconfigured in two or three places, not that the business needs new infrastructure.
4. "Do we need server-side tagging?"
This is a trap question, and a useful one. The correct answer in the majority of cases is "probably not, and here's how we'd find out."
Warning sign: an immediate yes. Server-side tagging is real and valuable when you can measure the loss it recovers, and an expensive answer to the wrong question when you can't. Someone who recommends it before measuring is pattern-matching on your budget.
5. "How do you name events?"
Listen for: a convention, stated without hesitation, and awareness of GA4's actual constraints — reserved names and prefixes, the 40-character limit, case sensitivity, and the fact that custom dimensions are a budgeted resource rather than something you add per event.
Warning sign: "whatever makes sense for your business." Naming is the thing that makes the data usable in two years, and improvised naming is why most properties become unqueryable.
6. "What happens to the dataLayer when our developers ship a redesign?"
Listen for: an acknowledgement that this is the main failure mode, and a plan — instrumentation in the application source rather than scraped from CSS selectors, and a spec the dev team owns.
Warning sign: confidence that it won't be a problem. It is always a problem. Tracking built on page structure breaks the moment the page changes, and nobody notices for a quarter.
7. "How will consent be handled?"
Listen for: Consent Mode v2 signals gating tags, and a clear statement that where processing happens doesn't change whether you need a legal basis.
Warning sign: any framing of server-side tagging as a way around consent requirements. That's a compliance liability being sold as a feature.
8. "What will you hand over?"
Listen for: documentation, container exports, the measurement plan, and a walkthrough with whoever maintains it next.
Warning sign: vagueness. Vague handover is how a project becomes a subscription.
9. "Who specifically will do the work?"
Listen for: a name, and an offer to introduce you before you sign.
Warning sign: "our team." Ask again. This question separates the person who impressed you in the pitch from the person who'll be in your account, which is the single biggest predictor of outcome when hiring an agency.
10. "What don't you do?"
Listen for: an actual boundary. Statistical modelling and data science, campaign management at scale, and creative are all reasonable things for a measurement specialist to decline.
Warning sign: nothing. A person claiming everything from attribution modelling to ad creative to development is describing a team — go back to question 9 — or overselling.
11. "How much will this cost, and what drives the number?"
Listen for: cost drivers named specifically: how many platforms need conversion endpoints, whether the event model exists, whether infrastructure is involved.
For reference, independent 2026 market guides put a full server-side tagging implementation in the region of $8,000–$18,500, with hosting from roughly $50–400/month depending on traffic. A GA4 and GTM implementation without server-side infrastructure is a smaller project.
Warning sign: a number with no drivers behind it, or a refusal to model cost before you commit.
12. "Can I see something you've built?"
Listen for: anything verifiable. Published tools, open-source work, documented methodology, products they've shipped.
Warning sign: only testimonials. This one needs a caveat, though — see below.
On testimonials, specifically
Testimonials are the weakest signal on this list, and I say that as someone who doesn't have publishable ones.
Measurement work sits on top of revenue data, conversion rates, and ad spend. The engagements most worth writing about are often exactly the ones covered by confidentiality, so the correlation between "has public case studies" and "is good at this" is much weaker than it looks. Meanwhile, unverifiable outcome claims — "increased conversions 40%" with no baseline, no attribution window, and no way to check — are trivially easy to write.
Better substitutes, in rough order of strength:
- Work you can inspect. Tools that run in your browser, open-source repositories, published specifications. Put in numbers whose answer you already know and see whether it agrees.
- Verifiable credentials. Google Ads and Google Analytics certifications tied to a real account.
- Documented methodology. A public description of how engagements run, including scope boundaries, is harder to fake than a quote.
- A reference call. Genuinely useful when available — just not a precondition.
Ask for whichever of these exist. Weight the ones you can check yourself most heavily.
The shortest version
If you only ask three:
- "What would you look at first?" — tests diagnosis before solution.
- "What if we don't need this?" — tests whether you're buying advice or a predetermined deliverable.
- "Who specifically will do the work?" — tests everything else you were told.
My own answers to all twelve are public: the engagement model, the five stages, and the explicit list of what I don't take on. The GA4 event name checker is question 5's answer as a working tool, and the Consent Mode modeling checker is question 7's.