Play the Accordion
Zoom out to update the strategy, then compress the learning into the next release
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 99%
Play the Accordion alternates deliberate strategic expansion with focused execution. The team first stretches outward: it states how the system is believed to work, identifies the desired outcome, and considers downstream effects and the longer-term direction. It then compresses that thinking into the smallest version capable of testing the belief and ships it. Once evidence arrives, the team expands again, reassessing the whole system in light of what it learned before choosing the next compact release. Repeated expansion and compression combine the directional value of roadmapping with the learning advantage of experimentation. The model guards against two failure modes: locally optimizing one feature without understanding its wider impact, and committing to a detailed long-range plan that cannot adapt when reality changes.
Origin
Extracted from Lenny's Podcast, where Tom Verrilli used a piano accordion as a memory device for alternating between strategic thinking and iterative shipping.
Core principles
- 01Pair long-range direction with short learning cycles
- 02Use every release to update the model of the problem
- 03Build the smallest version that can test a strategic belief
- 04Inspect second-order effects before advancing
- 05Avoid both blind iteration and rigid roadmaps
How to run it
- 1
Stretch to the Outcome
Zoom out and state what the team is ultimately trying to accomplish. Describe the current model of how the relevant system creates that outcome.
- 2
Trace the Implications
Examine how the proposed direction affects customers, adjacent systems, economics, and future options. Look for consequences hidden outside the immediate workflow.
- 3
Compress Into V1
Choose the smallest release or experiment that can test the most important belief. Define what evidence would support or challenge the direction.
- 4
Ship and Observe
Release the version and collect behavioral, qualitative, and operational evidence. Compare the result with the expected mechanism.
- 5
Stretch Again
Zoom back out and update the strategic model using the new evidence. Decide whether the direction holds or the next version must change.
- 6
Compress Into the Next Version
Translate the updated model into the next focused release and repeat the cycle. Continue until the broader outcome is achieved or invalidated.
In the wild
Whatnot sellers can show an item on video without creating a listing, which saves preparation time. Zooming out reveals that missing listings weaken search and prevent buyers from finding relevant streams. Requiring full listings appears to solve discovery, but another zoom-out shows that several minutes of entry per item would reduce seller throughput. The accordion motion reframes the task as finding a version that improves discoverability without imposing the full listing burden.
→ The team searches for a system-level solution rather than optimizing either seller speed or buyer discovery in isolation.
A team believes a discovery change will improve a business outcome. It builds the smallest test, observes the result, and then returns to the broader causal model. If the evidence supports the belief, V2 advances the same plan; if it contradicts the belief, the team changes V3 rather than following a fixed roadmap.
→ Each release creates both customer value and updated strategic direction.
Common mistakes
Throwing Spaghetti Forward
Rapid releases without a directional model generate activity but do not reliably move toward a valuable outcome.
Freezing a Multi-Year Roadmap
Detailed long-range commitments discard the advantage of updating direction from experiments and customer behavior.
Zooming Out Without Shipping
Strategic expansion creates no customer value until it is compressed into a concrete release.
Is it for you?
Best for
It is best for teams that can ship incrementally and update their direction from customer behavior or experimental evidence.
Not ideal for
It is not ideal for work where versions cannot be released safely or where feedback arrives too late to guide iteration.
From the transcript
“the point of the analogy of play the accordions.”
“you've got to constantly be zooming out and pushing back in, zooming out and pushing back in.”
“I I think the honest answer is you've got to do both. You can't overindex on either.”
From the episode
This CPO regrets that product management exists