A pilot in a new market produced a number that should have stopped the rollout. Forty five minutes per account in the new country against fifteen in the home market, three times slower on the same task, with the same tooling and the same training.
The response from the client was four words. The decision is already made.
That response was not stubbornness and it was not a failure of analysis. The commitment had been made months earlier: budget allocated, dates announced, a country lead hired, contracts signed with local providers. The finding was accurate and it arrived outside the window in which it could have changed anything.
Decisions have a window, and it closes early
Operational findings have value in proportion to what remains changeable. That makes the useful question about discovery a question about timing rather than about method.
Four phases, and the window closes at the end of the second.
| Phase | What is still changeable | What a finding can do |
|---|---|---|
| Framing | Whether to do it at all, and what for | Change the direction, or stop it |
| Option design | Which approach, scope, sequence and market order | Change the shape of the plan |
| Commitment | Budget, dates, hires, contracts, announcements | Adjust execution within a fixed plan |
| Execution | How the work gets done day to day | Improve delivery of the chosen path |
Most discovery gets commissioned in phase three. The budget exists by then, which is precisely why it exists by then: a diagnosis is easier to fund once there is a programme to attach it to. The programme is also the thing that has fixed the direction.
The result is a predictable pattern. A finding that would have changed the plan in phase two arrives in phase three and gets reclassified as an implementation risk. Nobody is being dishonest. The finding is genuinely no longer actionable as a direction change, because the direction has acquired dates, people and public statements.
Why is the window so much earlier than people expect?
Three mechanisms close it, and they close it before the formal approval.
Announcement precedes approval
Direction is frequently communicated internally before it is formally signed off. Once a market entry, a migration or a consolidation has been described to a leadership team as happening, the cost of reversal becomes reputational on top of financial. The window closes at the announcement rather than at the signature.
Hiring locks the shape
A country lead, a programme director or a specialist team is hired against a specific plan. Those commitments are harder to unwind than a budget line, because they involve individuals who relocated or resigned from elsewhere to take the role.
The pilot becomes evidence for the decision rather than a test of it
A pilot run after commitment is read as a delivery exercise. Its purpose shifts from answering whether this works to demonstrating that it can be made to work. The forty five minute result enters that frame as a training problem.
Programme management guidance treats this sequencing as structural. ISO 21502, which covers practices for project and programme management, places the definition of a business case and the assessment of feasibility before authorization, because the authorization is what converts options into commitments. The NIST AI Risk Management Framework places its Map function ahead of Measure and Manage for the same reason: context established after a system is in flight describes something that can no longer be redirected.
How do you know the window is closing?
The signals are observable and they appear earlier than the formal milestones.
| Signal | What it indicates | Window status |
|---|---|---|
| The question asked is "how" rather than "whether" | Direction already settled internally | Narrowing |
| A date has been communicated outside the team | Reversal now carries reputational cost | Nearly closed |
| A role has been posted or filled against the plan | Individual commitments exist | Nearly closed |
| A vendor contract is in legal review | Financial and contractual lock-in | Closed for direction |
| The pilot is described as a proof of concept | Pilot is demonstrating, not testing | Closed |
| Findings are being requested as "risks to manage" | Direction fixed, execution open | Closed for direction |
The first row is the cheapest early warning available and it costs nothing to listen for. A sponsor who asks how to make a market entry work has already decided to enter the market. A sponsor who asks which market to enter first is still in phase two.
Decision window
What a finding can change, by the phase it arrives in
| Arrives in | Changes direction | Changes scope | Changes execution |
|---|---|---|---|
| Framing | Yes | Yes | Yes |
| Option design | Yes | Yes | Yes |
| Commitment | No | Rarely | Yes |
| Execution | No | No | Yes |
The two columns on the left are where discovery pays for itself. The right-hand column is available at any time, and it is the smallest of the three returns.
What the forty five minute finding was worth, in each phase
The same number, four different values.
In framing, it is a market selection input. Three times the handling cost in this market against the home market, before any of the fixed costs of entry, changes which market goes first.
In option design, it changes the shape of the entry. A phased entry, a different operating model, or a local partner handling the slow step are all available, and each is a decision about the plan rather than a change to it.
In commitment, it becomes a staffing number. Three times the handling time means three times the headcount for the same volume, which is a budget variance to absorb rather than a reason to reconsider.
In execution, it becomes a performance target. The local team is now behind on a metric set by the home market, and the finding has turned into pressure on the people who inherited the condition.
The number did not change. What was still open around it did.
What to do when the finding arrives late
Late findings are not worthless and they are worth something different from what they would have been worth earlier. Four uses, in descending order of value.
Reset the baseline before the programme is measured. If the new market structurally takes longer, the home market number is the wrong benchmark. Documenting that before targets are set prevents a team being measured against a condition they did not create. This is the highest-value use of a late finding and it has a short window of its own, because baselines get fixed quickly.
Change the sequence inside the fixed plan. Direction is locked and order frequently is not. If three markets are in scope and one is three times slower, the order can usually still move.
Convert it into the removal of a specific step. A broad finding about speed gets absorbed as a risk. A specific finding about one handoff that takes four days in the new market and four hours at home is actionable at execution level, and the specificity is what survives the reclassification.
Record it as an input for the next decision of the same kind. The second market entry is a framing-phase decision, and it is the right place for a finding that arrived too late for the first one.
Where Horizon fits
Horizon is an AI-powered continuous discovery platform, and its relevance to the timing problem is duration. A diagnosis that takes months can only be commissioned by a programme that already exists, which puts it structurally in phase three.
Discovery Cycles run AI-led interviews asynchronously and in parallel with every person in scope, which compresses the time from question to finding to a matter of weeks. At that duration, a diagnosis fits inside option design rather than requiring a programme to host it. The decision about whether to run it stops competing with the decision it is meant to inform.
The Insights Dashboard ranks findings by impact and effort with traceability to the inputs behind them, which matters for a pre-commitment finding because an uncomfortable result arriving before a decision attracts challenge. A finding with its source attached survives that challenge. The Initiatives Dashboard turns the accepted findings into business cases with named owners, which is the form a phase-two decision needs.
At Trafilea, a cycle ran 43 interviews in two weeks and produced more than 10 findings, against a baseline of roughly a year for comparable work, with 129 hours of discovery effort avoided and a reported 310% return from day one. Two weeks against a year is the difference between a diagnosis that fits inside a decision and one that has to be attached to a programme. These figures describe one engagement and they are evidence rather than a guarantee.
The question worth asking any provider: how long from engagement to findings, and does that duration fit before our next commitment or after it?
Common mistakes
Commissioning discovery with the programme budget. The programme is the thing that fixed the direction, so a diagnosis funded by it arrives after the decision it should have informed.
Treating the formal approval as the deadline. Direction closes at the internal announcement, the first hire and the communicated date, all of which precede the signature.
Running the pilot after commitment. A pilot in phase three demonstrates rather than tests, and a bad result is read as an execution problem.
Letting the home market set the baseline. If the new context is structurally slower, the home number turns a condition into a performance failure for the team that inherited it.
Reporting late findings as risks. A risk register absorbs a finding without acting on it. A specific step with hours attached stays actionable.
Waiting for a complete diagnosis to say anything. Partial findings available in phase two outperform a complete diagnosis delivered in phase three.
Not asking what is still changeable. One question at the start of an engagement, and the answer determines what the work can possibly be worth.
FAQ
When should you run process discovery?
Before the direction is committed, which in practice means during framing or option design. The test is whether the question in the room is still whether to do something rather than how to do it. After budget, dates, hires and announcements are in place, findings can improve execution and will rarely change direction.
Is late discovery worth doing at all?
Yes, with a different purpose. The highest-value use of a late finding is resetting the baseline before targets are set, so a team is not measured against a condition they inherited. After that, changing the sequence inside a fixed plan and converting broad findings into specific removable steps are both available.
How do you tell whether the decision window is still open?
Listen to the question being asked. "How do we make this work" means the direction is settled. "Which of these should we do first" means it is not. The practical markers are a communicated date, a posted role and a contract in legal review, each of which closes the window further than a formal approval has.
Why does discovery usually get commissioned too late?
Because it is easier to fund once a programme exists to attach it to, and the programme is what fixed the direction. The sequence is administrative rather than deliberate, and it reliably places the diagnosis after the decision it was meant to inform.
How fast does a diagnosis need to be to fit before a decision?
Fast enough to run inside option design, which is typically a window of weeks. A diagnosis measured in months can only be hosted by an approved programme. One measured in weeks can inform the decision that would have created the programme.
The finding was right and the clock was wrong
A diagnosis is worth what remains changeable when it lands. The same number can redirect a market entry, reshape a plan, adjust a budget or become a target for a team that inherited the problem, and the only variable is the phase it arrives in.
That makes the first question of any discovery engagement a question about the calendar. What is still open, and when does it close.
See it. Fix it. Win it.