I once sat through a demonstration of a machine learning model that predicted equipment failure in a manufacturing business. It was genuinely impressive work. The accuracy was good, the visualisation was clear, and the room was pleased.
I asked one question. Who acts on the prediction?
There was a pause, and then a slightly awkward conversation about how the maintenance schedule was set by a third party under a contract with two years to run, and how the shift patterns did not allow for unplanned intervention, and how the person who would need to act on the alert was not in the room and had not been consulted.
The model worked. The project was theatre. Those two things are entirely compatible, which is exactly what makes this difficult.
01What is innovation theatre?
Innovation theatre is activity that produces the visible signals of innovation without changing how the business operates or what it earns.
The signals are familiar: pilots, proofs of concept, innovation days, a lab, a partnership announcement, a slide with a roadmap on it. None of these are bad in themselves. Every real innovation programme produces some of them. The difference is what happens after the demonstration, and in theatre the answer is nothing, or more precisely, another demonstration.
The reason it persists is that it is genuinely enjoyable and it is genuinely defensible. Nobody gets criticised for running a pilot. The work is interesting, the people doing it are engaged, and the outputs are impressive to anyone who does not ask the follow-up question. It also satisfies a real pressure that founders are under, which is to be seen to be doing something about a technology everybody is talking about.
What it does not do is produce a return, and the cost is not just the money. It is the credibility. A business that has run four pilots that went nowhere is measurably harder to get moving on the fifth thing, even when the fifth thing is right.
02What are the tells?
Five, in rough order of reliability.
Nobody's existing objectives improve if it works. This is the strongest signal by some distance. If the initiative succeeds completely, and no individual in the business has a target that gets easier, it will not survive contact with a busy quarter. It is competing against work with customers attached, and it loses.
The success measure is an activity, not an outcome. "We will run six pilots this year" is an activity. "We will reduce quote turnaround from nine days to two" is an outcome. Activity measures are how theatre reports progress, because they can be achieved regardless of whether anything worked.
The running cost has never been calculated. Building it is the cheap part and everybody models the build. What almost nobody models is what it costs to operate for three years: the licences at real volume, the person who maintains it, the retraining when the process changes, the thing you now cannot change because this depends on it. Initiatives die of running costs far more often than they die of build costs.
It is owned by whoever was most enthusiastic. Enthusiasm is a poor proxy for accountability. The right owner is the person whose numbers change, who often was not in the original conversation and is sometimes actively sceptical. That scepticism is useful information rather than an obstacle.
The pilot cannot fail. If the criteria for stopping have not been written down before starting, the pilot will not stop. It will be extended, rescoped, or declared a learning exercise. A pilot with no failure condition is a demonstration with a longer run time.
The question is never whether the technology works. It is whether anything about Tuesday afternoon is different once it does.
03The three questions that save ten weeks
When something is proposed, ask these before anybody builds anything. Together they take about ten minutes and they resolve most cases.
Which number changes? Name it. Not "efficiency" and not "customer experience", but a specific figure that exists somewhere in the business today and that somebody could show you. If the honest answer is that no current number moves, the proposal is not an opportunity yet. It might become one, but at the moment it is a capability looking for a use.
By how much, and who says so? The size matters, and so does who is willing to put their name to it. A 15% improvement claimed by the person proposing it is a very different item from a 15% improvement that the operations director thinks is achievable. If the person who owns the number will not endorse the estimate, that is the finding.
What does it cost to keep running, for three years? Build cost, then run cost, then the cost of the thing you can no longer change. Multiply the licence by real volume rather than pilot volume. Include the half a person who will maintain it. This question kills more initiatives than the other two combined, and it kills them early, which is the entire point.
If a proposal survives all three, it is worth spending money on. If it fails one, that is usually fixable by reframing rather than rejecting: find the number, find the owner, resize the ambition.
04What does real opportunity look like on paper?
Narrower than people expect, and considerably duller.
Real opportunity is specific about who benefits, and that person is usually named. It is specific about the current state, with a measurement rather than an impression. It has a defined smallest useful version that could be in production in weeks, not a phased programme with an eighteen month horizon. And it is owned by somebody whose job gets better if it works.
It is also, almost always, unglamorous. The initiatives that produce genuine returns in mid-sized businesses tend to be about quoting, scheduling, invoicing, handover, and getting information from one system to another without a person retyping it. They do not demonstrate well. Nobody wants to present them at a conference. They change what the business earns.
The glamorous version is not automatically theatre, to be clear. Occasionally the ambitious thing is the right thing. But it should have to clear a higher bar, not a lower one, and in practice it usually gets the reverse.
05Is there a place for exploration with no defined return?
Yes, and this is where I part company with the more austere version of this argument.
Some proportion of what a business does should be genuinely exploratory, with no number attached, because you cannot find opportunities you have not gone looking for. The mistake is not doing it. The mistake is doing it without saying that is what you are doing, so that exploratory work gets judged by delivery standards and delivery work gets excused by exploratory ones.
Name it. This quarter we are spending four weeks and a small budget finding out whether there is anything in this. Success is a clear answer, including a clear no. Then apply the three questions to whatever comes out.
Kept separate, both work. Blurred together, you get a business that is neither exploring properly nor delivering properly, running pilots forever, with a founder who has concluded that innovation does not work here.
It does. It just needs somebody in the room to ask who acts on the prediction.