Where we start
Session 1 gave you five categories to sort candidates into. Session 2 gave you a verdict to pass on each one. Neither of those tells you what to do first, because both are about a single case in isolation. Tonight the whole portfolio goes through at once, and the interesting results are all about the interactions: what one case blocks, what one field unlocks, and how much of a plan can be wanted by everybody and worth nothing.
★ What you walk out with tonight. Your own candidate list scored through both gates, the name of the one entity that unlocks the most on it, and at least one case you now know to stop defending.
Twelve candidates, twelve entities, two gates
6 min liveRidgeline Fitness is a 22-site gym chain. It makes its money on recurring memberships, with personal training and a cafe alongside. That last sentence is the whole of the second gate.
LiveWhat the bench computes, and why fit per week is the only rankable number4 min▶
Clears the gate. Every entity the case needs is held in full. Not planned, not partly, not "we could get that" - held. A case that needs one entity the business does not have is not a slow case, it is a different project with a data-collection programme in front of it.
Revenue fit. Nought to three, scored against how this business makes money. It is a judgement and the bench shows it rather than hiding it, so you can disagree and watch every conclusion move.
Weeks. Build effort of what you actually commit to. Note what happens in the first preset: the weeks include the blocked cases, because a plan that contains blocked work still spends the estimate on it.
Fit per week is revenue fit that clears, over weeks spent. It is the only one of the four you can rank on, because the other three can each be improved by doing something stupid. You can raise "clears the gate" by adding empty cases, cut "weeks" by cutting value, and raise "fit" by promising things you cannot build.
Self-studyHeld in full, held partly, not held: why the middle one is the dangerous state3 min▶
Three of Ridgeline's twelve entities exist but are incomplete: the cancellation reason is a free-text box the front desk fills in about a third of the time, member goals are collected on paper at some sites and never keyed in, and the injury log is a paper folder per site.
Not held at all is an honest state. Everybody can see the sensors are not on the machines. Held partly is the state that wastes projects, because the field appears in the schema, the query runs, and the answer comes back looking like an answer. A retention analysis built on a reason field filled a third of the time will produce a confident and worthless ranking of reasons.
The bench marks those cases conditional rather than clear, which is a deliberately uncomfortable word. It means: buildable, and not yet trustworthy. Session 4 is about what to do with that state, and it turns out to be the most valuable position on the whole board.
Five ways to pick, computed in front of you
★ 26 min · the main eventWork down the presets in order. Read the five numbers before the verdict line, and read the unlock table at the bottom last, because it is the one that changes what you do on Monday.
Read it in this order. Clears, then blocked, then fit, then weeks, then fit per week. The first preset is the list as it actually arrives from a round of workshops, and it is worth sitting with how bad it is before you look at any of the others.
Three results worth taking into a meeting
6 min liveEvery figure below came off the bench you just used, computed from the entity map rather than asserted.
| How you pick | Clears | Fit | Weeks | What it tells you |
|---|---|---|---|---|
| Everything everybody asked for | 6 of 12 | 12 | 64 | Half the plan does not clear the gate, and the estimate charges you for it anyway |
| Only what clears the gate today | 6 | 12 | 27 | The same six cases clear, so the same fit, for 37 fewer weeks |
| Chase the three exciting ones | 0 | 0 | 25 | Twenty-five weeks, three blocked cases, and a revenue fit of nothing |
| Finish one field first, then go | 7 | 15 | 33 | The only route to the highest fit on the board, and the only row with nothing blocked |
| Just ship the chatbot | 1 | 0 | 4 | The most-requested item on the list. Clears every gate but the one that matters |
LiveDropping the blocked third cost exactly nothing3 min▶
The first two rows have the same revenue fit: 12. Identical. The second row takes 64 weeks down to 27.
The reason is almost too simple to state, which is why it keeps happening: work that cannot be built contributes no value, but it does consume the estimate. A twelve-item plan where three items are blocked is not a twelve-item plan with a risk attached. It is a nine-item plan being costed as twelve, and the three will either quietly drop out in month four or drag a data-collection programme into scope that nobody agreed to.
This is the most robust finding on the bench, and the one to take into a planning meeting, because it does not depend on any judgement about revenue fit at all. It only depends on which entities the business holds, and that is a question with an answer.
LiveThe most-requested thing on the list is worth nothing3 min▶
The front-desk chatbot that answers questions about opening hours was tied for the most-requested item in the whole portfolio. It needs no data entities at all, so it clears the data gate instantly, and it takes four weeks.
Its revenue fit is zero. Ridgeline makes its money on whether members keep paying, and nothing about answering an opening-hours question faster moves that. The thing that makes this case instructive rather than merely silly is that it is genuinely easy, genuinely wanted, and genuinely harmless. There is no argument against it except the one the bench makes.
That is why the second gate exists. A portfolio filtered only on feasibility will fill up with exactly this: cheap, popular, clean, and empty. "Can we build it" and "does it pay" are different questions, and the first one is much easier to answer, which is why plans drift toward it.
Self-studyTwo numbers on the bench you should not rank on3 min▶
Worth naming, because the bench will happily print them and they do not mean what they appear to.
Fit per week, between the second and fourth rows. They come out at 0.444 and 0.455. That gap is not a result. It rests on a stipulated one-week cost for finishing the cancellation-reason field, a number chosen here rather than measured, and at two weeks the ranking reverses. What genuinely separates that fourth row is not its efficiency but its ceiling: it is the only way to reach a fit of 15, and the only row with nothing blocked.
Weeks, compared across rows with different scopes. The third row is cheaper than the second and produces nothing. Cheapness is only meaningful next to what clears, which is precisely why fit per week exists, and why even it has the limits above.
Every effort figure on this bench is stipulated. That is a real limitation and not a coy one: estimating build effort honestly is difficult, it is a separate skill, and a bench that pretended otherwise would be teaching you to trust the wrong number.
Running both gates without a bench
4 min liveFour passes, on paper, on the list you brought. It takes about twenty minutes for a list of a dozen and it removes the arguments that cannot be settled.
Write down the data entities first, before the cases. Not systems, not reports - entities. A member, a visit, a payment, a booking. Most lists of candidate use cases have fewer than fifteen distinct entities underneath them, and writing them out is what makes the blocking visible.
Mark each one held in full, held partly, or not at all, and be harsh about the middle one. "We have that field" is not the same as "that field is filled in". If you are not sure, it is partly.
Gate each case against that map. A case needing anything not held in full is blocked, and blocked is a category, not a delay. Do not let a blocked case sit in the plan with an asterisk.
Score revenue fit against how the business actually makes money, out loud, with somebody who owns the number. Disagreement here is the useful part and it is much cheaper now than in month four.
Before session 4
30 min- Finish the entity map for your own candidate list, with each entity marked held, partly or not at all. Bring it; session 4 is built on exactly this artefact.
- Find your zero-fit case. Most portfolios have one that is popular, cheap and does not move the money. Name it before somebody commits to it.
- Work out what the blocked cases are costing you in the current plan: add up their estimated effort and ask whether anybody has agreed to the data work that would unblock them.
- Ask one person who owns a revenue number to score your top five for fit. Note where they disagree with you, because that gap is the real finding.
Three questions
Clear all three and session 3 gets stamped in your passport.
1 · Dropping everything that does not clear the gate today, three blocked and three only partly held, took the plan from 64 weeks to 27 with the revenue fit unchanged at 12. Why did dropping half the list cost nothing?
This is the most robust result on the bench because it needs no judgement about revenue fit at all. It depends only on which entities the business holds, which is a question with an answer. Option A is false here: one blocked case carried a fit of 2 and one partly-held case carried the highest fit on the whole board, a 3.
2 · The front-desk chatbot needs no data, clears every gate, takes four weeks, is tied for most-requested, and scores a revenue fit of zero. What is it teaching?
What makes the case instructive rather than silly is that it is easy, wanted and harmless, so there is no argument against it except the one the second gate makes. C overcorrects: request volume is evidence about something, just not about revenue fit.
3 · On the bench, "finish one field first" scores 0.455 fit per week against 0.444 for "only what clears today". How should you read that difference?
A difference that flips when you change a number somebody chose is not a finding. The honest move is to say so and point at what survives the change instead, which here is the ceiling and the blocking rather than the efficiency.
What this session teaches, and where it came from
Every fact on this page, and its verification tier, is recorded in the course's source map.
Session 3 cheat sheet · pin this
Look again at the unlock table at the bottom of the bench. The entity that releases the most value on that board is not attached to any of the exciting cases: it is a free-text box the front desk fills in about a third of the time, and finishing it is a form change rather than a programme. Session 4 is about the data gate on its own, why the half-held entity is the most valuable position you can be in, and how to tell a form change from a standing job before you agree to either. Bring the entity map from tonight's homework.