The product flywheel

What the word is supposed to mean

A funnel takes money and attention in at the top and produces customers at the bottom. Want more customers next month, put more in at the top. The relationship is roughly linear and it stops the day you stop paying.

A flywheel is a loop where the output of one turn is the input to the next. More customers produce more reviews, more reviews produce more trust, more trust converts more of the same traffic into customers. Nothing extra went in at the top on turn two. The wheel is heavier and it holds its speed.

The reason the metaphor is worth keeping, despite how thoroughly it has been abused in pitch decks, is that it forces one specific question: what exactly did turn one leave behind that turn two gets to use? If the answer is nothing, you have a funnel with a nicer diagram.

The four questions

Name the number that carries over. Reviews, indexed pages, user-generated content, catalogue depth, integrations, referrals, trained data. It has to be a countable thing that exists at the end of turn one and is still there at the start of turn two. If you cannot name it, the loop is a story.

Know the turn time. A loop that completes in a week can be steered. A loop that completes in eighteen months is a strategy you will be judged on long before it can defend itself, and it needs a leading indicator standing in for it in the meantime.

Check that friction falls. The second hundred reviews should be easier to get than the first hundred, because there is now social proof, a habit, and a place to ask. If every turn costs the same effort, the wheel is not accumulating anything.

Check that it does not need proportional spend. This is the sharp edge. Paid acquisition producing customers who produce reviews that improve paid conversion is a real loop, because the improvement persists after the ad stops. Paid acquisition producing customers who churn in a month is a treadmill with good branding.

Loops that actually turn

Content into search into content. Publishing produces indexed pages, indexed pages produce traffic, traffic produces both the evidence of what to write next and the authority that makes the next page rank faster. This one is now double-sided: the same pages are what answer engines read when somebody asks the model instead of the search box, which is a ranking most companies are not checking.

Customers into reviews into conversion. The oldest one in commerce, and still the strongest, because it moves the conversion rate on traffic you already pay for.

Usage into data into a better product. Real when the data measurably improves the thing the customer came for, and a fiction when the data merely accumulates in a warehouse.

Users into invitations into users. The fastest loop when it works, and the one most often assumed rather than measured. It needs a genuine reason for one user to want another one there.

Why most flywheels stall in the same place

They stall at the handoff, not inside a stage.

The customers arrive and nobody asks them for a review. The reviews arrive and never appear on the page where the decision is made. The content is published and nothing links to it, so search never sees it. Each stage is fine on its own, and the loop is broken at the joint, which is exactly the part no team owns.

The other failure is subtler: a loop can be real and still be too slow to survive. If it turns twice a year, nobody can tell whether it is working, so it is judged on the outcome number instead, and quietly defunded in the second quarter. The fix is to instrument the handoffs and watch those weekly, because a leading indicator is the only honest weekly reading of a slow loop.

You can see somebody else's wheel from outside

This is the part that is genuinely useful about watching a competitor, as opposed to admiring them.

Review count is public and so is its growth rate. The number of indexed pages is public. Catalogue depth, integration directories, changelog frequency, the presence of a referral mechanism, whether their product asks for user-generated content at all: public, all of it, and each one is a stage in a loop somebody designed on purpose.

Read those over months and the wheel becomes visible. A review count climbing steadily while the catalogue holds still is a company whose loop runs through trust. A catalogue growing while reviews flatten is a company whose loop runs through range, and whose conversion problem is coming. The Benchmark holds the levels and Changes holds the rate, and the rate is the half that tells you whether the wheel is turning or just heavy.