Managed service or in-house? An honest map of which programmatic functions belong inside

Somewhere in most programmatic reviews, a question arrives that the desk on the other side of the table would rather came later. It is always asked politely. Why shouldn't we just take this in-house?

We published that question ourselves. It is question nineteen of the twenty we told iGaming operators to put to any programmatic partner, and our printed answer runs to eight words: Sometimes you should — and we'll tell you which parts.

This is the piece that makes good on it.

Almost everything written about managed service versus a self-serve seat is written by someone selling one of the two. A network explains why the work is too complex to bring inside. A platform explains how straightforward its seat is to operate. Both are answering a question that does not exist, because managed-or-in-house is not a single decision. It is a decision per layer, and the layers move inside on different timelines.

Programmatic is a stack, and you buy it by the layer

A trading desk does what an operating system does. It manages resources, it schedules processes, and it gives every campaign a consistent way to run. Budget, inventory and attention are the resources. Analysis, optimisation and reporting are the processes. Consistency — the same quality in every market regardless of who is at the keyboard — is the thing being bought.

That stack has four layers, and the in-house question lands differently on each one.

• The kernel — data. Pixel events, first-party seeds, conversion logs, path-to-conversion exports. Everything above it inherits its quality.

• The system — the platform. The seat itself: auctions, bid logic, optimisation goals, deals and supply paths. The platform can only execute what the kernel makes possible.

• The processes — the standing routines. Audience construction, pacing, reconciliation, testing, reporting. The part nobody puts in the business case.

• The interface — your people. Whoever reviews, decides and signs off. Judgement lives here, and nothing above automates it away.

Most in-housing decisions that go badly treat the stack as a single purchase. An operator takes the seat because the seat looked like the expensive part, and inherits the process layer without having built it.

What belongs inside a mature operator within two years

Some of this costs us money to say, which is precisely why it is worth saying in public rather than in a room where the answer can be tuned to the account.

The data kernel — and not in two years. Now. Pixels, audiences and campaign learning should be yours, in your seat, documented and portable, from the first week. That is our default contract position rather than a concession to be negotiated, and where a platform's structure genuinely limits what can transfer, that limit belongs in writing before signature rather than in a conversation on the way out. A desk that treats your audience intelligence as leverage against the day you leave is telling you what it privately thinks its own work is worth.

Steady-state trading in settled markets. Once the structure, the goals and the guardrails exist, holding an always-on campaign on pace in one mature market is maintenance rather than craft. It is also a meaningful share of what managed service is usually billed for. We would rather concede it than defend it.

The routine reporting layer. A report should answer four questions: what happened, why, what we did about it, and what happens next. Once event definitions, naming and the attribution window are fixed and agreed, producing the weekly against that template is a systemised task, and systemised tasks belong to whoever is closest to the business. Reporting hours are among the easiest for any desk to bill and among the hardest to justify once the taxonomy has stopped moving.

Creative production and the rotation calendar. Usually half inside already. Finish the job. No external partner will ever know your promotional calendar, your bonus rules or your market-by-market creative restrictions better than the people who live with the licence.

Budget authority across markets. Recommendations can come from outside. The allocation decision belongs to whoever carries the P&L, and a desk that resists that is protecting something other than your performance.

What tends to stay specialist

Our published answer names seven, and we have not revised the list:

• advanced setup and troubleshooting

• market expansion

• pDOOH and PMP development

• deep Trade Desk expertise

• attribution and path-to-conversion analysis

• independent performance review

• training your team through the transition, so no learning is lost on the way

There is a pattern running through that list. Each item is either infrequent — you enter a market or negotiate a private deal a handful of times a year, never often enough to keep a full-time skill sharp — or it requires someone with no stake in the verdict. Independent review is the clearest case. An internal team grading its own quarter is being asked to do something that no honest reporting structure asks of anyone, and the cost of getting it wrong is not a bad slide. It is a set of numbers nobody reconciled until the relationship had already been decided by them. We wrote up what that looks like in The 2% Problem.

The input nobody costs properly

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.

The first three can be bought. The fourth is the one that quietly fails, because it is never on the business case. A head of acquisition who now also owns trading is a head of acquisition with less time for acquisition. The skill is learnable and the tooling is available to anyone; what is scarce is a senior person whose calendar can absorb the auction on a Tuesday afternoon in a month when three markets are live and one regulator has changed its mind. Attention is the real product of any desk, internal or external, and it is the input most likely to be assumed rather than counted.

How to compare the two numbers honestly

You cannot run this comparison against a fee you cannot see, which makes fee transparency a prerequisite for the in-house decision rather than a nicety.

What should be on the table before you sign anything: media cost, platform fees, data costs, verification, ad serving where it applies, and the desk's own fee — itemised on one page, with pass-through costs identified as pass-through. 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, the desk has failed the question. The same applies to how the desk is paid: one disclosed model, stated in the agreement, so that nothing in the partner's economics breaks when your cost per acquisition falls.

Hold the in-house case to the same standard. A self-serve seat does not remove the platform fee, the data cost or the verification line. It moves the operating cost from an invoice to a payroll line and adds the recruitment, the ramp and the cover for the week that person is away. Both costs are real. Only one of them is easy to leave out of the spreadsheet.

What we would rather be

We would rather be your specialist layer for years than your everything for one contract cycle. That is not modesty; it is the only version of this business that survives a client getting good at it.

What that layer produces is easier to show than to describe. The Seasonal Engine Lens is one output of it: twenty-one months of a single operator's measured first-deposit data, built into an instrument you can move rather than a chart you have to take on trust.

There is a test you can run on any partner, including this one. Ask them to map which of your functions should be internal in two years, and watch whether the answer costs them revenue. If every function they nominate happens to be one they do not currently supply, you have learned something useful about how the rest of their answers should be read.

The twenty questions this piece comes from are set out in the iGaming operator's guide to programmatic, and you can score any desk against them — us, your incumbent, or the internal team you are considering building — with the partner scorecard. It runs in your browser, sends nothing anywhere, and asks for no email address. Whichever way the decision goes, you end up holding evidence rather than a pitch.

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.