The Behavioral Diagnosis
A journey map on steroids: every micro-step to the behavior, with the blocking psychology named at each one.
- Difficulty
- Advanced
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 95%
Before designing any intervention, map every single step a user must take to complete the target behavior — often 50 steps at a high level, or a 200-300 slide screenshot deck at product level — then overlay the specific psychology operating at each step. The map is what tells you whether a feature can plausibly carry the weight of the behavior change, and where to intervene.
Origin
The core toolkit of Irrational Labs (Kristen Berman), drawn from behavioral science's insistence on measuring what people actually do rather than what they say they'll do. Distinct from a standard UX journey map by the psychology overlay and the granularity.
Core principles
- 01Study what people actually do, not what they say they will do.
- 02Any work you put on the user is a cost you must prove is worth their time.
- 03Behavior change is complex, hard, and noisy — you must understand the full step chain before trying to alter it.
- 04The diagnosis, not the brainstorm, is where the light bulbs happen.
How to run it
- 1
State the behavior you're diagnosing
Take the uncomfortably specific behavior from step one of the Three B's — e.g. 'reduce your monthly spend' or 'book a doctor's appointment during onboarding'. The diagnosis is always relative to one behavior.
- 2
Map the high-level step chain, including pre-product steps
Write out every step required to complete the behavior — including the ones that happen before the user ever touches your product. There may be 50. For budgeting: know your budget, know where in the month you are, make a plan, then execute it repeatedly across the month.
Pro tip This is the fastest way to kill a doomed roadmap item. Once you see how many steps budgeting requires, it becomes obvious it cannot move spend.
- 3
Build the screenshot-level map
For an existing product you're improving (rather than a new one), assemble a deck of 200-300 screenshots covering every screen and micro-decision on the path to the behavior. This takes a long time; it is the work.
Watch out Do not shortcut this to a tidy five-box journey diagram. The insight lives in the granularity.
- 4
Overlay the psychology at each step
Attach the specific psychology blocking or driving each step — e.g. in a health app, at the 'view test result' step the barrier is information aversion (people don't want to see the result).
Pro tip Once a psychology is attached to a step, the brainstorm becomes concrete: 'what would decrease information aversion at this exact screen?' rather than 'how do we improve engagement?'
- 5
Design the intervention against the named psychology
Only now design. Each intervention should target one identified psychology at one identified step, so it can be tested and attributed.
In the wild
A popular fintech app's most-requested feature was budgeting — it came through support forums and interviews, and felt like table stakes. Irrational Labs couldn't talk them out of building it, but got them to run a proper experiment with 10,000 users: a control (just showing people how much they spend) versus two different budgeting variants. Berman's team wasn't surprised by the result, because the behavioral diagnosis showed the number of steps budgeting demanded of the user.
→ Zero change in average spend and zero change in spend variability, after a large investment of time and engineering effort. The team was shocked; the diagnosis had predicted it.
Common mistakes
Building the most-requested feature
What users request is what they say they'll do, not what they do. Budgeting topped every support forum and interview, and still moved nothing. The diagnosis is the check against demand signals that won't survive contact with real effort.
Assuming user work is free
Any step you add is a cost. If the behavior requires the user to remember, plan, decide and repeat across a month, be skeptical by default and prove the feature is worth their time — then measure whether they actually do it.
Is it for you?
Best for
A product team about to build a heavily-requested feature, or one that has just seen a big investment produce a null result and doesn't know why.
Not ideal for
Fast, low-stakes iterations where the cost of just shipping and measuring is lower than the cost of the mapping exercise.
From the transcript
“is what we call a behavioral diagnosis where you map out every single step so you'd have to like know what your budget is you'd…”
“once you do that behavioral diagnosis and you map out you try to understand what people actually do versus what they say they will do”
“the second would be that behavioral diagnosis and this is the toolkit in Behavioral Science where we study not what people say they will do…”
“and improving the current product we'll do a deck of 200 or 300 slides of screenshots where it's just very detailed analysis of the steps…”
“zero change in average spend zero change in spend variability”
From the episode
Using behavioral science to improve your product
Kristen Berman (Irrational Labs)