A commercial model is usually presented as a set of numbers: target margin by product, promotional intensity by segment, expected bonus cost, contribution by market. It reads like an economic plan.
It is not. It is a set of promises about decisions somebody will have to make every day — what to price at, when to lay off, which customers to restrict, how much to spend to reacquire a lapsed player. If nobody owns those decisions, or if two people own them with different targets, the plan does not survive its first quarter.
This is why commercial models and operating models fail together, and why designing them separately is such a reliably expensive mistake.
The classic split
The most common structural failure in a sportsbook is margin ownership sitting in two places.
Trading owns theoretical margin. Marketing or CRM owns promotional spend. Both have targets, and the targets are in tension by construction: every free bet, price boost and cash-back offer moves the effective margin that trading is accountable for, and none of them are trading's decision.
What happens next is predictable. Trading widens prices to protect the number. Marketing responds to the resulting competitiveness gap with more promotion. The book gets simultaneously less competitive and more expensive to run, and both functions can demonstrate they hit their targets.
Nobody in this loop is doing anything wrong. The model is wrong.
Three questions that expose the design
You can usually diagnose an operating model in a single session with three questions.
Who owns net margin after promotional cost? If the answer is "we look at it together in the monthly", nobody owns it. This is the most common answer and the most expensive one.
Who can suspend or restrict a customer, and who finds out? Restriction policy sits at the intersection of trading, compliance and commercial. When trading acts alone, marketing keeps spending to reacquire customers who will be restricted again. When commercial blocks it, liability accumulates.
Who decides what gets built next quarter? In many groups the honest answer is the platform supplier's roadmap. That is a viable model, but it should be a deliberate choice rather than a discovery.
Multi-OpCo groups compound all of this
In a group running several operating companies, each with local commercial autonomy, the failure modes multiply. Each OpCo optimises its own P&L. Shared technology has to serve conflicting requirements. Central functions accumulate coordination cost without decision rights.
The instinct is to centralise, and the instinct is usually half right. What genuinely benefits from centralisation is narrow: pricing and risk, platform and data, and brand-level positioning. What genuinely benefits from local ownership is also narrow: payments, regulatory relationships, customer service and local marketing.
The expensive mistake is centralising the second list because it is easier to see on an org chart than the first.
Designing them in one pass
A commercial model and an operating model should be produced by the same exercise, and it is not a long one. In practice it means answering, together:
- Where does margin come from, by product and segment, and who is accountable for it after promotional cost?
- Which capabilities are central, which are local, and what is the decision right in each case?
- What does the trading operation need to be able to do, and what is automated versus manual?
- What is the promotional envelope, who sets it, and how is it reconciled against trading margin?
- What does the technology have to support for all of the above to be possible?
The output is not a large document. It is an accountability map plus a small number of decisions that have actually been made rather than deferred.
Why this rarely happens
Because the two pieces sit with different people. Commercial strategy is owned by a chief commercial officer or a strategy function. Operating model design lands with a transformation office or HR, often a quarter later, once the strategy is already approved.
By then the strategy is a constraint rather than an input, and the org design is built to accommodate the existing structure rather than the plan. The plan quietly loses.
The fix is procedural rather than intellectual: do not approve a commercial model that has not named who owns each number in it. That single requirement surfaces most of the structural problems before they become expensive.