What would a truly transparent ad platform look like?

"Transparency" appears on every ad platform's about page and means something different on each. So here's a concrete ad platform transparency test — five questions a buyer should be able to answer — and an honest scoring of the industry against it, ourselves included. (Disclosure: published by BidSurvivor; we built the board to pass this test, and we say below where passing is easy for us.)

The five questions

1. Can you see the price before you commit? Not an estimate — the actual clearing price of the thing you're buying. The ad rate comparison lists published rates across several channels.
2. Can you see what you got? Impressions and clicks, from the platform, at the placement level.
3. Can you see the fees? What fraction of your dollar bought media versus intermediaries. Industry studies of programmatic supply chains have repeatedly found large fractions — sometimes approaching half — absorbed before reaching the publisher.
4. Can you see where your ad ran? Every URL, not a summary.
5. Can you see everyone else's results? The question almost nobody passes — and the only real answer to "is this platform worth it before I spend".

Question 3, with the number attached

The "approaching half" above is not rhetoric. The most cited study of where a programmatic pound goes is the one ISBA commissioned from PwC in 2020, which followed a group of UK advertisers' money through the open programmatic chain and found that publishers received roughly half of what the advertisers spent. Around 15% of the spend could not be attributed to any party at all — an "unknown delta" the study could not explain. The US advertisers' association ran a similar exercise in 2023 and reached a similar shape of answer. When the buyer, the seller and two trade bodies between them cannot say where a seventh of the money went, question 3 is not being answered by anyone.

The reason this matters for a small budget is that the fee is invisible precisely where it is largest. A $200 programmatic spend does not show you $100 of media and $100 of chain; it shows you a CPM.

How the industry scores

Search and social platforms pass 2 partially (your own dashboard), fail 1 (auction prices are reconstructible never), pass 3 nominally (their fee is the spread, undisclosed), pass 4 variably, fail 5 completely — you see benchmarks, published by the seller. Programmatic fails 3 famously and 4 chronically. Newsletter sponsorships pass 1 and 4 trivially (one price, one placement) and fail 2 more often than they should. EthicalAds deserves its flowers: published network CPMs pass half the test honestly.

The 2026 wrinkle is that the biggest platform is now automating the parts a buyer used to be able to see. An AI-managed search campaign reports what it chooses to report, at the level it chooses; the comparison with a placement that simply tells you its opens and clicks is in AI Max vs fixed price: which one tells you what happened?.

Our answer sheet — and why it's easy for us

BidSurvivor passes all five: clearing prices public, every brand's opens and click-throughs published, no intermediary fees (the bid is the whole price, $0 empty / $1.00 floor / $5,000 cap), one placement (the board itself), and — question 5 — everyone's numbers visible before anyone bids. Full honesty about why: passing is structurally easy at two slots a day, and the same transparency that proves the board works also publishes how small it currently is. We're not more virtuous than Google; we're small enough that the doors fit the building. The interesting question is why platforms whose numbers are good keep them private — we have a theory.

Where our answer sheet is weakest

Passing five questions does not make a platform good, and it is worth saying which questions the test doesn't ask.

It doesn't ask about targeting, because we have none. A slot shows one brand to whoever is looking; a search ad shows one brand to somebody who just typed what it sells. On question 5 we win; on "who saw it" search wins outright, and for most direct-response jobs that is the question that matters more.

It doesn't ask what a number means. We count "opens" (somebody opened a brand's card on the board) and "click-throughs" (somebody left for the brand's site) separately, and we only show an open-to-click rate once a slot has had at least twenty opens, because a rate built on three opens is a coin toss with a percentage sign. Traffic we can identify as bought or synthetic is flagged at source and never counted in either. All of that is a choice, and a different choice would produce different-looking public numbers from the same visitors — which is exactly why the definitions have to be published alongside the figures.

It doesn't ask what you get to keep. A placement that ends should hand you something. Ours hands you a link with rel="sponsored" — the attribute Google's documentation on qualifying outbound links says paid links should carry — plus a brand page that lists what every slot you held sent. Whether that is worth more or less than a "dofollow" link from a marketplace that ignores the guidance is argued in sponsored, nofollow, dofollow: what a paid placement should give you.

Why question 5 is the tiebreaker

Questions 1 to 4 are about your own purchase. Question 5 is about whether the platform's evidence survives contact with everyone else's experience, and it is the only one that protects you from the oldest trick in advertising: the case study. Every platform can show you a winner. A platform that shows you all of its buyers is showing you the losers too, and the ratio between the two is the number you would actually pay to know. The mechanism by which the losers vanish from every other platform's evidence is in every ad platform shows you the winners.

Using the test

Run the five questions over any platform pitching you. Every "no" is a place where someone else's judgement replaces yours, priced into your invoice. Buy from the platform with the fewest no's that still reaches your audience — and treat question 5 as the tiebreaker, because a platform confident in everyone's results has nothing to hide in yours. The full rules of the one platform we can vouch for are in how BidSurvivor works, including the two rules people find least comfortable: a lost slot is not refunded, and the numbers are published whether they flatter us or not.