12. What has to be true
Every chapter added lines to one list. This chapter collects them: each hypothesis with its number, its evidence or the word assumption, its owner, and the date by which it will be tested. The list is sorted so the assumption most likely to kill the business is at the top, and the top few are what must be defined this week. This is the deliverable of business design. Everything else was working.
- What is a 'what has to be true' analysis?
- How do you prioritize business assumptions to test?
- What is the final output of business design?
§The hypothesis: the business design of the first version can be written as one list of claims, each with a number, , an owner and a date, sorted so that the claim most likely to be fatal is tested first.
§Eleven chapters, one list. This is where it is collected.
§ 12.1The form#
§One row per . Five columns.
§The claim. In one sentence, falsifiable. "Sixty percent of guests shown the profile complete it unprompted." "A recognized guest visits ten percent more often across brands in a quarter." "A server will glance at the profile card before seating on half of shifts."
§The number. The threshold below which the claim is in doubt.
§The evidence. A quote and interview code, an internal data point, a benchmark, or the word "assumption."
§The owner. One name.
§The date and method. When it will be tested and how: a conversation, a concierge, a pilot, a page, a data pull.
§Roger Martin's version of the question, what would have to be true for this choice to be right, is the frame for every row: not "is this true" but "what would have to be true for the business to work," which turns beliefs into conditions and conditions into tests.
§ 12.2Sorting by what kills#
§McGrath's rule decides the order: test first the assumption that, if wrong, kills the plan. Not the easiest, not the most interesting. The fatal one. For most first versions the fatal assumptions cluster in three places: whether the first customer will pay what the sheet needs, whether the unit's cost line is what the sheet says, and whether the operation will do the one thing the model requires of it.
§Annie Duke's addition is a confidence level per row: how sure is the team, as a number. A row at ninety percent confidence with fatal consequences is watched. A row at fifty percent with fatal consequences is this week.
§ 12.3What must be defined this week#
§The top of the sorted list is a short set of rows with dates that are now. In the case, three. Whether guests would complete a profile with no reward: testable in act one's first month, owned by the lead, threshold sixty percent. Whether a recognized guest visits more: the central hypothesis, silent in the data, testable only in the pilot, owned by the product lead, with the quarter's visit data as the method. Whether a server will glance: testable in the discovery-loop conversations before act two and in the one-restaurant pilot, owned by the operations lead, threshold half of shifts.
§Everything else on the list had a date later than this week, and that was fine, because everything else had a row.
§ 12.4The list and the ladder#
§The fourth whitepaper's had goals, one number per bet. This list is longer and lower: it holds the assumptions underneath the goals, the ones about cost, capacity and payment that the goals presume. When a goal is read at ninety days, this list says which assumptions the goal's number confirms or denies. The two documents are kept together.
§ 12.5What has to be true#
§This chapter adds nothing. It collects. The list, sorted, with the top rows dated this week, is what business design produces, and Part V is about how teams manage not to produce it.
One list. Number, evidence, owner, date. Sorted by what kills the business first. The top three are this week.
- Rita McGrath and Ian MacMillan, Discovery-Driven Planning, Harvard Business Review (1995). hbr.org/1995/07/discovery-driven-planning
- A.G. Lafley and Roger Martin, Playing to Win (2013), on 'what would have to be true.' rogerlmartin.com/lets-read/playing-to-win
- Annie Duke, Thinking in Bets (2018), on beliefs held with confidence levels. www.annieduke.com/books