Most engagements fail at the question, not the analysis. A team commits months to answering something that was never framed to be answerable, and no amount of good work rescues a bad question. The Engagement Question-Framing Diagnostic pressure-tests the named question before scope is set, checking whether it is singular, decidable, owned, and time-bound. Run it in an hour with the sponsor in the room, and you either sharpen the question into one worth answering or you surface that there is no real question yet, which is the cheapest finding you will ever buy.
The question is the riskiest artifact in an engagement
Every engagement rests on a question. It might be phrased as a goal, a mandate, or a problem statement, but underneath it is a question the work exists to answer. When that question is well framed, ordinary analysis produces a usable answer. When it is badly framed, brilliant analysis produces an answer nobody can act on, because the question was never shaped to yield a decision. The most expensive failures in professional work are not analytical failures. They are framing failures that everyone agreed to before the meter started running.
The Engagement Question-Framing Diagnostic exists to catch that failure while it is still cheap to fix. It runs before scope is set, before the plan is built, before anyone commits time. It takes the named question of the engagement and tests it against a small set of properties that separate answerable questions from unanswerable ones. The diagnostic does not answer the question. It decides whether the question is worth answering, and if it is not yet, it shows exactly what is missing. Buying that finding in an hour is one of the highest-return moves available to anyone about to commit a team.
The reason this works is that bad framing hides in plain sight. A question like "how do we grow faster" sounds legitimate and passes unchallenged in most kickoff meetings. The diagnostic forces the room to notice that it names no single decision, sets no bar for what counts as an answer, and belongs to no one in particular. Naming those gaps out loud, before scope, is what the tool is for.
The four tests a question must pass
The diagnostic runs the named question through four properties. A question that fails any one of them is not yet ready to fund. The value is not a pass or fail score; it is the specific rewrite each failed test demands. The table below shows the tests, what each asks, and the failure signal that tells you the test was failed.
| Test | What it asks | Failure signal |
|---|---|---|
| Singular | Is this one question, or several bundled together? | The word "and" joins two decisions, or answering half still leaves you stuck. |
| Decidable | Could the answer come out either way and change what we do? | Every plausible answer leads to the same action, so the question is rhetorical. |
| Owned | Is there a named person who will act on the answer? | No one in the room can say "I will decide based on this." |
| Bounded | Is the answer testable against a stated bar and a date? | There is no definition of "good enough" and no deadline the answer serves. |
| Grounded | Do we know what evidence would settle it? | No one can name the data or observation that would change their mind. |
Consider a worked example. A COO brings the question "should we restructure the commercial organization?" to a kickoff. Run through the tests, it fails three. It is not singular: restructuring bundles a design question with a sequencing question with a change-management question. It is not bounded: there is no bar for what restructuring must achieve, so no answer can be judged right. And it is only half owned: the COO owns the design, but the CEO owns the go decision, and she is not in the room. The diagnostic rewrites it into a fundable question: "by what date and against what revenue-per-rep target should the CEO decide whether to consolidate the two regional sales teams into one?" That version is singular, decidable, owned by a named person, bounded by a metric and a date, and grounded in data the team can actually gather. The engagement can now be scoped, because there is finally a real question underneath it.
Running the diagnostic before scope
- Get the actual decision owner in the room, not a proxy. The test for "owned" cannot be run by someone who will not make the call, and framing done without them tends to get overturned later.
- Write the named question on the board verbatim as the sponsor states it. Test the real words, not a cleaned-up version you improved on the way to the whiteboard.
- Run all five tests before rewriting anything. Rewriting after the first failure hides the other gaps and produces a question that fixes one flaw while keeping the rest.
- For each failed test, capture the specific rewrite it demands, not just the fact of failure. The output of the diagnostic is a sharpened question, not a verdict.
- Refuse to scope until the question passes all five. A failed question is a stop, and stopping here is far cheaper than stopping three months in.
Where question framing goes wrong
- Testing a proxy instead of the decision owner. A framing that the real owner never endorsed unravels the moment they see the answer. Fix it by refusing to run the "owned" test without the person who will act.
- Improving the question before testing it. A tidied-up version passes tests the original would fail, hiding the real gap. Fix it by testing the sponsor's exact words first.
- Accepting a bundled question because each part sounds reasonable. Two decidable questions joined by "and" still cannot be answered as one. Fix it by splitting on the conjunction and scoping the piece that matters most.
- Treating "decidable" as satisfied by any answer. If every answer leads to the same action, the question is rhetorical and the engagement is theater. Fix it by asking what you would do differently for each possible answer.
- Skipping the bar and the date. Without a stated "good enough" and a deadline, the answer can never be judged complete. Fix it by forcing the sponsor to name both before scope.
Before scope is committed
- The named question is written verbatim and tested as stated.
- The real decision owner confirmed they will act on the answer.
- The question is singular, with any bundle split apart.
- A bar for "good enough" and a decision date are both named.
- The evidence that would settle the question is identified.