The launch metric is green, but the support calls suggest people are leaving before account creation. Can we check what the metric excludes before we call this a success?
The dashboard and the customer calls are describing different parts of the journey. We should not choose between them without checking the definition.
The current sample starts after account creation. Anyone who leaves before that point is missing from the denominator, so the green number is narrower than we thought.
Who agreed the success definition before launch? I want to understand the decision, not have the metric interpreted as my personal preference.
I am uploading the metric spec and the raw funnel. The current definition is valid for one step, but it does not cover the customer outcome we discussed.
This is not really a disagreement between the dashboard and the calls. We are using a metric whose boundary was never made explicit.
Agreed. Let us add the missing step, document the limitation, and decide whether the launch call needs to be revisited.
The classic product warning is easy to picture: the highest-paid person shares an opinion, everyoneโs evidence develops stage fright, and the roadmap changes before lunch.
Hence the HiPPO.
At our launch review, Victor arrived two minutes late and said almost nothing. He asked one question about retention, wrote a note, and spent the rest of the meeting listening with the unsettling competence of a leader who has read the pre-read.
The most powerful participant was already there.
It occupied the central screen. Cell G14 was green. A small arrow leaned upward. The dashboard headline said Activation +18%, and the room responded with the quiet relief normally associated with a medical test and an unexpectedly affordable restaurant bill.
Bea advanced to the next slide.
Priya did not.
โWhat happened to the people who left before creating an account?โ she asked.
The dashboard had no opinion on those people. Technically, it had no people. It had a denominator, and the denominator had been selected during launch planning when the team believed the most important question was whether newly registered users completed setup.
The metric was correct.
It was also wearing Victorโs lanyard.
Opinion is not the enemy
Product decisions require judgment. Evidence is incomplete, futures are not observable and somebody eventually has to choose.
A senior leader may know something the room does not: a financing constraint, partner risk, legal exposure or strategic commitment. Pretending every decision can be settled by an experiment is another form of theatre. Some choices define the environment in which experiments become possible.
The problem is not that an influential person has an opinion. The problem is when source, assumptions and decision rights remain hidden.
โWe are doing this because the CEO wants itโ is crude but legible. โThe data says soโ can be harder to challenge even when the data reflects a narrow choice made by somebody else.
The green activation metric contained at least four opinions: when the journey began, which users counted, which behaviour represented value and how long the team should wait before judging. Those opinions had been converted into a definition, then a query, then a chart. By the time they reached the review, they looked like weather.
Maya asked Bea to walk backward from the cell.
Bea, to her credit, did not defend the dashboard as though it were a family member. She explained the event, cohort, exclusion and time window. The team discovered that paid acquisition had changed the mix of visitors. More of the people who created accounts completed setup, but a larger share of visitors never created accounts at all. The chart described a healthier middle of a leakier funnel.
Both statements were true. Only one had received a green background.
The quiet forms of authority
Titles are only one costume authority wears.
There is the dashboard HiPPO, where a number becomes unchallengeable because its definition is invisible.
There is the deadline HiPPO, where โwe need this by Q3โ closes discussion without revealing the dependency, commitment or cost of being late.
There is the customer escalation HiPPO, where one urgent account becomes the entire market because revenue is concrete and opportunity cost is not.
There is the research HiPPO, where a confident synthesis conceals who was sampled, what was asked and which evidence did not fit.
And there is the deck-owner HiPPO, the person who controls order, framing and what fits on a slide. They may have no formal authority while quietly deciding which facts arrive before the decision and which live in the appendix after everyone has mentally left.
These forms matter because teams trained only to resist senior opinions may miss the authority embedded in process. They congratulate themselves for being evidence-led while leaving the evidenceโs constitution unwritten.
Activation increased 18%. The funnel declined 11%.
- Included: people who created an account
- Excluded: visitors who left before registration
- Changed: acquisition mix during the launch window
- Decision risk: optimising a healthier middle of a leakier journey
Experiments do not remove judgment
Teams often respond by asking for an experiment. That can improve the decision, but it does not eliminate choices about metrics, segments, duration and acceptable trade-offs.
Microsoft researchers summarising experimentation across thirteen organisations describe practical challenges that extend far beyond running a statistical test: trustworthy metrics, organisational culture, infrastructure and the interpretation of results all matter. The useful lesson is that experimentation is a socio-technical capability, not a machine that converts disagreement into truth.
An experiment can tell you whether a particular change affected a defined measure for a defined population during a defined window. It cannot decide which customer outcome deserves protection, whether a short-term gain justifies a long-term cost, or whether the organisation should enter the market at all.
In our review, the launch had included an experiment. The variant improved completion after account creation. That result remained valid. The mistake was promoting it from โthis part of the flow improvedโ to โthe launch is working.โ
Victor finally spoke.
โWhat decision are we making?โ
This question had the annoying effect of shrinking thirty minutes of analytical theatre into one sentence.
The team was deciding whether to increase paid acquisition and move engineers away from onboarding. Under the broader funnel view, both actions looked premature. They kept the improved setup flow, paused the spend increase and gave Tomโs team one cycle to investigate pre-account abandonment.
Victor had influenced the outcome, but not by overruling evidence. He had clarified the decision boundary. Seniority can distort a room; it can also make hidden stakes visible. The job is not to remove authority. It is to make its use inspectable.
A decision preface
Intercomโs decision-making matrix offers a practical way to widen input while keeping responsibility clear. The broader principle is useful: expertise, input and final decision authority are different things. Confusing them creates either dictatorship or a committee with excellent attendance and no steering wheel.
Before an important review, write a short preface:
- Decision: what must be chosen now?
- Decider: who has the authority and responsibility to choose?
- Inputs: which evidence and expertise should shape the choice?
- Assumptions: which definitions or constraints are being treated as fixed?
- Revisit trigger: what new signal would reopen the decision?
This makes a disagreement easier to diagnose. Are people contesting the evidence, interpretation, objective, risk tolerance or decision right? Those are different conversations. Mixing them produces the familiar meeting where everyone is correct about a different question.
Bea updated the metric spec in the review itself. The green cell became one part of a balanced view: visitor-to-account conversion, setup completion, first-value behaviour and an early retention signal. Priya added qualitative evidence from people who left. Maya wrote down why the team paused acquisition.
The dashboard did not become less useful. It became less royal.
The authority audit
When a decision feels strangely predetermined, ask where authority entered:
- Who defined the objective?
- Who selected what counts as evidence?
- Who framed the options?
- Who owns the deadline or commercial commitment?
- Who can make and revisit the final decision?
Sometimes the answer is a senior person. Sometimes it is the query written three months ago. Sometimes it is the absence of a person whose experience did not survive the funnel.
The audit is not an invitation to relitigate every decision. Teams need speed. It is a way to spend disagreement where it can change the quality of the choice, rather than performing evidence after the choice has become socially irreversible.
At the next review, Victor entered on time. The dashboard showed four measures and one explicit caveat. Bea had added a small note beneath the activation chart: Green is a colour, not a conclusion.
Leon objected to the typography. Everyone else agreed to keep it.
My take
The HiPPO is not always the loudest person. It is whatever the room is least permitted to question.
Do not replace senior opinion with metric worship. Expose the assumptions, name the decider and clarify which decision the evidence can actually support. A spreadsheet is a useful witness. It should not chair the meeting.
Questions to take back to your team
- Which current metric contains the most consequential hidden definition?
- Who is absent from the denominator of your most celebrated dashboard?
- Which deadline is treated as natural law even though somebody chose it?
- Can your team distinguish expertise, input and final decision authority?
- What evidence would genuinely cause the last major decision to be revisited?
Pick one green metric and walk backward from colour to query to definition to judgment. You may not find a villain. You will probably find a surprisingly well-dressed opinion.