Can the wisdom of crowds be applied to project status?

What the effect actually requires

The popular version of "wisdom of crowds" — that groups are simply smarter than individuals — isn't true, and believing it leads to committees. James Surowiecki's account is narrower and more useful: crowds outperform experts when four conditions hold.

• Diversity of information. Members hold different pieces of the picture.
• Independence. Each judgement is formed without being shaped by the others.
• Decentralisation. People draw on local, specific knowledge rather than a shared summary.
• Aggregation. A mechanism combines the individual judgements into a collective one.

Where those hold, the individual errors are somewhat independent and tend to cancel; what's left is signal. Where they don't, the errors correlate and you get a confident group that's confidently wrong.

How a project team scores

This is the argument Genchi is built on, so it's worth walking the conditions one at a time and being explicit about which ones come free and which had to be designed in.

Diversity — naturally strong

A backend engineer, a mobile developer, a designer and a QA lead each see a different face of the same project. The person three weeks into an integration knows things nobody else does. This is exactly the kind of distributed information the effect requires, and it's the reason a project team is a better candidate than, say, a group of executives who all read the same report.

Decentralisation — naturally strong

Same reason. People are drawing on what they've personally encountered rather than on a shared briefing document. A status meeting actually erodes this: after everyone has heard the same summary, their subsequent judgements are based on the summary, not on their own evidence. It's one reason Genchi asks for the check-in before and independently of any team discussion, rather than collecting a show of hands at the end of one.

Aggregation — trivially solvable

This is arithmetic. It's also the condition most conventional processes fail, because their aggregation mechanism is a person writing a summary — which is judgement, not aggregation. See why bad news gets softer as it travels up.

In Genchi the aggregation is literally arithmetic, performed on the server, with no human step between the responses and the number a leader sees.

Worth being precise about what "aggregation" means here, because a single average would throw away the most useful part. Genchi shows the aggregated score and a visualisation of the individual votes that produced it — the values, not the voters. You can see that four people said 4 and one said 2. You cannot see which person said 2.

Independence — the one that breaks

In any ordinary team setting, independence is almost impossible to maintain. If the tech lead says "yeah, we're in decent shape" before anyone else speaks, everyone else's assessment is now anchored to that. Not because they're deferential — because that's how anchoring works, and knowing about it doesn't switch it off.

Your team already has a technique for this. Planning poker exists precisely because estimates given publicly and sequentially converge on the first number spoken. The simultaneous reveal isn't ceremony; it's the mechanism that preserves independence long enough for the aggregation to mean something.

A confidence check-in is the same problem, distributed over time instead of a meeting. It needs the same protection, which is why Genchi collects each response privately and shows only the aggregate. Nobody sees the team's current score before answering, and nobody can see who answered what afterwards. That is the single design decision the whole approach rests on — without it, the other three conditions produce a well-aggregated measurement of what people think their manager wants to hear. The argument in full: why project confidence votes should be anonymous.

What this gets you that one estimate doesn't

Two things, and the second is underrated.

Error reduction. One person's optimism bias is the whole estimate. Six people's optimism biases are six partially independent errors, and the aggregate is less extreme than the most optimistic member. It doesn't eliminate the bias — more on that below — but it stops one person's disposition determining the reading.

Visible disagreement. A single estimate has no variance. Six responses do, and the spread carries information the average doesn't. Five people at 4 and one at 2 is a very different situation from six people at 3.7, though both average to roughly the same place — and the first is far more interesting.

This is why Genchi displays the individual vote values alongside the aggregate rather than collapsing everything into a single number. A mean would report those two teams as equivalent. Seeing the spread means the outlier is visible at a glance, which in our research was the specific thing leaders said they wanted: one described wanting to know when "someone is seeing something the rest aren't."

An engineering leader we interviewed put the appeal succinctly:

"Really like that it's so simple, so trivial. Made for collecting a lot of data. Detect trends sooner. Get the average of the crowd."

The objection: status isn't a guessing game

One of our interviewees pushed back on the whole premise, and it's the strongest objection we heard:

"Not comfortable crowdsourcing status. I want it done in a consistent way." — project manager

The concern is legitimate. Classic crowd-wisdom demonstrations involve estimating a quantity with a true value — the weight of an ox, the number of beans in a jar. "Will we hit the deadline?" has a true answer only in retrospect, and the responses are subjective assessments rather than measurements.

Two responses. First, the alternative isn't an objective measurement — it's one subjective assessment, made by the person most exposed to the consequences of pessimism. Aggregating six subjective assessments isn't obviously worse than trusting one.

Second, consistency is a fair requirement and it's a property of the question, not the number of respondents. Everyone answering the same fixed question on the same fixed scale at the same cadence is more consistent than prose summaries, not less.

Where it fails

Correlated error. The serious limitation. If the whole team has been told the vendor will deliver on time, they'll all be confident, and averaging six people who believe the same wrong thing produces a confident wrong answer. Independence protects against social influence within the team; it doesn't protect against everyone being misled from outside. One interviewee noted his team "rarely disagreed on the success of the team" — which reads as consensus, and might equally be an absence of diverse information.

Small teams. Aggregation needs numbers, for signal quality and for anonymity. Below about five people, both get thin.

Unclear goals. If people don't share an understanding of what success means, they're answering different questions and the average combines incompatible things.

This is what Genchi is

Genchi is the wisdom of crowds applied to project status, and there's no more to it than that. Every team member answers the same question at a set cadence — how confident are you that we'll achieve our goal? — on a one-to-five scale, in about two seconds.

What you see is both halves: the aggregated score for the initiative, and the individual vote values that make it up, shown without attribution. The aggregate tells you where the team is. The spread tells you whether they agree. The trend across recent check-ins tells you which way it's moving.

Mapped against the four conditions:

• Diversity and decentralisation come free, from asking everyone doing a different part of the work rather than one person summarising all of it.
• Independence is the part that had to be engineered: private responses, no visible running total, no attribution afterwards.
• Aggregation is arithmetic on the server, with no editorial step — and the individual values remain visible alongside the aggregate, so disagreement within the team isn't averaged away.

The lineage isn't novel and we'd rather not pretend it is. Genchi takes the ergonomics of the standup thumb check, the independence mechanism of planning poker, and the longitudinal record of a niko-niko calendar, and points the combination at a single question about whether the goal will be met. Three practices your team already trusts, made to persist and to travel beyond the room.