Innovation Center Governance Models That Actually Work
Innovation centres are rarely killed for failing to innovate. They are killed because their governance made failure look like incompetence, or because nothing they produced was ever adopted. Both are governance problems, and both are decided before the first bet is taken.
The core difficulty is that innovation and delivery need opposite governance. Delivery governance asks what shipped, rewards predictability and treats failure as a defect. Innovation governance asks what was learned, rewards decisive abandonment and treats failure as expected output. Applying the first to the second produces safe, incremental work that was not worth doing offshore.
This guide compares four governance models, sets out the stage gate structure that makes any of them workable, and covers the adoption mechanism that determines whether the centre survives.
Key points
- Delivery governance applied to innovation produces safe work and no invention
- Funding model is the governance decision that matters most
- Stage gates should test learning, not progress toward a deliverable
- A high kill rate is evidence of discipline and should be reported as such
- Adoption must be designed in at gate one, not discovered at the end
The four models
Four governance structures appear consistently in practice. They differ in how much the centre decides for itself and how the money flows, which between them determine almost everything else.
| Model | Who sets the agenda | Funding | Best suited to | Main risk |
|---|---|---|---|---|
| Directed | Core business assigns the questions | Project by project | First innovation mandate, low trust | Becomes a delivery pod |
| Sponsored portfolio | Sponsor sets themes, centre chooses within them | Annual allocation | Most organisations, most of the time | Sponsor turnover kills it |
| Autonomous | Centre chooses within a strategic boundary | Annual allocation, discretionary | Mature centre with a track record | Drift from company priorities |
| Internal venture | Centre pitches, an internal board funds | Stage gated tranches | Companies with strong venture discipline | Heavy process overhead |
The sponsored portfolio model is the right answer for most organisations. Directed governance is a reasonable starting point but should have a scheduled end date, or it becomes permanent by default.
Why the funding model is the real governance decision
Everything else follows from how money reaches the centre. Project by project funding means each initiative must be justified in advance, which reintroduces an approval gate and biases the team toward proposals that are easy to defend. Easy to defend means predictable, and predictable means not worth exploring.
An annual allocation changes the behaviour immediately. The centre can take a bet that will not resolve for three quarters, can stop a bet without having to explain a cancelled project, and can run several attempts in parallel knowing most will fail. That is the operating freedom the mandate requires.
The condition is portfolio level accountability. The centre is answerable for what the whole portfolio produced over the year, not for the outcome of any individual bet. Without that, annual funding simply becomes project funding with extra steps.
Stage gates that test learning
Whatever model is chosen, the gate structure does the real work. The question at each gate should be what has been learned and whether the bet still looks worth continuing, never what percentage complete the work is.
Is the question worth answering?
A short framing exercise: what would have to be true for this to matter, roughly what is it worth, and who in the business would receive it. Most ideas should stop here, cheaply.
Is it technically plausible?
A rough prototype answering the single riskiest technical question. The output is a judgement, not an artefact, and building a polished demo at this stage is a warning sign.
Does it hold under real conditions?
Real data, real constraints, real scale characteristics. Most technically plausible ideas fail here, and that failure is valuable because it is now specific.
Will the business actually adopt it?
Confirm the receiving owner still wants it and can take it. Running this gate late is the single most common cause of wasted innovation spend.
Transfer and hand over
Move it to a team that will run it, funded separately. The centre should not operate what it invents, or within a year it becomes an operations team.
Two disciplines make this work. First, killing a bet at a gate is a successful outcome and should be recorded and reported as one. Second, the gates must be cheap. If reaching gate two costs a quarter of effort, the centre will only propose ideas it is confident about, which defeats the purpose.
The adoption mechanism
The characteristic failure of innovation centres is not invention; it is that nothing gets picked up. Prototypes are built, demonstrated, admired and abandoned, and after two years the centre is closed on the reasonable grounds that nothing reached a customer.
This is preventable with two structural decisions. Name a receiving owner in the core business at gate one rather than gate four, and treat the absence of a plausible receiver as a reason not to start. An idea with no identifiable home has already failed; it just has not been told yet.
The second decision is to hold productionisation budget separately from exploration budget. If the centre has to fund industrialising its own successes, it will run out of money exactly when something works, which is the worst possible moment for that to happen.
Not the person who will admire it, the person whose team will run it. If nobody will claim it at the start, that is the cheapest possible moment to learn that.
Exploration funding and productionisation funding must be distinct, or the centre will stop exploring in order to finance its own transfers.
The receiving team should be involved before handover, not at it. Adoption failures are usually surprises that a single earlier conversation would have prevented.
The headline metric should be outcomes adopted by the business. Counting prototypes rewards volume and produces a demo factory.
What to report, and what to stop reporting
Innovation reporting fails when it borrows delivery metrics. Throughput, velocity and percentage complete are all meaningless here and actively harmful, because they reward choosing problems small enough to finish inside a reporting period.
What an innovation centre review should lead with
Illustrative weighting rather than measured research. The point is the ordering: adoption first, resolved bets second, and delivery metrics essentially absent.
Governance failures to avoid
Each of these is a reasonable governance instinct applied to the wrong kind of work.
- Reviewing the centre in the same forum and format as delivery teams
- Requiring each initiative to be approved in advance, which reinstates the gate autonomy was meant to remove
- Treating a killed bet as a failed project in reporting, which teaches the team never to kill anything
- Leaving the adoption question until the work is finished
- Funding productionisation out of the exploration budget
- Letting the centre operate what it builds, which converts it into an operations team within a year
- Relying on a single sponsor with no succession plan, which is how most centres die
Choosing where to start
Most organisations should not start with autonomous governance. Trust has to be earned in both directions, and a centre with no track record given full discretion tends to produce work the business does not recognise as relevant.
Directed, with an end date
The core business sets the questions. This builds credibility and shared vocabulary. Put a review date in the plan so it does not become permanent by inertia.
Sponsored portfolio
The sponsor sets themes, the centre chooses within them, funded by annual allocation with portfolio level accountability. Most centres should stay here.
Autonomous, if earned
Move to a strategic boundary and full discretion only if the centre has demonstrated adopted transfers. Autonomy without a track record is usually withdrawn at the first difficult quarter.
Protect the sponsor relationship
Sponsor turnover kills more innovation centres than poor results. Build the relationship with more than one senior stakeholder from the start.
Governance decides the output before any work is done,The funding model, the gate questions and the adoption route determine what kind of work the centre will produce. By the time you are reviewing results, those decisions have already set the ceiling.
Frequently asked questions
Which governance model should we start with?
Directed governance with an explicit end date, moving to a sponsored portfolio in the second year. Starting with full autonomy rarely works because the centre has no track record and the business has no basis for trusting its choices.
How do we stop delivery governance creeping back in?
Review the innovation mandate in a separate forum with a separate agenda, and never in the same meeting as delivery. The creep is almost always a consequence of shared review meetings, where delivery questions get asked because that is what the forum is for.
What is a healthy kill rate?
High. If most bets survive their gates, the centre is choosing problems too safe to be worth exploring. A portfolio where the substantial majority of bets are stopped before production is behaving normally and should be reported as functioning correctly.
How much should the innovation budget be?
Small enough that failure is affordable and large enough that the work is real. What matters more than the amount is that it is an allocation rather than a series of approvals, and that productionisation is funded separately.
What if the sponsor leaves?
This is the most common cause of death for innovation centres. Build relationships with more than one senior stakeholder from the beginning, and keep a current record of transferred outcomes so the centre s value does not rest on one person s memory.
Should the innovation centre report into the GCC lead?
It can, provided the mandates are governed separately with different scorecards and a ring fenced budget. What fails is appearing as a line in the capability centre s delivery plan, where it will be reallocated to delivery work the first time a deadline is at risk.
How do we handle a bet that keeps looking promising but never resolves?
Set a decision date at each gate and hold it. Bets that persist without resolving consume the portfolio and are usually a sign that the gate question was never sharp enough to be answered. Killing them is a governance function, not a technical judgement.
How long before governance should be reassessed?
Annually, and always after a sponsor change. Governance that suited a centre with no track record is usually too restrictive by year three, and a centre operating under governance that no longer fits will quietly stop taking interesting bets.
Sources & further reading
- NASSCOM — https://nasscom.in/
- McKinsey & Company — https://www.mckinsey.com/
- Deloitte — https://www.deloitte.com/
- Everest Group — https://www.everestgrp.com/
Govern innovation so it survives its own failures
Hexominds sets up innovation centres with the funding model, stage gates and adoption route defined before the first bet is taken.