Most portfolios don't fail because of bad ideas— they fail because the system was never designed to kill anything.

Projects accumulate. Budgets get split. Attention gets diluted. And somewhere in the middle of a portfolio that looks busy, the zombies are eating your budget while your best bets starve. Projects that should have been stopped long ago: not killed, not funded, just kept alive by a system that defaults to inertia.

Your team already knows which project it is. The problem is that knowing isn't enough. Zombie projects don't survive because they're strong. They survive because no one built a system that could stop them.

Building that system —one where the kill decision is the natural output of honest data— is what separates portfolios that produce breakthroughs from portfolios that produce slide decks.

"Zombie ventures are not a sign of bad people," says Ben Yoskovitz, founding partner of Highline Beta and co-author of Lean Analytics. "They are a sign of a bad system. You need to build a system where ruthlessness is the natural output of looking at the data honestly."

That system has two parts: the metrics that surface the truth early, and the decision-making process that acts on it. This article is for the innovation leader who already suspects which projects should go and wants the system to back them up. It covers what to measure, how to make the kill decision stick, and why the hardest part usually has nothing to do with the data.

Innovation portfolios fail in one of two ways. There's the single big bet that gets protected past the point of reason, because too much has been invested for anyone to let it fail. And then there's the spray-and-pray portfolio, where a dozen half-funded ventures consume resources without clear criteria for what success looks like, until the portfolio quietly becomes a graveyard of slow deaths.

The portfolio math is straightforward. Of roughly 30 validated concepts, maybe 10 get built, and of those, 1 scales. So the pipeline has to be full: 15-20 ideas in exploration, 5-8 in validation, 2-3 scaling — all running in parallel. Because the moment you stop ideating to focus on scaling, the pipeline goes cold, and six months later there's nothing behind the current batch.

However, this isn't the part that trips innovation leaders up. Where programs still break down, even ones that have the pipeline running and the stages in place, is the measurement system underneath. The metrics chosen at each stage are what determine whether weak concepts get caught early or drift through unchallenged. That's where zombies come from, and that's where the fix has to start.

Every corporate innovation program has an operating model. Stages, gates, deliverables — discover, validate, build, scale. And if you line them up side by side, they all look remarkably similar. The framework is not the differentiator. What separates portfolios that make clean kill decisions from portfolios that debate the same stalled projects all the time, is the metrics chosen at each stage. And the two stages where that choice matters most are discovery and validation.

Getting the metrics right here means that by the time you reach a kill decision, the answer is already clear: the data has been building toward it. Getting them wrong means the whole system is shaky, and every portfolio review turns into a debate.

There are three measurement failures that show up frequently in these stages:

That third failure deserves extra attention, because it's also the most common way that well-intentioned frameworks accidentally create zombies.

Example: One company built a solid stage-gate system with clear goals, indicators, and methods at each stage. So far, so good. But then they added a column for "minimum expectations": conduct over 50 face-to-face interviews, ideate over 100 concepts, test with over 50 subjects. The teams started optimizing for the checklist. They hit all the numbers, ticked all the boxes, and moved projects forward that should have been stopped, because the system was measuring effort rather than evidence. As a result, the framework became a zombie factory disguised as a process.

The fix is to focus Discovery-stage criteria on what the market told you, not what the team did. At this stage, the questions should be:

If a concept can't answer these six questions with evidence from the market, it hasn't earned the right to move forward.

Once concepts are scored and compared, each surviving concept is condensed into a one-pager so leadership can evaluate it quickly: the problem, the solution, the value proposition, the strategic fit, the next steps, and simple red/yellow/green signals on desirability, viability, and feasibility.

Surviving Discovery doesn't mean a concept is safe. Validation is where the expensive mistakes get prevented, because what comes next is an MVP budget, a team, and real resources. The questions shift from "is the problem real?" to "will this actually work as a business?"

By the end of Validation, the team should be able to say clearly: this is the real problem, this is what the solution should look like, this is who our early adopter is, this is the scope for an MVP, and this is how the business model and pricing work. If any of those sentences come out fuzzy, the concept isn't ready for the resources it's about to receive.

Projects that coast through Discovery on a promising technology or an interesting problem space get exposed here. The question that separates momentum from evidence: is there real proof this will work as a business, or are we running on excitement?

Scorecards can surface the information needed to decide whether a project should keep going. What they can't do is make someone act on it.

The scorecard is on the screen. The reds are visible. The data points in one direction. And yet the project keeps going: not funded properly, not stopped cleanly, just lingering at low intensity because nobody has the guts to kill it.

When a project is evaluated and the evidence says it shouldn't continue, there are only three real moves:

This diagnosis is rarely where things stall. It's the step after, carrying that conclusion into a room where the incentives, the politics, and the unspoken rules all push toward "let's keep going."

The reason most people can't kill a project is not a lack of frameworks. It's a lack of safety.

Annual budget cycles force teams to request funding before they know whether something is working — and if you don't ask for the money, you're guaranteed not to get any. So people ask, even when the evidence is telling them to stop, because opting out of the budget cycle feels like opting out of their own relevance.

What often happens is that teams present the data, assess it against criteria, and then stop there, handing the whole thing to a venture board or executive stakeholders without ever saying what they think should happen. The decision gets made by people who weren't in the weeds, without a recommendation from the people who were.

But budget cycles are not the only blocker. The pattern shows up in other ways: individual performance reviews that don't reward honest assessment, so killing your own project feels like killing your own career. Cultural norms that treat any form of stopping as failure rather than learning. A review process that doesn't create space for kill recommendations. Or sometimes the simplest version: one specific person in the room won't let go, and nobody wants to be the one to challenge them.

The result? The structures that are supposed to fund innovation end up protecting zombies instead.

The structural fix is to separate the recommendation from the decision. The process has four steps:

Step three is where most organizations have a gap. The person doing the work should own the recommendation. Leadership owns the final call. But without that recommendation, leadership is flying blind, and the default becomes "let's keep going."

The innovation professional running the project needs to be the one comfortable enough to put their hand up and say: this should stop. Not because it's easy, but because they're the one with the clearest view of the evidence. If that recommendation has to come from someone else, it's already too late.

Pick your weakest project and walk through the four steps above. Where did you get stuck?

If you stalled at Step 1 or 2: your problem is measurement. The data either doesn't exist or the criteria were never defined.

If you stalled at Step 3: ask yourself honestly — is it because the evidence is genuinely unclear, or because making the recommendation feels unsafe? If it's the latter, your problem is the room, not the data.

If you stalled at Step 4: your problem is governance. There's no clear decision-maker, which means nobody owns the call, which means nothing gets stopped.

The single shift that makes kill decisions stick is treating a parked project not as a failure of the people who worked on it, but as a portfolio allocation decision: redirecting limited resources to where they'll create the most value.

When teams internalize this, everything changes. Every parked project generates value if the learning is captured. Every failed experiment makes the team sharper for the next cycle. The point isn't to celebrate failure: it's to celebrate learning, and to build a system where that learning compounds. A three-week experiment that produces a clear learning is not failure. A three-year project that consumed tens of millions and taught nothing is.

The moment stopping is treated as a legitimate, respected outcome rather than something to be avoided, the zombie problem will start to dissolve.

Everything in this article, the scorecards, the filters, the kill decisions, the reframe, describes how the system should work. RBC is what happens when someone actually runs it.

The Royal Bank of Canada built a venture studio with Highline Beta and committed to the discipline for five years. Not a quarter-long pilot. Five years of running the full cycle. Hundreds of ideas entered the funnel. About 15 internal ventures were actually built and launched. Four startup acquisitions were made. Over 13 investments. All deployed based on what the situation demanded, not what felt most comfortable.

And then came the success that most case studies leave out. The killing. Constant, methodical, ongoing. Ideas that looked promising in a workshop but collapsed under primary research. Ventures that passed discovery but stalled in validation. Projects that got built, launched, found users, and still got parked because the strategic case wasn't strong enough or the traction wasn't steep enough. For five years, the default answer was "no, not this one" far more often than "yes." Every kill freed up money, people, and attention for something stronger.

It's easy to imagine most organizations easing off after year two. RBC kept going. Four ventures made it through. Dr.Bill does medical billing, putting RBC in front of high-value wealth management clients. Houseful builds services around the housing journey, differentiating in a commoditized mortgage market. Ownr handles business registration, because every new business needs a bank account. Mydoh helps kids manage money, capturing a generation that won't inherit their parents' bank. Each one creates value independently and feeds customers back to the core business. A few others were absorbed into core operations, smaller wins from the same process.

Every winner has two things: its own standalone business model, and a strategic hook that feeds value back to the core bank. That's not a coincidence. That's discipline, applied relentlessly over time, doing its job.

Resources got freed. Attention got focused. The oxygen went to the things that were working. If RBC had kept all fifteen alive, splitting the budget, rotating leadership attention, avoiding the politics of shutting things down, none of the four would have broken through.

None of this requires a revolution. It requires a scorecard with honest thresholds, a review process where recommendations are expected, and a culture where parking a project is treated as good portfolio management rather than a personal defeat. Small structural changes, applied consistently, are what separate portfolios that produce breakthroughs from portfolios that produce slide decks.

RBC didn't transform its innovation pipeline with a single dramatic decision. It did it by making hundreds of small, honest ones: scoring rigorously, recommending clearly, killing early, and redirecting resources to the projects that earned them. Five years of that discipline produced four ventures that now drive real value back to the core business.

The same opportunity is sitting in your portfolio right now. The scorecards are here. The process is here. All that's left is the courage to start.

Pick the weakest project on your list today, run it through the filters, and make that recommendation you've been putting off. The first kill is the hardest: everything after that is just discipline.

Alexa Dembek, former Chief Technology and Sustainability Officer at DuPont, knows what walking away from a project she believed in feels like.

She spent years building DuPont's energy storage business from zero on a direct CEO mandate. "I remember crying the day we shut it down. But it was the right decision."

Previously, Ben argued that the biggest blocker to killing projects isn't the data, it's the culture. When stopping is perceived as failure rather than learning, no system will save you. Alexa tackled this culture gap head-on with a ritual called Dead Projects Day, an annual celebration where killed projects aren't punished but honored, and the learning is extracted in front of the entire innovation organization. Here's how to run one yourself.

At DuPont, the event is held every year around Halloween, complete with costumes, cupcakes with orange frosting, gummy worms, and buttons that read Fall in love with growth, not projects. The format is intentional: it frames stopping as something worth celebrating, not something to sweep under the rug.

You don't need a Halloween theme. You need a recurring, visible, calendar-blocked event where killed projects are the sole agenda. What matters is that it's predictable: teams should know the day is coming and start thinking about which of their projects belongs there. Alexa says people now come to her saying "I've got a project that should be on Dead Projects Day." That's the behavior shift you're designing for: teams volunteering their own projects, because stopping is respected here.

On the day of the event, open the session by making every killed project from the past year visible. DuPont calls this the Innovation Graveyard. Every project that was stopped gets named and displayed, so the room can see the full picture of what was parked and why.

When the whole room can see that dozens of projects were parked, and that every team is represented, killing a project stops feeling like an individual failure and starts looking like a normal part of how the innovation portfolio works. It reframes the narrative: these aren't mistakes on a wall. They're decisions that freed up money, people, and attention for work that actually has a chance to deliver.

The core of Dead Projects Day is three teams each presenting the story of a dead project: the original ambition, what actually happened, and what the team learned. DuPont calls these ghost stories. Each is followed by a panel discussion where the room digs into the patterns together.

The ghost stories force those questions into the open. Hearing a colleague walk through the real timeline of a project they invested themselves in, including where the warning signs showed up and why they kept going, lands differently than reading about common innovation pitfalls in a book.

The final part of the session shifts from looking back to looking forward. Alexa uses this time to work through techniques that help teams decide whether to pivot, pause, or persist on their current active projects. The pitfalls and red flags surfaced by the ghost stories become something every team can apply immediately.

This is where Dead Projects Day pays for itself. Three dead projects produce learnings that sharpen dozens of live ones. Teams leave not just with a better understanding of what kills projects, but with the language and permission to raise those concerns on their own work.

It’s important to note that Alexa didn't delegate Dead Projects Day. She led it personally, as the most senior innovation leader in the organization. The seniority in the room shapes how safe it feels for teams to present honestly. The title slide for her next Dead Projects Day says it all: "If we don't innovate, we lose." When leadership shows up and says that publicly, killing a project stops being a confession and starts being a contribution.

Imagine inheriting 550 people, 50 active projects, and a mandate to build a billion dollars in new revenue, inside an 90-year-old Japanese company where more than 80% of the revenue comes from a single product that's going extinct.

This is what Dirk Schapeler's first weeks at Niterra, the company formerly known as NGK Spark Plugs, looked like. He was the first outside hire at a senior level in the company's history. Seventeen thousand employees, four billion USD in revenue, and a leadership team that had prepared for the decline of combustion engines for twenty years but with only moderate success to diversify. Every internal attempt had been led by the same people: all engineers, all career lifers, none of whom had ever worked outside the company. "If we could," Dirk says, "we would probably make spark plugs until the end of life."

Dirk spent weeks reviewing every one of those 50 projects. Weeks of presentations on video calls during COVID, listening to pitch after pitch. Most followed the same pattern: teams had built around a technology they were proud of, then gone looking for someone who might want it. The engineering was impressive, to say the least, but underneath those pitches, there was often no customer validated problem that they would be willing to pay for, no identified customer and no evidence of demand.

Dirk set out on a quest to find out how. Soon enough, he introduced MIT's Disciplined Entrepreneurship framework, the 26-step methodology that forces teams to start from the other end: what is the actual problem? Who has it? Will they pay to solve it? He ran every single project through it. The ones that couldn't answer those questions got stopped.

He also gave teams a pitch deck template, modeled on what a startup would use, that they could download and fill in. This was a mental exercise, designed to force people to think through the things they had to consider: the problem, the customer, the business case. Whether they used his exact template or something better didn't matter, as long as they actually went through the thinking.

Eventually, 80% of the portfolio didn't make it. People who had spent years on these projects were, in Dirk's words, "still annoyed". But the framework gave the decisions a logic that was traceable and defensible. It allowed a cultural outsider to stop zombie projects that no insider had been willing to touch in two decades.

But Dirk didn't stop there. Of those 550 people originally assigned to his team, he recognized that the vast majority had never worked outside the company and had only ever worked on combustion engines components such as spark plugs and sensors. As a next step, he hired 50 outsiders, startup founders, entrepreneurs, and rebuilt the team by blending them with the insiders who had deep company knowledge and relationships. The combination gave him something neither group could have provided alone.

Then came the second problem. Leadership set annual revenue targets for his team. "Hit this number every year on your way to a billion," they said. Chasing yearly sales pushed the team toward whatever could generate quick cash: short-term wins that looked good on a dashboard but would never become sustainable businesses. "You basically end up working on NFTs," Dirk jokes.

So Dirk went back to leadership and proposed a different model. Instead of annual revenue, measure the team on the value of the businesses and investments they develop and successfully transfer back to headquarters. Leadership agreed. That single shift changed everything downstream: what areas the team explored, how they placed bets, and how they engaged with the rest of the company. The conversations with stakeholders moved from "justify your existence" to "here's something valuable we built for you."

Today, Dirk runs his portfolio the way he wished someone had run it before he arrived. He keeps a backlog of around 20 opportunity areas but only doubles down on two or three at a time. When one is either transferred to the business or killed, it frees up space for the next one. He measures how quickly things move through the pipeline, not how busy the team looks. And he keeps his team deliberately insulated from the core business, not isolated, but protected from having to justify every move daily, so they can actually build something worth presenting when the time comes.

Most innovation leaders will recognize pieces of Dirk's situation in their own day-to-day. The portfolio full of projects that exist because nobody stopped them. The metrics that accidentally reward the wrong behavior. The gap between what leadership says they want and what the system actually incentivizes. What Dirk's experience makes concrete is that you can't fix any of these standalone.

The methodology, the metrics, the team composition, the relationship to the core business, the way leadership measures progress: they're all one system.

And in this case, it took an outsider’s fresh perspective to change it.

For more insights about how the role of the innovation function and the nature of innovation management are changing, check these curated resources.

That’s it for today.

Hope you enjoyed this third edition of the Compass. Next time, we’ll have a look behind the scenes of a long standing successful employee-driven innovation program within a large bank. These types of programs can deliver solid results, if done well, as we’ll see.

Hans Balmaekers
Founder, the Compass and Chief @ Innov8rs

PS- best to move this newsletter to your primary inbox, or ‘whitelist’ our domain, to ensure we don’t end up in your promo or spam dungeons…

PPS- feel free to forward this newsletter to all the portfolio leads and Zombie Project owners in your company and network. Sharing is… indeed!