Growth Loops · Session 06 · Dates TBD
Ship one measurable product growth loop
Ship AIPhoenix, ArizonaPress → to begin
A circle drawn around a funnel does not make it compound.
A loop earns the name when one turn creates an output that becomes a measurable input to another turn.
The previous five sessions built the measurement, first value, repeat value, retention, and re-engagement layers. Tonight connects one proven behavior into a repeatable mechanism.
Act one
Start with value already observed in the product. The loop should amplify that value exchange, not hide a weak funnel inside a circle.
Funnel
Growth loop
Funnels remain useful inside loops. They describe the conversions that determine how much of each turn survives.
The added context gives the original user a reason to return and can expose another teammate to the product. The value exchange, not the invite count, is the mechanism.
| Output that can re-enter | Necessary value | |
|---|---|---|
| Collaboration | A participating teammate | Shared work improves for both sides |
| Content | A useful public or indexed artifact | The artifact answers a real need |
| Data | Better structured state or recommendations | Use makes the next result better |
| Retention | Saved context that reduces future work | Returning is easier or more valuable |
| Marketplace | Supply or demand that attracts the other | Both sides complete the core exchange |
A mechanism fails when the output benefits the company but not the participant expected to create it.
Act two
Write the loop as a sequence of observable states, then attach conversion and elapsed time to each crossing.
If the output cannot be named, counted, and observed re-entering, the diagram is a narrative — not yet a model.
| Question | Illustrative measure | |
|---|---|---|
| Eligible input | How many can start? | Activated workspaces with a recurring job |
| Conversion | What share completes value? | Useful shared result per eligible workspace |
| Output | What does completion create? | Qualified teammate exposures |
| Re-entry | What share becomes a new input? | Exposed teammates who complete a value action |
| Cycle time | How long does one turn take? | Median time from share to teammate value |
One-turn output is the product of eligible input and each conversion through re-entry. Track volume and time separately so a slow loop is not mistaken for a weak one.
Exercise: trace a real turn
model input → value → output → re-entry
5 observable boundaries mappedconversion + cycle time attached2 assumptions marked unverifiedweakest crossing: recipient reaches first value
Choose one recent product example and follow it through every event. Record where the output re-entered, where it stopped, and which step lacks enough data to estimate honestly.
Do not fill missing cells with intuition. An unverified assumption is a useful result because it tells the experiment what to learn.
Weak conversion
Slow cycle
The same total output can come from a strong slow loop or a weak fast loop. The product intervention should match the actual constraint.
Act three
The smallest useful experiment changes one crossing, preserves the value exchange, and defines what evidence will earn more investment.
A loop metric can rise because low-quality outputs increased. Guardrails protect the value exchange on both sides of the turn.
| Question | Evidence | |
|---|---|---|
| Did the mechanism move? | Primary conversion or cycle time | Compared on the defined audience |
| Why did it move? | Diagnostic events | The intended path changed |
| Was value preserved? | Quality and trust guardrails | No hidden cost traded away |
| Did another input result? | Output observed at re-entry | The loop actually closed |
| What next? | Decision rule and caveats | Ship, revise, or stop with an owner |
A local metric win without observed re-entry may improve a funnel step without strengthening the loop.
Live build: close one instrumented turn
launch → replay turn → inspect events → decide
audience + assignment verifiedvalue, output, re-entry events matched the modelguardrail failure path testedwrote 06-growth-loop/experiment-readout.md
Ship the bounded intervention and complete the loop as every affected role. Inspect each event and property, then record the first ship, revise, or stop decision without extrapolating beyond the observation window.
A working prototype plus trustworthy events is the finish line tonight. Compounding is observed over future turns, not declared at launch.
A dashboard without an owner is storage. Put the decision and next evidence date in the same place as the metric.
| Artifact | Capability | |
|---|---|---|
| 01 · Measurement | Events and scorecard | Trust the journey data |
| 02–03 · Value | Activation and engagement | Improve first and repeat value |
| 04–05 · Durability | Retention, risk, re-engagement | Find decay and earn a return |
| 06 · Loop | Mechanism and experiment | Decide from a closed turn |
The system is not the six dashboards. It is the shared definitions, validated events, bounded experiments, and decisions they support.
A loop compounds only after the output comes back.
Next
The deck, experiment contract, and 30-day cadence live on the session page. Keep the definitions versioned, the events validated, and the ship–revise–stop decision attached to every turn.
shipai.club · Free, always