10 Questions to Ask Your Media Agency About Where Your Budget Actually Goes

In short: you don’t need to be a fraud expert to hold a media agency to account. Ask for your verified human reach, your invalid-traffic rate, whether filtering is pre-bid, and for device-level proof — then watch how confidently each one is answered.

The ANA’s supply chain study found that only around 36% of a programmatic dollar reaches a real consumer (ANA, 2023). The remainder disappears into fees, made-for-advertising inventory, non-viewable impressions and invalid traffic. None of that is a secret, and none of it requires technical expertise to ask about.

What follows is a working list. Take it to your next review and pay as much attention to the manner of each answer as to its content — the difference between a partner who measures delivery and one who resells a dashboard shows up in how quickly the numbers appear.

1. What share of my spend reached verified humans?

The opening question, and the most revealing. If the answer is “all of it”, you are being shown served impressions, not independently verified reach.

A good answer: a specific percentage, with the method used to arrive at it.

2. What was my invalid-traffic rate?

Every campaign has one. A credible partner knows the number for your account rather than quoting an industry average, and the programmatic average sits around 20% (Juniper Research, 2023–2028 forecast).

A good answer: your rate, broken down by channel and publisher.

3. Do you filter pre-bid, or only report post-bid?

This is the question with the most money attached. Pre-bid filtering stops the impression being bought at all; post-bid reporting documents a loss you have already taken, and credits are discretionary.

A good answer: pre-bid, with a description of what gets blocked before the auction.

4. Can I see device-level proof?

Aggregated dashboards are easy to flatter — averages hide the placements doing the damage. Device-level evidence is considerably harder to dress up.

A good answer: yes, with a sample you can inspect rather than a summary slide.

5. How do you handle made-for-advertising sites?

MFA inventory passes standard brand-safety filters because nothing on it is unsafe — it is simply built to arbitrage your budget. It once absorbed 15% of programmatic spend (ANA, 2023).

A good answer: an exclusion list that is actively maintained, not inherited from the DSP.

6. What’s my real frequency per person — not per ID?

Duplicate and recycled advertising IDs inflate “unique reach”, so the same person can be counted many times over. Ask for frequency against people, not identifiers.

A good answer: a deduplicated figure, with an explanation of how duplicates were resolved.

7. How do you verify geography?

Location data is routinely wrong, and spoofed geo is one of the cheapest ways to sell worthless inventory into a premium market. Ask whether targeting is confirmed by recent device activity or simply declared in the bid request.

A good answer: confirmation from observed device behaviour, not IP lookup alone.

8. What share of budget reached the target audience specifically?

Reach is meaningless if it is the wrong people. Verified humans outside your audience definition are a cleaner form of waste, but waste nonetheless.

A good answer: on-target percentage reported alongside verified reach, not instead of it.

9. Who provides the verification — you, or an independent source?

A seller grading its own delivery is not verification, and the same applies to an agency reporting on media it bought. This is the structural question behind all the others.

A good answer: a named third party with no stake in the media being measured.

10. Will you run a controlled pilot against a BAU control group?

The closing question, and the one that settles it. Identical creative, identical pricing, identical landing experience — business as usual on one side, verified delivery on the other. Willingness to be tested on your own data is the clearest signal there is.

A good answer: yes, and a suggested design for it.

What the answers tell you

Notice that none of these questions asks an agency to be perfect. They ask it to be measured. A partner who takes delivery quality seriously will have most of these numbers to hand, will be candid about the ones they don’t yet track, and will offer proof rather than reassurance.

Vagueness, defensiveness, or “the platform takes care of that” are themselves answers — and usually not the ones you were hoping for. So is a confident number with no method behind it. The useful follow-up to any figure on this list is simply “how do you know?”, and the quality of that second answer tells you more than the first.

Work through the list once a quarter rather than once a relationship. Delivery quality drifts as inventory mixes change, and an agency that answered well in January may be buying from somewhere else entirely by the autumn. The point is not to catch anyone out; it is to make verified delivery a standing item, so that the numbers are produced as routine rather than as a response to suspicion.

If you want independent answers to any of these, we’ll verify a live campaign for you.