TOOL

The Behavior Layer

AARRR tells you where the funnel is leaking. It never tells you why. This worksheet runs the why in order: five questions between the metric and the fix. Nothing you type leaves this page.

01

What changed?

McClure, D. (2007). Startup Metrics for Pirates: AARRR! Presented at Ignite Seattle, August 2007. Source
Funnel stage

Where in the relationship did the number move? If your work is not a funnel, pick the closest and keep going.

AcquisitionDid they arrive?SuggestsAdoption · GreenDot · prompt, then motivation

The number, the direction, the size, the window. No explanation yet.Example: Activation fell from 41% to 33% over six weeks.

02

What kind of change?

Rodden, K., Hutchinson, H., & Fu, X. (2010). Measuring the User Experience on a Large Scale: User-Centered Metrics for Web Applications. Proceedings of ACM CHI 2010, 2395-2398. Source

What should be true for the person, written from their side rather than yours.Example: A new person reaches their first useful output in one sitting.

What you would observe if that were happening, or if it were not.Example: They finish workspace setup instead of abandoning partway through it.

The number that tracks that signal.Example: Setup completion rate, and time to first output.

Prior, a suggestionMost often at this stage: Adoption

03

What are we asking a person to do?

Fogg, B.J., & Hreha, J. (2010). Behavior Wizard: A Method for Matching Target Behaviors with Solutions. Persuasive Technology, PERSUASIVE 2010. Lecture Notes in Computer Science, vol 6137. Springer. https://doi.org/10.1007/978-3-642-13226-1_13 Source

Color is the kind of change you are asking for. Row is how long it has to last.Example: GreenSpan. Do a behavior that is new to you, for a period.

Green · newBlue · familiarPurple · increaseGray · decreaseBlack · stopDot · one timeSpan · a periodPath · from now on

All five columns, scrolls sideways

No behavior selected yet. Pick the cell that matches your ask.

Write it as the person would describe it, not as a feature name.Example: Finish setting up my workspace in one sitting, before I have seen anything useful.

Prior, a suggestionMost often at this stage: the typical ask is GreenDot

04

Why isn't it happening?

Fogg, B.J. (2009). A Behavior Model for Persuasive Design. Proceedings of Persuasive 2009. Source

Did anything tell them to act at the right moment? A screen they never reached does not count.

How much they want this at this moment, not in general.

How cheap it is in time, thought, and nerve. Count steps, decisions, and unknowns, not screens.

Example: motivation 70, ability 25. They wanted the outcome and the task was expensive.

Fogg describes the motivation-ability tradeoff qualitatively. He never published an equation for the action line. The curve here is a readable approximation, and the useful signal is which side you land on, not the exact distance.

Signal

Motivation and ability are both sufficient. Nothing told them to act. Use a signal: a plain reminder, placed where the person already is.

What you concluded, and why. This comes after the plot, not before.Example: Ability. Two fields were added to setup three weeks before the drop. Motivation was unchanged and the prompt still fired on schedule, so the cost of the task went up while the payoff did not move.

Prior, a suggestionMost often at this stage: Prompt, then motivation. Ability is rarely the constraint before signup.

05

What is the smallest move?

The smallest single move aimed at whatever you named above.Example: Default the two new setup fields, and let people edit them later.

What you expect to see if your diagnosis holds.Example: Setup completion returns toward 41% within two weeks with no drop in output quality.

The one that matters. If you cannot fill this in, go back to layer 04.Example: Completion recovers while quality falls, which would mean the fields were load-bearing and the answer is sequencing rather than removal.

Notes

Fogg renamed the third element of the Behavior Model from trigger to prompt in late 2017. This page uses prompt throughout.

Fogg's 2009 paper described 35 behavior types. The 15-type grid from the 2010 Behavior Wizard paper supersedes it, and this page uses the 15-type grid.