The Approval Chain That Kills Creativity
A few years ago, a mid-sized retailer tested a loyalty idea that broke from category convention. It didn’t survive first contact with the org chart. By the time it reached launch, seven people had improved it in ways that made perfect sense individually. What launched looked exactly like what every competitor already had.
Nobody in that chain made a bad call. That’s what makes this worth examining. The problem isn’t that people say no to good ideas. It’s that a good idea can pass through six sets of reasonable hands and come out the other side unrecognisable, and everyone involved will tell you, honestly, that they were just doing their job.
The usual complaint about approval processes is that they’re slow. True, but it’s the least interesting thing wrong with them. The real damage happens earlier and quieter: people start shaping ideas around what they expect the process to tolerate, long before anyone official has weighed in. The approval chain stops being a thing that happens to an idea and becomes a thing that happens inside the head of the person having it.
Nobody sets out to build a bad process. Reviewers get added because someone wants a decision to benefit from more expertise, and that instinct is sound. The trouble is what “more expertise” turns into once it’s operationalised. Six perspectives brought together to think about a problem is one thing. Six perspectives each given the standing ability to stop the work is another, and organisations routinely build the second while believing they’ve built the first.
Woolley and Malone’s research on collective intelligence is useful here mostly because of what it ruled out. Groups didn’t perform better simply because they contained smarter individuals. What correlated with performance was closer to how evenly people got to speak, and how well the group read each other. Later work has picked apart exactly which factors matter most, so it’s worth not treating this as a formula. But the basic point survives the argument: putting good people in the same room, or the same chain, doesn’t automatically combine their thinking. Somebody still has to design how the thinking moves.
Most companies never make that design explicit. They use one word, “stakeholder,” to cover at least five different jobs: contributing information, challenging an option, offering expert advice, making the call, and holding actual veto power. Collapse those into one category and something predictable happens. Being invited to weigh in starts to feel like having the right to block. Nobody voted for that shift. It just accumulates.
The part of this that’s hard to see, and therefore easy to ignore, is what happens upstream of any actual rejection.
Give someone enough experience with a review process, and they’ll learn its shape without anyone teaching it to them. They’ll know which kind of proposal triggers a finance objection, which one makes brand nervous, which one needs a citation nobody has. And the next time they’re generating options, they won’t hand the organisation the full range and let it choose. They’ll do the choosing themselves, quietly, before the meeting.
This is worth sitting with because it inverts the usual story about creative failure. We tend to picture a bold idea getting shot down in a room. More often it never makes it to the room. There’s a version of rejection that happens out loud, and a version that happens in someone’s head three weeks earlier, and only the first one shows up in any record of what the organisation considered. You cannot audit the ideas nobody proposed.
If there’s a single sentence worth keeping from all of this, it’s that the real cost of heavy approval isn’t the good ideas leadership turns down. It’s the ones that never get written down at all.
Here’s a pattern that shows up in almost every version of this problem. Finance is doing its job when it tightens the numbers. Legal is doing its job when it flags exposure. Brand is doing its job when it asks for consistency. Every function along the chain is protecting something real, and every one of them is, by design, looking at the idea through the lens of what could go wrong from where they sit.
Run an idea through five or six of those lenses in sequence, and each one takes a small bite out of the thing that made it interesting. Nobody along the way is acting against the organisation’s interests. Nobody is even wrong, exactly, in the moment they make their request. But there’s no equivalent function whose job is to ask what the idea would lose if this particular caution won. Risk has people whose salary depends on catching it. Upside mostly doesn’t.
Which suggests a genuinely useful habit for anyone sitting in one of these review seats: alongside “what risk am I reducing here,” it’s worth asking what you might be quietly removing that isn’t yours to give back.
One mistake shows up constantly enough that it deserves its own section: treating idea generation and idea evaluation as though they need the same room, the same people, and the same conditions.
Keum and See’s research on hierarchy and innovation found something that sounds almost too simple to matter: hierarchy tends to get in the way when a group is generating options, and tends to help when a group is choosing between them. Generation wants independence, some friction, room for a bad idea to exist for five minutes without anyone naming it bad. Selection wants the opposite: named criteria, real trade-offs, someone with the standing to actually decide.
Most companies run both in the same meeting. “Bring your boldest thinking” tends to arrive alongside a room that includes the CFO, legal, and whoever runs brand. Everyone present has technically been invited to imagine freely. What they’ve actually been asked to do is produce something that survives the people in the room, which is a much narrower task and a much duller one. The fix isn’t complicated to state, even if it’s harder to build: let ideas take shape before anyone with real authority is allowed near them, then bring authority in deliberately, later, once there’s something worth judging.
It would be satisfying to end on the usual note here: less hierarchy, more autonomy, trust the team. The research doesn’t actually support that as a general rule, which is a more interesting finding than the tidy version would have been.
A large communication-network study, over 1,600 participants, found that centralised structures could help groups adapt to changing conditions, but only when information moved in both directions, out from the centre and back in. Once influence started flowing mainly one way, outward, the benefit vanished. Centralisation wasn’t the villain. One-directional centralisation was.
So, the better question isn’t whether authority should sit at the top or spread across the team. It’s which parts of the process actually need to spread, and which need to stay narrow. Sensing what’s happening in the market, challenging an assumption, bringing specialist knowledge- those can involve a lot of people without causing harm. Making the final call usually can’t. Twenty people might have something useful to say. Two might have a genuine right to stop the work. One needs to decide. The mistake most organisations make is quietly promoting all twenty into the second category.
None of this argues for less governance. It argues for governance that knows what it’s governing.
A workable starting point is naming roles explicitly for each kind of recurring decision, rather than defaulting to “stakeholder” for everyone in the room. Who’s closest to the problem. Who has information the originator doesn’t. Whose job is specifically to poke holes in it. Whose expertise should shape the decision without giving them the power to end it. Who actually owns the call. Who has a genuine, defined right to stop it, tied to something real like legal exposure or safety, not just seniority. Who has to execute once it’s decided.
That last category, true veto holders, is usually much smaller than people assume. When McKinsey’s researchers mapped decision networks inside one large consumer company, they found that 40 per cent of the interactions around a typical decision added no real value- people who’d been looped in out of habit rather than necessity. Fixing that wasn’t about writing a better chart. It was about being honest that most of the people in the room were there because someone once thought it seemed safer to include them, and nobody had gone back to check.
There’s a useful way to name the property an organisation is actually after here: how much good information can reach a decision without every source of that information also gaining the power to block it. Organisations that get this wrong tend to solve every expertise gap the same way, by adding the expert to the approval list. Do that enough times and every useful voice becomes another checkpoint.
The alternative is keeping voice and veto as genuinely separate things. People can raise a concern, bring evidence, argue for an alternative, without that involvement quietly hardening into a permanent stop button. That’s a harder thing to design than it sounds, because the instinct to reward useful input with more authority is a natural one. It’s also, in most cases, exactly backwards.
There’s a related pattern worth naming on its own: the ideas most likely to matter are often the ones with the least precedent behind them, which means they attract the most scrutiny at the exact moment there’s the least evidence available to resolve it. Familiar ideas glide through because everyone’s seen something like them before. Unfamiliar ones get stopped and re-stopped precisely because nobody has. That’s a strange thing for a process built to protect the organisation’s future to be doing to the ideas most likely to shape it. A cleaner rule would tie the intensity of review to how consequential and how reversible a decision actually is, not to how novel it looks or who might happen to notice it.
Most companies are fluent in measuring what a decision produced. Almost none of them measure what it cost to get there in time, attention, and rework. That’s worth fixing, because the numbers tend to be uncomfortable. How long between having enough information and actually deciding. How many layers a proposal has to clear. How many people can say no relative to how many will answer for the outcome. How often something gets sent backwards. Whether ambitious proposals quietly get revised more than safe ones, which is usually a sign the process is doing something other than what it claims to be doing.
Surveys of executives consistently find only a small minority believe their own organisation is genuinely good at making decisions, and by some estimates executives spend close to 40 percent of their time on decisions- a share that climbs past half once they're deep enough into the senior ranks that most of it happens in meetings. That’s not a minor inefficiency sitting off to the side of the real business. For a lot of companies, it functionally is the business, just poorly instrumented.
None of this is an argument against oversight, or against the people doing it. Almost every layer in a typical approval chain got added for a defensible reason at the time: a near miss, a compliance gap, an executive who wanted visibility into something that mattered. The problem isn’t that any single layer was a bad idea. It’s that organisations are much better at adding checkpoints than at retiring them once the reason for them has passed. What’s left behind, eventually, is a structure made almost entirely of old caution, largely disconnected from anything currently at risk.
The goal was never an organisation where everyone gets a vote. It’s one where everyone’s relevant knowledge can actually reach the decision without everyone also picking up the power to end it. Good ideas don’t usually die because someone senior says no. They die more slowly than that, worn down by a sequence of people whose jobs quietly reward finding a reason to hesitate over finding a reason to move.
