The Hierarchy of Engagement
Three stacked levels — core action, accruing benefits, self-perpetuation — that decide if a consumer product endures
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 95%
A three-level pyramid for building enduring consumer (especially consumer social) products. Level one: get users completing a single defined core action. Level two: make the product get better the more it is used, so users have more to lose by leaving. Level three: make the product self-perpetuating via network effects, growth loops and re-engagement loops. Each level must be earned before the next one is chased, and the framework is a prioritisation lens for the roadmap, not a metrics dashboard.
Origin
Created by Sarah Tavel, first PM at Pinterest (where she led the discovery team), after moving into venture at Greylock. She kept meeting consumer founders presenting up-and-to-the-right signup, download and MAU charts — vanity metrics that said nothing about whether an enduring product was being built. The irritation became 'a pebble in my shoe' and she wrote and distilled it into the hierarchy. The Level 3 loop vocabulary overlaps with the growth-loop literature; the YouTube core-action correction (subscribing, not watching) came from Shishir Mehrotra, YouTube's then-CPO, who wrote to her after the original post.
Core principles
- 01MAUs, signups and downloads are vanity metrics — they do not test whether the product endures.
- 02There is almost always ONE core action that both proves the user understands the product and predicts their return.
- 03A retentive product gets better the more you use it, and gives the user more to lose by leaving.
- 04The core action itself must be the thing that makes the experience better over time.
- 05Level 3 (self-perpetuation) is what makes cheap, organic growth possible; almost no consumer product can buy its way to mass scale.
- 06Persistent identity is a precondition for accruing benefits and mounting loss — pure anonymity blocks Level 2.
- 07The levels build on each other: do not chase virality before retention, or retention before the core action.
How to run it
- 1
Define and instrument the core action
Pick the single action that proves comprehension and predicts return — pinning for Pinterest, sending a snap for Snapchat, sending a message for WhatsApp, writing a note for Evernote. Express it as a metric over a real cadence (Pinterest used weekly active pinners) and make it the only thing new features are judged against.
Pro tip Choose an action that scales to enough users and that you would be happy to have your whole roadmap optimise for — ask 'if I optimise for this, what do we end up building?'
Watch out Measuring follows, likes, click-throughs and time on site all at once with no chosen north star means experiments produce contradictory data and no decision.
- 2
Rebuild activation around the core action
Treat new-user experience success as: did this cohort come to understand the mental model well enough to complete the core action? Strip the new-user surface back so the one thing you want them to do is unmissable.
Pro tip Judge the NUX by core-action completion rate, not by signup completion.
Watch out Founders copy the fully-expressed product they aspire to (Instagram, TikTok, Roblox) and give new users far too much to take in — Tavel calls this the first and possibly largest mistake she sees.
- 3
Engineer accruing benefits and mounting loss (Level 2)
Make each core action feed back into the user's own experience so the product improves with use, and so leaving costs them something. Pinterest's Picked-For-You feed turned each pin into personalisation; each board became a repository of identity the user would not abandon.
Pro tip Ask two questions of every feature: does this make the product better the more it is used, and does it increase what the user loses by leaving?
Watch out Anonymity kills this level — with no persistent identity there is nothing to accrue and nothing to lose, so users can churn and return three months later to an identical experience.
- 4
Verify retention with cohorts before spending anything
Track weekly cohorts on two axes: the percentage of each vintage still completing the core action over time, and the activity level within those cohorts. You want the classic smile curve, not a leaky bucket. Until cohorts plateau, retention work is not finished.
Pro tip TikTok did not put serious user-acquisition spend behind the product until the first two levels were solid; only then did the roughly billion dollars of paid growth make sense.
Watch out Founders routinely present growth charts having never built a cohort table. Growth without a plateau in the cohorts is a bucket you are refilling.
- 5
Convert user energy back into the product (Level 3)
Treat every tap and click as kinetic energy and convert as much of it as possible back into the experience. Maximise the network effect first (each Pinterest pin creates a new edge in the graph that improves recommendations for everyone), then identify and tune the growth loops (sharing, metadata for SEO, collaborative experiences) and re-engagement loops (Pinterest notifying you that someone repinned your pin).
Pro tip The strongest re-engagement loops are ones the company does not have to manufacture — a user's action produces the notification that pulls another user back.
Watch out Loops that depend on push notifications can invert at scale: as House Party and Clubhouse users followed more people, notification volume became overwhelming, users tuned out, and the flywheel that should have accelerated broke down.
In the wild
In 2011 it was not obvious whether Pinterest was a social network optimising for a follow graph, like Twitter and Instagram at the time. The team ran a bottoms-up analysis of every possible action and a top-down question about what Pinterest was actually for, and converged on pinning. The north star became weekly active pinners, and the Picked-For-You algorithmic home feed was built so that pinning made the product better with every use.
→ Every experiment could then be judged by one rule: if the percentage of people pinning or the number of pins per cohort did not go up, the experiment failed. Follows turned out not to matter much — few users actually followed each other.
Evernote is extremely retentive at the user level: thousands of documents, search that always finds what you need, so a lifetime user like Tavel cannot leave. But there is no loop where one person's use improves or attracts another's — collaborative journals were tried and did not work.
→ Evernote had to pay to acquire users and eventually tapped out. Level 2 without Level 3 means a sticky product with no cheap growth engine.
Tavel originally published that YouTube's core action was watching a video. Shishir Mehrotra, YouTube's CPO, corrected her: that was true early, but analysis showed subscribing was the real core action. Creators care about YouTube because subscribers are where their audience accrues; viewers come back because they subscribed to creators whose content resonates.
→ The core action served both sides of the network — a sign, Tavel says, that you have got it right — and it demonstrated that the core action can legitimately change as a company evolves.
Common mistakes
Optimising vanity metrics
Signups, downloads and MAUs go up and to the right without telling you whether anyone is engaged. Opening the app is not being a user; completing the core action is.
Shipping the fully-expressed product to new users
Founders build toward the mature product they admire, so a new user faces too many surfaces at once. Attention gets diffused away from the one action that most matters, and activation collapses.
Betting on pure anonymity
Talented founders keep hitting this wall. Pseudonymity (Twitter, Reddit) still builds a persistent identity with accruing benefits and mounting loss; pure anonymity gives neither, and invites the bad behaviour that degrades the community.
Growing without geographic or segment focus
One founder grew a dating/friend-making app through TikTok, which is not geographically constrained at all — user numbers rose but the product could never be retentive because the density that makes it work was never built anywhere.
Is it for you?
Best for
Consumer and consumer-social founders, product leaders and growth teams prioritising a roadmap when the product is live but growth is not compounding
Not ideal for
Enterprise/B2B software or niche products with no ambition of mass-market scale — the framework is most useful where social loops are possible; Evernote-style tools can only reach Level 2
From the transcript
“there're almost always is this action which I call the core action of that product that forms the foundation of the product”
“the test for me of whether you're building a product that has the ingredients to create a retentive product on a micro level just at…”
“the biggest thing that you can do is a network effect”
“but there's other Loops too that you have to identify and then maximize and these are the kind of growth and re-engagement Loops”
From the episode
The hierarchy of engagement
Sarah Tavel (Benchmark, Greylock, Pinterest)