Why Some Innovation Recipes Explode (and Others Fizzle)

Every team has run a workshop that felt productive in the room and evaporated by Monday. Every team has also experienced that rare sequence where one activity fed perfectly into the next, momentum built, and by the end of the week something real existed that hadn't before.

The difference is usually not the methods but the bonds between them.

This is the third post in our series on the chemistry of innovation. The first argued that methods work best in sequences, not isolation. The second explained how to read the Periodic Table of Innovation. This one gets practical: how to spot stable molecules, avoid unstable ones, and build your own.

Stable molecules: what makes them work

A stable molecule has three properties.

Each step transforms the input. The output of method A is meaningfully different from its input, and that transformed output is exactly what method B needs. In the sequence "Challenge Brief → Problem Interview → Solution Interview → Lean Canvas," each method takes what came before and reshapes it. The Challenge Brief produces a focused problem statement. The Problem Interview turns that statement into evidence about whether real people experience the problem. The Solution Interview tests whether they'd use a proposed fix. The Lean Canvas captures all of that as a structured business hypothesis. Nothing is repeated. Each step adds something new.

The difficulty escalates gradually. Good molecules start simple and get harder. "Facts, Assumptions, and Doubts → Problem Reframing → How Might We" takes a team from a low-stakes sorting exercise (what do we know, what do we assume, what's unknown?) into a more demanding reframing exercise, and then into creative generation. By the time you're writing How Might We questions, you've built the context and confidence to do it well. Start with How Might We on a cold team and you get generic, surface-level questions.

The molecule ends with a decision or an artefact. Every good sequence produces something concrete at the finish. A validated hypothesis. A tested prototype. A prioritised backlog. A growth model. If your molecule trails off without a clear endpoint, it's not a molecule, it's a to-do list.

Unstable molecules: how to spot them

Unstable molecules look reasonable on a whiteboard but fall apart in practice. Here are the patterns.

The gap bond

This is the most common failure. Two methods sit next to each other in a sequence, but the output of the first doesn't connect to the input of the second. There's a gap in the logic.

Example: "Brainstorming → Development Sprints." You've generated a pile of ideas and now you're... building one? Which one? Based on what? The gap between "we have ideas" and "we're writing code" is enormous. You need filtering (DFV Matrix), specification (MVP Specification or Product Requirements Document), and probably validation (Prototype Test Plan or Usability Testing) before you're ready to commit engineering effort.

When you spot a gap bond, the fix is almost always to insert an intermediate method that bridges the two. Brainstorming → DFV Matrix → MVP Specification → Development Sprints is a stable sequence. Each step narrows and sharpens.

The echo chamber

This is when you stack too many methods from the same family without ever crossing into another. Three research methods in a row. Two canvases back to back. A strategy session followed by another strategy session.

Example: "Business Model Canvas → Value Proposition Canvas → Lean Canvas." Each of these is a good method. But this sequence is three consecutive framing exercises with no testing in between. You're just rearranging the same assumptions on different templates. The molecule looks busy but produces no new information.

The fix is to interrupt the echo with a method from a different family. "Business Model Canvas → Problem Interview → Lean Canvas" breaks the pattern by inserting real customer data between two framing tools. Now the Lean Canvas is informed by evidence, not just by a reshuffled version of the Business Model Canvas.

The premature heavyweight

This is when the sequence jumps to an Advanced method before the foundation is laid.

Example: "Challenge Brief → Product/Market Fit." You've written a problem statement and now you're trying to prove product-market fit? You've skipped customer discovery, solution validation, prototyping, and probably an entire Build-Measure-Learn cycle. Product/Market Fit is an outcome, not a step you can shortcut to. It's the end of a long molecule, not the second element in a short one.

The fix is to respect the difficulty gradient. If an Advanced method appears early in your sequence, you've almost certainly skipped steps.

The kitchen sink

This is the opposite problem: a molecule with too many elements. Twelve methods in a row, covering everything from foresight to delivery. It looks comprehensive. In practice, it's a three-month programme disguised as a sequence, and no team will maintain coherence across that many steps.

Good molecules are typically three to five methods long. If you need more than that, you probably need two separate molecules with a clear decision point between them. Run the first molecule, make a decision based on what you learned, then start the second if needed.

Building your own molecules

Sometimes the pre-built sequences don't fit your situation. Here's how to construct your own.

Step 1: Name your start state and end state. Be specific. "We have a hunch about a customer problem" is a start state. "We have a validated MVP specification ready for engineering" is an end state. The clearer these are, the easier it is to pick methods.

Step 2: Pick your last method first. What's the method that produces your desired end state? If you need a validated MVP spec, your last method is MVP Specification. If you need a growth model, it's probably Building Your Flywheel or Pirate Metrics. Work backwards from the end.

Step 3: Ask what that method needs as input. MVP Specification needs a validated solution concept. What produces that? A Solution Interview, or a Usability Test on a prototype. Now you're building the chain.

Step 4: Keep working backwards until you reach your start state. At some point, the input required by a method will match what you already have. That's your first element. The chain from first to last is your molecule.

Step 5: Check for the failure patterns. Any gap bonds? Any echo chambers? Any premature heavyweights? Trim the kitchen sink if it's longer than five methods.

Molecules from the field

Here are a few sequences we've seen work consistently across different organisations and contexts, with notes on why the chemistry holds.

The stuck team unlocker: Problem Reframing → How Might We → Mind Mapping → Rapid Ideation

This works because stuck teams don't need more brainstorming. They need the problem reframed first. Problem Reframing shakes loose the fixed assumptions. How Might We converts the reframed problem into opportunity questions. Mind Mapping lets the team explore connections visually without the pressure of committing to ideas. Rapid Ideation then capitalises on the fresh perspective with a time-boxed generation burst. By the time you hit ideation, the team has genuinely new material to work with.

The full open innovation pipeline: Innovation Mapping → Crowdsourcing → Hackathons → Accelerator Program

This one's unusual because it's designed for organisations sourcing ideas externally. Innovation Mapping identifies the landscape and gaps. Crowdsourcing pulls in external ideas to fill those gaps. Hackathons compress the best ideas into working prototypes. The Accelerator Program gives the most promising prototypes a structured path to viability. Each step filters and intensifies. You start with hundreds of inputs and end with a handful of investment-ready ventures.

The metrics alignment molecule: One Metric That Matters → North Star Metric → Key Performance Indicators

This is a short, tight sequence for teams drowning in dashboards. It forces a brutal prioritisation: what is the one thing that matters right now? Then it connects that to the longer-term North Star. Only then do you build out the supporting KPIs. Most teams do this backwards (build a dashboard, then argue about what matters) and wonder why no one looks at the numbers.

The deep tech development path: Technology Readiness Levels → Prototype Test Plan → Pilot Canvas → Scaling Operations

This molecule exists because hardware and deep tech can't follow the same rapid iteration loop as software. TRLs give you a shared vocabulary for readiness. The Prototype Test Plan structures what you're testing at your current TRL. The Pilot Canvas designs a real-world trial. Scaling Operations maps what it takes to produce at volume. Each step respects the reality that physical things take longer and cost more to iterate than code.

The meta-lesson

The methods aren't the hard part. Picking the right method for a given situation is not that difficult once you understand the table. The hard part is the connections, knowing what should come before and after. Knowing when a sequence needs another element and when it needs fewer. Knowing which bonds are strong and which are just wishful thinking on a whiteboard.

That's what separates teams that run workshops from teams that ship. The workshops are the same, but the chemistry is different.

Explore the Periodic Table of Innovation to see all elements. Browse our Playlists for pre-built molecules.

Or pick a method, check its valence, and start building your own. Custom Playlists coming soon.

---

Ready to explore the elements? Browse the Periodic Table of Innovation to see all 70+ methods, filter by methodology family, and start building your own molecules.