How ThumbAd Answers the 20 Questions — Our Due-Diligence Answers, On the Record

A trading desk publishing its own exam answers is a slightly odd thing to read, so it is worth saying plainly why this exists. Last month we published a due-diligence guide telling iGaming operators exactly what to demand from a programmatic partner — twenty questions, and how to score the answers. Publishing a standard and then declining to be measured by it would have been the most revealing thing we could possibly have done.

So here is the whole exam, answered. Every question, our short answer, what it means in practice, and the line telling you how to hold us to it. If you have already run the partner scorecard, this page is where its closing instruction — now run the same twenty on us — is supposed to land. Score us against what is printed below. Nothing here is written to be admired; it is written to be checked.

This is open, and that is a deliberate change

The answer key was originally set to sit behind a form, like the guide it accompanies. That was wrong, and we have changed it. Gating our own answers to our own published exam contradicts the entire premise of publishing them — you cannot ask an industry to be more transparent and then charge an email address for your own transparency.

So: this document is open. No form, no gate, no email. The guide that sets the standard stays gated, because it is a thirty-page instrument built for operators who want it and a fair trade for a contact detail. Our answers to it are not. Read them, quote them, print them, use them on your incumbent.

Three ways to use this

Read it before meeting us, and the first meeting starts three meetings deep. Bring it to the meeting and score us live against what is printed — we will bring the artifacts each answer references. Or use it against your current partner: every answer here is a benchmark for what the same question should produce elsewhere.

Five answers commit us to bringing something physical to the room rather than a claim:

• Q07 — yes to raw path data. We bring an anonymised path-to-conversion export.

• Q10 — yes, we cut good-looking channels. We bring the specific story: client, channel, data, aftermath.

• Q14 — every fee, line by line. We bring a sample fee sheet you can keep.

• Q18 — decisions, not metrics dumps. We bring real weekly, monthly and quarterly examples.

• Q20 — a lost account, honestly. We bring the account, the reason, the process change.

One reading note: where an answer depends on a specific contract, platform structure or market, we say so in the answer. Certainty we don't have is not something we print — that discipline is also part of the answer.

Questions 01–06 · Data and audiences

Q01 — Who owns the pixel and the audiences you build?

The short answer: You do. In your seat, documented, portable.

Client ownership of pixels, audiences and campaign learning is our default contract position, not a concession you have to negotiate. Segments are built in your advertiser structure, named to a documented taxonomy, and written up so they make sense without us in the room.

Where a platform's seat structure or privacy rules limit what can physically transfer, we put that in writing before signing — you'll know exactly what leaves with you and what is platform-bound. What we will never do is treat your audience intelligence as leverage against the day you want to leave.

Hold us to it: ask for the audience taxonomy document in month one. If it doesn't exist, this answer was theatre.

Q02 — Show me an audience you built from behaviour, not demographics.

The short answer: Here's one: started registration, never finished.

That single behaviour outperforms any demographic label, because it's a person telling you exactly where they stopped. Our iGaming segments are built from events like it:

• Registration started, not completed

• Registered, never deposited

• Lapsed depositors flagged for reactivation from CRM

• High-intent visitors by funnel depth, frequency and recency

• Product- and market-page behaviour, not "interest in sports"

• Exposed on CTV or DOOH, reachable again on mobile and display

"Males 25–44, sports interest" is a census category, not an audience. We build from what people did.

Hold us to it: pick any live segment in your account and ask us to show the behaviours that define it.

Q03 — How do you prove a new audience beats what it replaces?

The short answer: A named test, designed in writing before launch.

Every challenger audience gets a test card before it spends: the incumbent it must beat, the success metric, the budget split, the runtime, the minimum conversion volume, the attribution window, the exclusion logic, and the decision rule — scale, iterate or kill.

The metric matches your economics, not our convenience: FTD or deposit value where backend data is connected, and where it isn't yet, results are labelled platform-attributed — never dressed up as confirmed customers.

Hold us to it: ask to see the test card before the test, and the verdict after. Both should exist as documents.

Q04 — How do you handle consent and cookieless markets?

The short answer: Segment by segment, market by market — with consent status documented.

"The DSP handles it" is not an answer, so we don't give it. Every segment carries its legal basis and consent status in its documentation, and addressability is planned per market.

• First-party and consented CRM data where the basis is clean

• Contextual and inventory-based targeting where identity is thin

• UID2 / EUID and publisher audiences where supported and lawful

• Measurement expectations adjusted honestly — not pretended around

In restricted markets we change the strategy and say so, rather than running the same tactics and hoping the reporting hides it.

Hold us to it: ask for the consent status of any segment we run. The answer should take minutes, not a legal review.

Q05 — What happens to my data if we part ways?

The short answer: It leaves with you — and the map of what transfers is agreed before we start.

Setup, naming conventions, audience definitions, reporting logic and campaign learning are documented from day one, which is what makes them portable. Before signing, we walk through what is contractually and technically transferable versus what any platform keeps regardless of partner.

No hidden lock-in dressed up as partnership. A desk that needs captivity to keep clients is telling you what its work is worth.

Hold us to it: ask for the offboarding clause before you need it. Ours is written to be read.

Q06 — How is my data separated from your other clients?

The short answer: Client-scoped structures. No cross-account mixing, ever.

Your data lives in your campaigns, your advertiser structures and your projects — separated at the account level of every platform we touch, from The Trade Desk to measurement and reporting tools.

In a regulated category like iGaming this is more than hygiene: a segment that has quietly absorbed another brand's data is a compliance incident waiting to be discovered. Client-specific data stays client-specific, without exceptions for convenience.

Hold us to it: ask us to diagram your account structure and where its boundaries sit. It's a one-page answer.

Questions 07–13 · Measurement and proof

Q07 — Can I see raw path-to-conversion data for my campaigns?

The short answer: Yes — and we bring an example export to the first meeting.

Path-to-conversion analysis is core to how we work, not a special request: full journey sequences with every touch, channel, creative and device, first and last exposure, touch counts, time to conversion, and view-through versus click-through roles.

Where privacy rules or platform policy limit granularity in a market, the export says so in its notes — the limitation is disclosed, not smoothed over. You will never be asked to accept screenshots as data.

Hold us to it: ask for the sample export in the sales process and time the response.

Q08 — Show me last-touch and multi-touch side by side for one campaign.

The short answer: Side by side in every review — with the disagreement explained, not hidden.

The two models answer different questions, so of course they disagree: last-touch flatters retargeting and lower-funnel display; multi-touch reveals what CTV, video and prospecting actually contributed upstream. The interesting part is the gap between them — that's where budget decisions live.

We will never tell you multi-touch is "too complex." We will tell you precisely what each model can and cannot prove, and reconcile both against your backend where it's connected.

Hold us to it: in any QBR, point at a channel and ask what each model says about it. Expect two numbers and a reason.

Q09 — How do you reconcile platform FTDs with my backend?

The short answer: Monthly, in writing, with every gap explained.

The reconciliation checks the boring things that create the gaps: event definitions, attribution windows, time zones, currencies, deduplication, naming and UTM logic — then compares platform-attributed FTDs against your backend-confirmed customers, line by line.

Until your backend is connected, platform FTDs are labelled platform-attributed in every report we send. Numbers that have never been checked don't get presented as facts.

Hold us to it: ask for last month's reconciliation note. If a desk has never written one, their numbers have never been audited.

Q10 — When did you last recommend cutting a channel that looked good?

The short answer: We cut channels that look good — when the path data shows the credit is borrowed.

A channel earns a cut when it converts only on last-touch, when its players don't deposit twice, when frequency has passed the point of persuasion, or when pause tests show the conversions would have happened anyway. We run those pause tests and spend reviews as standing practice.

The specific story — which client, which channel, what the data showed, what happened after the cut — we tell in person. Our clients' accounts are real businesses, not anonymised marketing material. Ask, and you'll get names, numbers and the follow-up.

Hold us to it: ask the question in the room. A desk that has never recommended spending less is selling media, not managing it.

Q11 — How would you design an incrementality test on my account?

The short answer: Designed to your volume — or we tell you you're not ready yet.

The design follows your market count, conversion volume and budget: geo splits and matched markets where you have several territories, audience holdouts where you don't, pause-reactivation tests where budget discipline demands it. Every design is written before launch: hypothesis, test and control, KPI, runtime, minimum volume, exclusions and the decision rule.

If your account can't yet support a statistically defensible test, the honest recommendation is a measurement-readiness phase and a clearly-labelled directional test — not a fake experiment that flatters everyone.

Hold us to it: ask what minimum volume the test needs. A desk with a real methodology answers with a number.

Q12 — What's your attribution window, and who can change it?

The short answer: It's in the contract. Changing it requires your sign-off.

The window is agreed per objective before launch — tighter for lower-funnel registration and FTD activity, longer where CTV and DOOH influence is genuinely part of the journey — and it's documented where both sides can see it.

It does not move after results are visible. Window-shifting is the oldest trick in performance reporting, which is exactly why the change control belongs to you, not to us.

Hold us to it: check your current contract for the window. If it isn't written down, it's already movable.

Q13 — Which of my KPIs — NRC, FTD, RDC — do you actually optimise toward?

The short answer: The one your economics point to — usually FTD, with RDC as the quality check.

Cheap registrations that never deposit are a cost, not a result, so we don't stop at them. The working hierarchy: qualified exposure up top, registration started and completed mid-funnel, FTD as the primary optimisation event, and repeat deposit as the standing check that the players we buy are worth having.

Where FTD or RDC signals aren't yet available in-platform, we build the plumbing — postbacks, offline conversion imports, CRM reconciliation — rather than optimising to whatever happened to be measurable.

Hold us to it: ask which platform event the algorithm is literally optimising toward today. The answer should match your P&L.

Questions 14–20 · Money, operations and fit

Q14 — Break down every fee between my budget and the publisher.

The short answer: Line by line, on one page — shown before you sign.

Media cost, platform fees, data costs, verification, ad serving where it applies, and ThumbAd's fee: itemised, with pass-through costs identified as pass-through. The structure varies by channel and buying method; the transparency doesn't.

You should be able to trace every euro from your budget to a publisher and account for the difference. If any line of that trace requires trust instead of arithmetic, we've failed the question.

Hold us to it: ask for the fee sheet in the first commercial conversation, not the last.

Q15 — Do you take rebates or arbitrage media?

The short answer: One disclosed model, agreed in writing — you always know exactly how we're paid.

We operate as a transparent managed service: our compensation is stated in the agreement, and whether media is bought at cost or under another agreed structure is written where you can point at it. No compensation that depends on you not reading the contract.

The test of any model is whether our incentive bends away from your efficiency. Ours is built not to: nothing at ThumbAd breaks when your CPA falls.

Hold us to it: ask us to state, in one sentence in the contract, every way ThumbAd is compensated on your account.

Q16 — Who actually trades my account, and how senior are they?

The short answer: Named people, introduced before you sign — and you can ask their account load.

ThumbAd is a senior-led desk: the people in your kickoff are the people in your account. You'll know by name who owns strategy, trading, reporting and escalation, and complex work — multi-market iGaming, pDOOH, attribution — gets senior hands, not junior handoffs.

Ask the account-load question directly and we will answer it: the current number of accounts carried by the specific traders who would be on yours, given in the meeting rather than printed here, because the honest figure moves as the book moves and a number in a brochure would be stale the week after we wrote it. Attention is the real product, and a desk that won't give you the number is telling you it's a bad one.

Hold us to it: meet the actual trader before signing. "The team" is not a person.

Q17 — How do you handle responsible-gambling and licensing rules in my markets?

The short answer: We ask about your licences before we show you a single reach number.

Market context comes first: licensed territories, geo and age restrictions, responsible-gambling messaging, self-exclusion obligations, bonus and creative rules, platform and app-store policies. All of it shapes the plan before activation — it is planning input, not post-launch friction.

We do not learn on your licence. In this category, a compliance mistake doesn't cost budget; it costs the thing the whole business stands on.

Hold us to it: notice which comes first in our pitch — questions about your licence conditions, or slides about reach. The order is the tell.

Q18 — What does your weekly and quarterly reporting actually contain?

The short answer: Decisions, not metrics dumps.

Every report answers four questions: what happened, why, what we did about it, and what happens next. Weekly: performance versus target by market, delivery and pacing, what changed in the account and what we changed in response. Monthly adds the full funnel, backend reconciliation and cohort maturation. Quarterly adds path-to-conversion findings, test verdicts, one honest miss, and the plan.

One consistent visual system across all of it, so a number means the same thing in the weekly note and the board deck.

Hold us to it: ask for a real (anonymised) example of each cadence before signing. We'll send all three.

Q19 — Why shouldn't I just take this in-house?

The short answer: Sometimes you should — and we'll tell you which parts.

In-housing works when you have the scale, the platform access, the technical support and — hardest of all — the sustained senior attention to run it well. When you do, the honest move is to say so, and we do: several capabilities belong inside a mature operator.

What tends to stay specialist: advanced setup and troubleshooting, market expansion, pDOOH and PMP development, deep Trade Desk expertise, attribution and path-to-conversion analysis, independent performance review, and training your team during the transition so no learning is lost. We'd rather be your specialist layer for years than your everything for one contract cycle.

Hold us to it: ask us to map which of your functions should be internal in two years. Watch whether the answer costs us revenue. We have since published that map in full — including the functions it costs us money to concede — in an honest map of which programmatic functions belong inside.

Q20 — Walk me through an account you lost and why.

The short answer: Yes — a real one, with what we changed afterwards.

Every desk that has traded for over a decade has lost accounts: to in-housing, to budget cuts, to market exits, to measurement disagreements, to procurement. Pretending otherwise is the least trustworthy sentence in any pitch.

We tell the story in person — a specific account, what happened, what we could have done differently, and the operating change it produced — because real losses involve real clients, and we don't anonymise them into content. What we'll commit to in print: the story exists, it's specific, and it ends with a process improvement you can still see in how we run accounts today.

A partner who never reports a miss isn't miss-free. They're hiding them — and the guide you're holding lists that as red flag number ten. The anonymised version of what unreconciled numbers do to a relationship is written up in The 2% Problem.

Hold us to it: ask in the first meeting. The quality of this one answer predicts the honesty of every future report.

Now score us

The companion guide ends with a six-criteria scorecard and a 90-day evaluation plan. We're inviting you to run both — on us. Score each criterion 1–5 with evidence attached; anything scored on impression defaults to a 3. We consider 24 of 30 the bar worth scaling with, and we expect to be held to it.

• Measurement honesty — path exports on request, models compared openly, misses reported unprompted. Evidence: Q07, Q08, Q09, Q20.

• Audience craft — behavioural segments you own, tested against controls, documented. Evidence: Q01, Q02, Q03.

• Market fluency — licences, payment rails and market rules asked about before answered about. Evidence: Q04, Q17.

• Fee transparency — every euro accounted for, in writing, before signature. Evidence: Q14, Q15.

• Communication quality — named senior people; reports that explain decisions. Evidence: Q16, Q18.

• Compounding evidence — documented taxonomy, versioned learning, honest verdicts that persist. Evidence: Q03, Q05, Q11.

And run the same scorecard on your incumbent in parallel. Whatever you decide, ninety days from now you'll hold two evidence files instead of two pitches — which was the entire point of writing any of this down.

Run the scorecard here. It scores all twenty questions in your browser, sends nothing anywhere, and asks for no email address.

Take it with you

Twenty answers. Zero footnotes hiding. Everything a desk hopes you won't ask, answered in writing before you asked it. The meeting version comes with exports, fee sheets and one uncomfortable story told properly.

Every answer above is the complete text — nothing is held back for a download. The designed 24-page edition is here too: How ThumbAd Answers the 20 Questions (PDF) — no form, no email.

ThumbAd is an independent programmatic trading desk in Oslo, trading on The Trade Desk across Europe, Africa, North America and Latin America since 2012, with betting and gaming at the core.