PINO^®

Problem → Insight → Narrative → Outcome

A four-stage discipline for analysis work. Each stage has an entry condition, an exit condition, and a characteristic way of failing. An analysis is "PINO" when all four stages cleared. Anything else is a stage that got skipped.

The four stages

P — Problem

What result are we going after, and which decision unlocks it?

Start at the result. A decision is not the destination — it is the vehicle that gets you there. Framing the stage this way keeps the team pointed at a number that has to move, instead of at the satisfaction of having made a call.

Entry	A result someone owns: a number that has to move, with a size and a deadline.
Exit	One written sentence — the target result, the decision that unlocks it, the metric that tracks it, and the threshold that would change the call.
Fails as	Starting from the table that happened to be available. Or treating the decision as the finish line: the call gets made, the deck gets applauded, and nobody ever checks whether the result moved.

I — Insight

What does the data say that we didn't already believe?

The data stage. Full toolkit, no purity contest about method — SQL and the warehouse, cohort and segment work, hard-won knowledge of how the operation actually behaves, AI leverage for exploration and speed. The only constraint is that the finding survives scrutiny.

Entry	Sources reconciled, base defined, baseline established before the result is known.
Exit	A finding that would have surprised the room a week ago — sized, segmented, uncertainty declared, and with the plausible rival explanations already ruled out: selection, seasonality, mix shift, a control group that was never comparable.
Fails as	A number that is true, correct, and tells nobody anything new. Symptom: the reaction is "right, we knew that."

If it can't survive "but couldn't it just be…?", it doesn't move to N. Kill it here, cheaply, rather than in front of the committee.

N — Narrative

Is it ready to face executives — and will it survive the room?

This stage is not about being right. It is about being adopted. Four gates, in order:

Strategy — align before you present. Walk the idea past the key stakeholders privately, before any data hits a screen. Does it hold up next to the initiatives already running? A surprise in a C-level room is a no by default, regardless of the quality of the work. (For findings whose whole value is that they contradict the room, see the escape hatch below — this gate is not a filter for inconvenient news.)
Tie out to the big numbers. Does the analysis reconcile with the KPIs everyone already accepts? If your number disagrees with the official number, you explain the gap first — before someone in the room finds it. Never make executives choose between your analysis and their dashboard.
MECE. Mutually exclusive, collectively exhaustive. No overlapping buckets, no unexplained residual, no segment that quietly appears in two places.
Pyramid, 30 seconds. Answer first, then the three supporting arguments, then the evidence underneath. If you can't deliver it in 30 seconds, you don't have it yet — you have material.

Escape hatch: when the finding contradicts what the room believes

Gates 1 and 2 exist to prevent ambush, not to filter out inconvenient findings. A finding whose entire value is that it contradicts the accepted view still goes forward — under a heavier standard, not a lighter one. You don't soften the finding. You relocate the surprise.

First, classify the contradiction.

Numerical — our number disagrees with the accepted number. The deliverable is a bridge: walk from the official number to ours one line at a time, each delta named, quantified and attributable — different base, different date cut, a different definition of active, cancellations treated differently, one source silently excluding a channel. Every step reconciles; nothing is left as "methodology differences." Present it as a waterfall where possible.

Rule: the divergent number never appears without the bridge on the same slide. And the framing is never "the dashboard is wrong" — it's "these two numbers answer different questions; here's the one that answers ours."

Interpretive — the numbers agree, the mental model doesn't. The deliverable is evidence: state what would have to be true for the existing belief to hold, then show that it isn't. Prefer evidence the room already trusts — their own dashboard, a prior pilot, an operational fact the frontline will confirm on the spot. Borrowed credibility beats new credibility.

Then pre-align the contradiction itself. Take the bridge, or the evidence, to the owner of the number or the belief — privately, before the room. Gate 1 still applies: they simply don't get surprised in public. If they accept it, you walk in with a co-author instead of an opponent. If they don't, you walk in with two readings and the test that would settle which is right — a defensible position. A public argument is not.

What this hatch never becomes: a reason to leave the finding in a drawer. If it cleared I and the only obstacle is that it's unwelcome, the hatch is mandatory, not optional.

Entry	An insight that survived the technical challenge.
Exit	A pitch that opens with the recommendation, reconciles with accepted numbers, splits MECE, and has been pre-walked with the people who could kill it.
Fails as	A technically flawless analysis that dies in the room — because it surprised someone, contradicted a dashboard, or took four slides to reach the point.

O — Outcome

What gets done, by whom, by when — and who is watching?

Entry	Decision taken, narrative approved.
Exit	Action plan with an owner per action; timeline with milestones; governance defined — cadence, forum, and where the numbers live; guardrails; and a re-read date on which the target result from P is checked.
Fails as	"Next steps: continue monitoring." Symptom: nothing changes in the operation and the same deck reappears next quarter.

The re-read in O is what opens the next P. The cycle closes; it doesn't end.

Review checklist

Use when reviewing someone else's work — or your own, before it leaves your hands.

I can name the result this analysis is chasing, not just the decision.
The baseline was defined before the result was known.
The finding is sized — not just directional.
Rival explanations were named and tested, not merely acknowledged.
Treatment and control were comparable before treatment, and I can show it.
The numbers tie out to the accepted KPIs — or the gap is explained on my slide, not discovered on theirs.
The stakeholders who could kill this already know it's coming.
If my number contradicts an accepted number, the line-by-line bridge is on the same slide.
If it contradicts a belief, I've stated what would have to be true for that belief to hold — and shown that it isn't.
The owner of the contradicted number or belief saw it before the room did.
It splits MECE, and I can deliver the whole thing in 30 seconds, answer first.
Every action has an owner, a date, and a forum where it gets followed.

Items 10–12 apply only when there's a contradiction to handle; skip them otherwise and score out of nine.

All applicable items: ship it.
One or two gaps: name them out loud before presenting.
Three or more: it isn't PINO yet.

Where analyses actually die

Two distinct graveyards, and they require different defenses.

Technical death, at I. The segment is real, the uplift is real — and what was never ruled out is that the treated group was the group that was going to stay anyway. Selection effects are the single most common cause of a result that looks strong in the pilot and evaporates at rollout. The comparability check is the highest-leverage question in the entire cycle.

Organizational death, at N. The analysis is correct and still goes nowhere: it contradicted a number the executive committee already believes, it landed as a surprise on someone who had a competing initiative in flight, or it buried the recommendation on slide nine. Being right is the entry ticket to this stage, not the outcome of it.

And the quiet one, at O. A validated, well-received finding with no owner and no date is a cost, not an asset.

Worked example

P — Cost per save in retention is unsustainable and the result we need is a lower cost per save at a stable save rate, by end of cycle. The decision that unlocks it: replace the fee waiver with a capped voucher. Threshold: save rate holds within X points at materially lower cost.

I — Acceptance concentrates where the voucher covers close to 100% of the monthly fee. Below that ratio acceptance drops; well above it, breakage rises and we pay for a benefit nobody uses. Checked against the obvious rival: is this the voucher working, or just the highest-spend segment, who were the least likely to leave anyway?

N — Pre-walked with retention, pricing, and finance before the deck existed. Reconciled against the official churn and revenue-at-risk numbers so the room never has to choose sides. Segments split MECE. Opens with the recommendation, not with the methodology.

O — Capped voucher for the segment where the ratio holds; downgrade offer where it doesn't. Owner per action, rollout milestones, guardrail on cost per save, monthly forum, and a re-read date where we check the result from P — not just whether the thing shipped.