The Curious Case of Innovation ROI Problem, and When It Starts.
- Pinaki Bhowmick
- 6 days ago
- 15 min read
Most organisations blame weak ideas, poor execution, or lack of funding. The real culprit usually appears much earlier... long before the first pilot, business case, or budget request ever exists!
Imagine a typical Monday morning review. The innovation head/Leader walks in with a slide that says: pilots and scale-ups are down again this quarter.
Familiar... isn't it? Normal, one would think! Just another week - eh?
As you might imagine - nobody in the room is surprised. Somewhere between "great idea in the workshop" and "what we built & how it helped" - most innovation programs quietly lose the plot. Funny thing is... everyone seems to have made peace with this... somehow! Another 'cost of doing business'?
Been there is such a room? I'm sure you'd know what I'm talking about!
Except, it isn't the cost of doing business.
This is the cost of doing one specific thing wrong.
And most leadership teams aren't even aware... what that thing is!!!!!
PART 1: First, let's look at what available data tells us about ROI
On one hand, the returns do not look promising
Here's what the data says, and it isn't kind.
BCG's Innovation Playbook puts the pilot-to-scale number at 22% - that's the share of pilots that survive past the scaling phase.
McKinsey found under 30% of pilots ever reach wide adoption, with 84% of companies stuck in what they call "pilot purgatory" for more than a year.
Fortune 1000 data on proof-of-concept to scale conversion lands around 25%.
MIT's 2025 "GenAI Divide" report broke the drop-off into stages that make it worse, not better: of the organisations that evaluated an enterprise-grade AI system at all, only 20% got as far as a working pilot, and just 5% reached production. Ninety-five percent of the generative AI pilots that do get built deliver no measurable P&L impact, not because the models don't work, but because of what MIT calls the learning gap, the failure to fold a tool into workflows, structures, and culture that were never redesigned to receive it.
Now, we did have this conversation last year too - but frankly, at that time it seemed to be a little premature - specially on the context of GenAI success rates!
Honest submission - based on available data, having this discussion now in mid-2026 doesn't seem any less stranger as well.
Yet, on the other hand, the money hasn't slowed down at all.
By the US Bureau of Economic Analysis's own third estimate, AI-related capital investment, data centres, compute hardware, networking, drove roughly 74% of America's GDP growth in the first quarter of this year. That's close to the entire growth story of the world's largest economy sitting on one category of spend.
The capital is moving, at a scale few technologies have ever commanded this early. What MIT's data says, in the same report, is that the organisations receiving all of that capital still mostly aren't absorbing it. The bottleneck was never the size of the cheque, it seems!
At DX&Beyond, roughly 60% of ideas from our programs move to pilot or scale.
Wait. I'm not writing this to wave that number around. A number without a mechanism is marketing wearing an insight's clothes. I'm writing this because the mechanism behind it is more useful than the number itself, and because before I even started this piece, I wanted to know whether anyone else could show comparable proof.
That search (what you've seen above), turned out to be its own finding.
But there is a reason why that I went out there looking for the ROI stats.
I'll be straight about this, because a number like 60% from a firm that isn't yet a household name deserves scrutiny, not applause. Before writing this, I went looking specifically for a recent, named, global example, published in the last two years, showing a genuinely high idea-to-implementation ratio, something that would let you compare our number against a credible external benchmark rather than just taking my word for it.
I didn't find one. Not because nobody is doing good work, but because almost nobody appears to be measuring, or at least publishing, this specific ratio.
HYPE Innovation's 2025 State of Corporate Innovation Report found that:
80% of organisations focus their innovation effort on the early, ideation stage, and 30% or fewer ever track what happens to concepts after that point.
Turning ideas into outcomes was named the single biggest obstacle by 87% of the innovation leaders HYPE surveyed.
Most companies, in other words, simply don't have one to report, because they stopped measuring at the exact point where the real story begins.
The closest thing I found to a counter-example was rather a bit of apples-to-oranges match, but it's worth mentioning - rather than pretending it doesn't exist.
ICF's Innovation Incubator, a structured programme for utility companies, runs ideas through an eight-stage process - from concept to scaled deployment in 12 to 18 months, and reports over 140 completed pilots with roughly 90 more in flight. That's not the same ratio we track, ICF isn't measuring who was in the ideation room, and utilities aren't wrestling with the same politics as a corporate innovation team.
But the underlying logic rhymes: a deliberately staged process, with the right people positioned at each gate before the next stage begins, produces a very different survival rate than letting ideas wander through an organisation looking for sponsors after the fact.
There's one more data point worth sitting with, if only because of its scale.
LEGO's Ideas platform has around 2.8 million contributing members who have submitted more than 135,000 concepts. By LEGO's own published thresholds:
A concept needs 10,000 public supporters before it's even reviewed for production
And by some public estimates only around 7% of the submissions that clear that bar actually become a set.
LEGO isn't trying to get a business unit head to co-own a fan's train design, so the comparison only goes so far. But it's a useful gut check: even with unmatched crowd validation, at a scale most of us will never touch, most ideas still don't cross the finish line unless someone with the authority to build them is already standing on the other side, ready to receive what's coming.
That's really my whole argument. Readiness on the receiving end, not enthusiasm on the submitting end, is what decides whether an idea survives.
Read it again!
PART 2: Most Innovation Programs Play Kabaddi Backwords (and then blame a poor Innovation ROI)
Watch a good kabaddi team defend... and you'll notice something before the raider even crosses the halfway line: The defenders are already holding hands, already reading his eyes and his footwork, already deciding silently who covers the ankle and who covers the escape route.
All of that coordination happens before the raid starts. Nobody links hands after the raider is three steps in and already touching people for points.
Most innovation programs run the play in reverse. Here is how.
An idea gets born in a workshop
The idea gets sharpened inside an innovation team's backlog
And dressed into a business case
And sometimes even quietly piloted in a corner.
Only once it (the idea) has a shape, a cost, a name... does anyone walk it over to the CFO, the COO, the relevant business head, and ask them to suddenly form a chain around something that's already deep inside the circle, already claiming budget, already touching three other departments' plans.
That's not coordination, it's a team trying to link hands mid-raid, after the raider has already scored. Of course he gets away. You'd get away too, if the chain only formed once you were already past it.
Beyond the sport analogy
There's real behavioural science behind why this fails. Norton, Mochon and Ariely's research on what's now called the IKEA effect found that people value things they helped build significantly more than an equivalent thing handed to them ready-made, because effort creates psychological ownership.
Separately, published research on stakeholder engagement in innovation adoption makes a sharper point still: it isn't only whether stakeholders get involved, it's when. Bring them in after the decision is made and you get polite nodding in the meeting, followed by quiet non-cooperation later. But bring them in while the decision is still being shaped, and they defend the idea like it's theirs. Because at that point, it is.
The Metric Behind The Metric
So here's the number we actually track, and it isn't idea-to-pilot conversion - as one of the potential ways of proving Innovation ROI. That's the output.
What we track sits upstream of it: how many of the eventual pilot's future co-owners, whoever will have to fund, staff, or defend this thing later, were physically in the room during ideation itself, hands already linked, before the idea crossed anywhere near a business case.
When that number is high, the 60% mostly takes care of itself.
When it's low, no amount of downstream stakeholder management fixes it, because by then you're asking people to form a chain around a raider who's already sitting in their half, and everyone at the table knows it, even if nobody says so out loud.
Where CIOs, CTOs and CDOs walk into the exact same trap
If there's one function that should already know better, it's technology and data.
In practice, the pattern shows up there just as often, wearing a different badge.
Ask most CIOs or CTOs what their innovation programme delivered this year, and the answer often arrives as a single number: three hundred ideas generated, five hundred, a thousand... It's an easy figure for a board slide, and it's close to meaningless on its own.
Eric Ries flagged exactly this kind of number as a "vanity metric" over a decade ago, and the criticism has only sharpened since: a count of submissions tells you about enthusiasm, not about whether any of it survived contact with a P&L.
Here's the part that rarely makes the slide.
Call a cross-functional "what matters most" workshop, the kind meant to surface genuine opportunities from the business side, and more often than the tech team likes to admit, what comes back is a list of existing bugs, complaints about system performance, and a handful of features that were promised eighteen months ago and never shipped.
Nothing wrong with any of that as work.
It simply isn't innovation, not even the little-i kind.
It's business-as-usual wearing an innovation workshop's name tag.
McKinsey's research puts a number on why: technical debt now consumes roughly 40% of the average IT balance sheet, and close to a third of CIOs say more than a fifth of the budget earmarked for "new" work quietly gets redirected to fixing what already exists.
Ask a room what matters most, and if that room has been carrying that debt for years, it will name the debt.
When something that does resemble genuine innovation surfaces from these sessions, it usually arrives as a pile, a use-case backlog pulled together from whichever cross-functional stakeholders happened to be in the room, unvalidated, and often quietly contradicting priorities two departments over.
The uncomfortable part, the one I've watched play out more times than I can count, is that the technology team rarely pushes back on any of it. Not because every idea is sound, but because questioning a business stakeholder's framing carries a risk that questioning a system log never does: the risk of looking like you don't understand the business well enough to have an opinion on it.
Amy Edmondson's research on psychological safety, and Google's own well-documented Project Aristotle study of what makes teams effective, both land on a version of the same finding: people stay quiet in a room not because they lack the insight, but because the room hasn't made it safe to be the one who says "I don't think that's actually the problem." Technologists, sitting one perceived rung below the business owner in that room, feel that risk most acutely, and it shows up as silence dressed as agreement.
CDOs live a version of this with even less room to hide, and the latest survey data makes the shape of it sharper than it was even a year ago. Deloitte's 2026 Chief Data and Analytics Officer survey found the highest percentage it has ever recorded, 93%, naming human issues, culture and change management, as the real barrier to AI and data value. Only 7% pointed to the technology itself.
That's about as close as a large-sample survey gets to confirming, in the data leaders' own words, that this piece's whole argument is correct: the constraint was never the model, or the platform, or the use case. It's whether the humans downstream were ever brought into the room.
And yet the same wave of 2026 research shows something that complicates the tidy version of this story, which is worth sitting with rather than smoothing over.
90% of firms now report having appointed a Chief Data Officer, more than a third have gone further and appointed a dedicated Chief AI Officer, and roughly 70% of respondents describe the CDO role itself as "successful and established."
Put that next to the tenure data: Chief Data Officers still average only around 30 months in the seat, more than half leave within three years, and close to three in ten say, when asked directly, that they don't see a future in the position at all.
Set that against a CIO's roughly four and a half years, or a CEO's nearly seven, and you get a strange picture, a role that most organisations now consider embedded and successful, and that keeps losing the people who hold it faster than almost any other seat in the C-suite.
No - its not that one of those numbers is wrong. It's that "the role is established" and "the person in it can't get buy-in fast enough to survive it" are both true at once, and that tension is exactly the CDO's version of a use-case inventory that impresses the board without a chain of hands ready to receive any of it.
PART 3: Three Frameworks Worth Stealing
Cross-functional co-ownership is the non-negotiable one, the load-bearing wall in this whole argument. But if that's the only tool you walk away with, you'll be well-aligned and still under-proven, because alignment isn't the same as evidence.
Here are three that work together, not in sequence exactly, but as three lenses on the same problem: who's in the room, what can you actually claim, and what's your idea's honest weak point before someone else finds it for you.
The Chain Before the Raid
Before any idea moves past the workshop stage in our programs, we make the sponsor answer one uncomfortable question: whose hands aren't in this chain yet, who will need to say yes later? Then we build what we call the Chain Before the Raid. Three names, no more.
Let me be honest about what that looks like in practice, because written down it can sound suspiciously clean, as if three names appear and the raider gets tackled on cue.
It isn't clean, and it isn't fast. Getting these three people into the same room, and getting them to genuinely agree on anything, routinely takes months, not weeks. The moment you put an operational owner, a budget gatekeeper and a business sponsor together early, everything that was previously being managed by simply not talking to each other surfaces at once: silo-based wishlists nobody had reconciled, practical dependencies nobody had mapped, real disagreement on risk exposure and on what's strategically urgent this quarter versus next year, and, more often than any of us like admitting out loud, plain data hoarding and authority power plays, because whoever controls the information or the sign-off has usually been protecting that leverage for a while and won't hand it over just because a workshop asked nicely. Uff!
None of that resolves in a single session. What a well-designed engagement produces, over that time, is a loosely agreed version everyone can live with, not the final one. That's the actual job of the intervention, and it's the whole engagement doing that work, not one workshop: surfacing friction early and often enough that by the time the idea reaches a business case, the fights have already happened where they're cheap, in the room, on a whiteboard, instead of later in a steering committee where the same disagreement now has a budget line attached and everyone's more dug in.
So, the three roles this is designed to force into the open:
One operational owner, the person who will actually have to run this once it's real, not just approve a slide describing it.
One resource gatekeeper, whoever controls the budget or headcount this idea will eventually need, positioned in the chain before the ask, not summoned at the ask.
One deliberate skeptic, someone whose entire job in that room is to find the flaw now, while it's cheap to fix, instead of six months later in a steering committee where saying no has become politically expensive for everyone involved.
All three stand in the chain during ideation, not after, and getting them genuinely aligned rather than merely present is the slower, harder part of the work. It routinely adds months to what looks, on a project plan, like a straightforward ideation phase. But the ideas that survive that friction arrive at the pilot stage pre-negotiated instead of post-negotiated. Slower, messier start. Genuinely faster, cleaner finish.
This role only works if it comes with explicit permission to be wrong out loud. Put a technologist in the skeptic's chair opposite a business sponsor, or a business owner in the skeptic's chair opposite a technical lead, without naming that permission first, and you haven't added a skeptic. You've added a fourth quiet person to the chain, nodding along for the same reasons everyone else in that room usually does.
The Three Numbers That Matter
If cross-functional buy-in is the mechanism, this is how you prove it's working, and it's the piece most programs skip, which is exactly why HYPE found 87% of innovation leaders naming "turning ideas into outcomes" as their biggest obstacle. You cannot prove ROI on a funnel you're not measuring past its first stage.
Track three numbers, not one:
Number one: ideas submitted, or use-cases catalogued. The vanity metric. Easy to inflate, easy to celebrate in a townhall or a board slide, almost meaningless on its own. This is the number most CIOs, CTOs and CDOs already have, and proudly report.
Number two: ideas that reach a named, cross-functional business case, meaning an operational owner, a resource gatekeeper and a sponsor have all put their name to it before a single rupee or dollar of pilot budget is spent. This is the upstream metric described above, and it's the one almost nobody tracks, because it requires admitting how few ideas actually clear that bar.
Number three: ideas that reach pilot or scale. The lagging output everyone reports, when they report anything at all.
The ratio between numbers two and three is usually close to whatever number your organisation would be proud to publish. The ratio between numbers one and two is where the truth lives, and it's exactly the gap
Deloitte's 93%-versus-7% finding points at from a different angle: everyone already agrees the barrier is human, almost nobody has built a metric that actually measures the human part. If your organisation can state number one confidently and goes quiet on number two, you don't have an innovation problem yet. You have a measurement problem, and it's hiding the innovation problem from you.
For a CHRO or Chief Learning Officer, this framework does double duty: number two is also a capability audit, because chronically low conversion from submission to named ownership usually means the organisation hasn't built the muscle to run these conversations, not that the ideas were weak.
The Pre-Mortem, borrowed and adapted
Cognitive psychologist Gary Klein published a technique in the Harvard Business Review nearly two decades ago that still hasn't made it into most innovation playbooks: before a project launches, gather the team and ask them to imagine it's eighteen months from now and the initiative has failed completely. Then have everyone write down why.
Run this before an idea leaves the ideation stage, with the same three people from the Chain Before the Raid in the room, and you get something closer to an actual ROI case than most business cases produce, because you're surfacing the specific ways the return doesn't materialise, rather than assuming it will.
A pre-mortem run with the operational owner in the room usually surfaces two or three genuinely fatal risks that a standard business case template never asks about, because templates ask "what's the opportunity" and rarely ask "who quietly kills this in month four, and why."
For a Chief Experience Officer or a COO sitting on the sponsoring side of a pilot, this is also the fastest way to build a credible ROI narrative before you have real numbers. You're not promising an outcome. You're demonstrating that you've already stress-tested the ways it could fail, and priced that risk in. Boards and finance teams trust that kind of proof more than an optimistic projection, and they should.
The honest caveat
None of this is an argument for lining up thirty defenders. A kabaddi side plays seven a side for a reason, too many hands in the chain and nobody knows whose job it actually is to make the tackle.
Consensus-by-committee kills speed just as reliably as isolation kills adoption, and I've watched well-meaning inclusivity turn a six-week ideation sprint into a six-month opinion-gathering exercise that goes nowhere.
The Chain Before the Raid works precisely because it's three names, not thirty. The goal isn't democratising every decision. It's identifying the two or three people whose later "no" would actually kill the idea, and getting their objections on the table while the idea still costs nothing to change.
The months it takes to align three genuine decision-makers is friction earning its keep. The months it takes to gather opinions from thirty people rarely arrive anywhere at all.
Cross-functional buy-in negotiated after the raid is already underway isn't buy-in. It's a scramble that gets minuted as if it were a plan.
So, a question back to you
What's your organisation's current idea-to-pilot conversion rate?
And before you answer, a harder question underneath it: do you actually track number two, the ideas that reached a named cross-functional business case, or only numbers one and three, the ones that make for an easy slide?
Most CHROs and COOs I ask this to go quiet for a second.
So do most CIOs when I ask how many of last year's cataloged use cases have a named business owner attached, and most CDOs when I ask what happens to the backlog once the roadmap review ends.
That pause usually tells me more than any number would. If you're a CTO, a CDO, or a Chief Innovation Officer reading this and you do track it, I'd genuinely like to know what you found, because as this piece admits upfront, that data point is rarer than it should be, even in a year when the capital behind it has never been larger.
Comments