In this article

Suppose 1,000 active accounts each share a project with two new recipients. Fifteen percent of those recipients sign up, and 40% of the resulting accounts become active. Under these illustrative assumptions, the sharing produces 120 active accounts.

That is a useful acquisition contribution. It is also a loop yield of 0.12 new active accounts per original account, before duplication or other acquisition influences. Drawing an arrow back to the beginning does not make those 120 accounts replace the original 1,000.

A growth loop earns its name through a repeatable mechanism: an output from one cycle becomes an input to another. To judge its contribution, you need to follow that return path beyond the first attractive number.

Follow the output into another person’s decision

For a collaboration product, the proposed mechanism might be that customers create useful projects, share them with colleagues, and expose those colleagues to a reason to use the product for their own work. Some recipients then create projects and share again.

Every transition needs a reason. The original customer shares because collaboration helps finish the project. The recipient engages because the shared work matters to them. Creating a separate account or project must serve another need. Accessing a colleague’s project does not automatically supply that need.

This is where many proposed loops remain incomplete. An invitation can create an account without creating a future creator. A public page can attract readers who never publish. Those people may still contribute value, but the original mechanism must account for what they actually do next.

Brian Balfour’s essay on launching for growth argues for a continuing growth mechanism beyond the launch. Applied to a sharing feature, the practical question is what keeps generating useful exposure after the announcement and incentives have passed.

Trace one cycle in verbs and name its output. Then write the first action of the next cycle. If the connection depends on an unexplained leap from “views” to “more growth,” that is the uncertainty to investigate.

Keep the funnel inside the loop

A funnel can show where recipients fail to reach the next meaningful state. It remains useful even when the broader system repeats.

In the opening example, the recipient funnel is straightforward:

Illustrative stageCountConversion from the previous stage
New recipients of a shared project2,000—
Recipients creating an eligible account30015%
New accounts becoming active12040%

The loop asks where those recipients came from and whether the 120 accounts generate another wave. The funnel locates losses within that journey. The two views support different questions about the same mechanism.

Use a consistent unit. If the output is accounts, several invited people joining one existing account should not become several new accounts in the loop calculation. Likewise, repeated invitations to the same recipient do not create new reach. Deduplicate at the unit that matters to the model and state what you cannot identify reliably.

Define activation by the contribution required for the next cycle. A recipient who reads a shared project has reached one kind of value. A creator who produces and shares a new project has supplied the mechanism with another output. Reporting both stages is more informative than giving them the same active-user label.

Calculate another generation

If each of the 120 new accounts produces the same first-cycle yield, the next generation would contain 14.4 active accounts in expectation. After that, about 1.7. These are model outputs, not forecasts or fractions of literal customers.

In this simplified one-generation-per-account model, the induced generations shrink because the yield is below one. A loop can still reduce the external acquisition needed to grow. It need not sustain the whole company unaided to be worth improving.

Be careful with the model’s boundary. Existing users may share repeatedly over their lifetime. A first-month yield measures a different contribution from a lifetime yield. If repeated sharing is central to the mechanism, model the later opportunities rather than assuming every account produces only once. Avoid counting the same recipient in several cycles.

Yields above one deserve equal skepticism. Early participants may have unusually large networks. Incentives may attract people who do not repeat the behavior. A small community can exhaust its reachable audience quickly. Observe later generations before extending an early coefficient indefinitely.

Put time and effort on the arrows

A cycle that completes in a day behaves differently from one that takes a quarter. Record the lag between useful output, recipient exposure, activation, and subsequent contribution. A launch cohort can look promising long before enough time has passed to observe repetition.

Separate waiting from missing behavior. A quarterly planning tool may require another planning cycle before its recipients share their own work. A daily collaboration tool with no second-generation output has a different problem. The product’s cadence determines when the evidence can mature.

Also name the work that keeps the mechanism running. If employees create every shareable asset and persuade every recipient to participate, the model includes an operating expense and a capacity limit. That can be a viable acquisition method; it should not be presented as effortless customer-driven distribution.

Balfour’s Universal Growth Loop describes reinforcement between growth, resources, and the ability to solve further problems. It also notes that such systems can reverse. That broader perspective helps expose dependencies a customer-sharing diagram may leave out: revenue must fund the work, and the work must generate a useful next cycle.

Check whether the loop adds growth

Tracking a shared link to a signup establishes a path. It does not prove the person would otherwise have remained unaware of the product. They may already be evaluating it or responding to another campaign.

Use a suitable comparison when you need to estimate the incremental effect of a sharing intervention. In a collaborative system, consider how assignment affects connected users: recipients can encounter links from several accounts, and treatment can cross the boundaries of a simple user-level test. Where the design cannot isolate the effect credibly, describe the result as attributed acquisition and retain the uncertainty.

Measure the cost imposed on existing customers too. More invitations may reflect an intrusive prompt rather than better collaboration. A change that produces extra signups while making sharing less useful can damage the behavior supplying the loop.

Once the mechanism is explicit, the next product decision becomes narrower. If people share but recipients cannot understand the artifact, improve that experience. If recipients get value but never become creators, investigate whether creation serves their work at all. A larger share button cannot repair a missing reason for the next generation to exist.

Sources

Something we should clarify?

Suggest a correction