Marketing frameworks: an assumption register to steer pilots

What to take away

What the register holds

A register lists what must be true for a plan to work. It turns a framework from a workshop output into an input for pilot design and staged funding. A plan that hides its assumptions cannot say what a pilot is meant to prove.

Every familiar model carries such assumptions: SWOT, Porter's Five Forces, the Business Model Canvas, the 4Ps, jobs-to-be-done, AARRR and OKRs. The register is a marketing framework example teams already keep in some form. What changes is how directly it feeds the decision on what to test and what to fund.

A worked case, in words

Take a team planning to offer a paid add-on to existing customers. The plan rests on three assumptions: customers want the add-on, the email list reaches them, and support can absorb the extra load. No figures appear because this is an illustration, not a case study.

The first assumption decides whether the offer exists at all, so it ranks highest on the value of the decision. The third is cheap to check against the support rota. The second is the slowest to answer. The team pilots the first on a small audience and writes the stop rule before the test starts.

If the pilot shows no effect, the result goes into the register as it stands. The offer does not return next quarter under a new name with the same untested claim behind it. The other two assumptions stay open and can be tested later, if the plan survives.

Three ways to rank what gets tested

Rank each assumption on three bases. Each asks a different question of the same register, and an assumption that scores well on all three is the obvious place to start.

GOV.UK's alpha-phase guidance makes the same point for services: it tells teams to focus on testing the riskiest assumptions. Marketing teams can borrow the habit without copying the government process.

Funding follows the evidence

Staged funding releases money as assumptions are retired, not as a single annual block. Each stage has a question, a bounded test, an owner and a stop rule.

A rolling plan in a marketing planning process can then change as the evidence comes in. A small first audience keeps an early mistake cheap, and a stop rule tells the team when to end a test instead of extending it. Name one owner for the register.

Neutral results stay on the record

A pilot that shows no effect is still evidence, and the register is where it lives. Without it, a rejected idea returns in the next planning round under a new name.

Pair the register with forecast tracking. Forecast error and data quality sit beside campaign performance, and both help decide whether an assumption has really been retired or only stopped being discussed.

AI enters the register

AI can summarize research, classify evidence, generate options and identify inconsistencies. It can also introduce unsupported output, protected-data exposure, bias and automation error. An assumption an AI tool proposes is still an assumption.

NIST's AI Risk Management Framework FAQ describes AI RMF 1.0 as a living, voluntary document and notes that the 2025 U.S. AI Action Plan tasked NIST with revising it. Record the guidance version, the use case, the inputs and the review route, which is already a line on the marketing frameworks checklist.

Prepare without chasing fashion

Audit current frameworks, list the assumptions each one rests on, and test one live decision. Adopt a new practice only when it handles a defined uncertainty better than the existing decision record.

Let register upkeep share the campaign calendar rather than pause it. Finance, legal and privacy review only the inputs they own, and each assumption keeps one named owner.

Common questions