For a strategy module I had to analyse DAZN — the global sports streaming platform — using the standard toolkit: Porter's competitive strategies, the five forces, competitor analysis, environmental trends, a stakeholder matrix. I went in expecting to write about the product. Streaming quality, the app, pricing tiers, the emotional pull of watching live with millions of other people. That is what you see as a user, and it is what the company markets. The five forces analysis kept pulling me somewhere else. The force that decided everything Supplier bargaining power: high. Not "somewhat high" — the highest-leverage finding in the whole analysis. Broadcasting rights are concentrated in a very small number of organisations. Leagues and federations own the thing customers are actually paying for, and they auction it on their own schedule, to whoever bids most, for a fixed term. Now put that next to customer bargaining power, which came out medium-high: monthly subscriptions with no lock-in, several substitutes, and viewers who follow a sport rather than a platform. That is a squeeze from both ends. Costs are set by suppliers with structural leverage. Revenue depends on customers who can leave the month the rights they cared about move elsewhere. Every part of the product I had planned to write about — the interface, the streaming stack, the recommendation logic — is table stakes. Necessary, entirely insufficient, and not where the competitive position is decided. The position is decided upstream, in a negotiation the user never sees. Why I have kept thinking about it Because once you see that structure you start finding it everywhere, and the tell is always the same: the thing everybody benchmarks is not the thing that binds. In AI, the conversation is about models. Which one scores better, which one is cheaper per token, which one handles longer context. Those are real questions and they are the equivalent of comparing streaming quality. The forces that actually constrain an AI company sit upstream of all of it. Access to compute, and what it costs in energy. Rights to the data you trained on, and whether that permission survives a change of licence. Dependence on a handful of providers who set your input prices, on their timetable, and who are also — occasionally — your competitors. A team can win every benchmark it enters and still have no defensible position, because the benchmark measures the part of the stack where leverage is weakest. The question the framework actually asks The five forces are taught as a template to fill in, and that is how I first used it: five boxes, five paragraphs, a mark. Filling in boxes is not analysis. The analysis is noticing that one of the five is doing all the work and the other four are context. For DAZN, one force explained the strategy: exclusive content, local market focus, revenue diversification. All of it is a response to supplier power. Read the strategy without that force and it looks like a series of product decisions. Read it with the force and it looks like what it is — a company organising itself around the one thing it does not control. So the question I take into anything I evaluate now is not "how good is this?". It is: which of the constraints on this business is the binding one, and is the thing being measured anywhere near it? Most of the time the answer is no. That is not because people are careless. It is because the binding constraint is usually upstream, unglamorous and hard to measure, and the thing that is easy to measure is right there, with a leaderboard next to it.