LLenny's Podcast
← All frameworks
ProductivityBangaly Kaba (YouTube, Instagram, Facebook, Instacart)

Understand Work (Understand → Identify → Execute)

Budget intentional time to understand a problem first, so your team ships fewer things with a far higher win rate.

Difficulty
Moderate
Time to result
~months to results
Steps
4
Confidence
97%

A planning discipline that inverts the common anti-pattern of jumping straight to building. Instead of Identify → Justify → Execute (someone proposes an idea, you pull data to justify it, then sink time into it), you run Understand → Identify → Execute: first understand from first principles what's actually going on. 'Understand work' is placed on the roadmap as an explicit, owned deliverable, run in parallel with execution so learnings compound into a velocity multiplier.

Origin

Kaba credits this to an old Facebook framework ('understand, identify, execute'); he does not take credit for it, describing himself as its shepherd — carrying it to Instagram, Instacart, Twitter and YouTube.

Core principles

  • 01The anti-pattern is identify-justify-execute: building something you feel is good and back-filling justification.
  • 02Understand work must be an intentional affordance — a named roadmap item with an owner — or it never happens.
  • 03Every function (PM, eng, data science, marketing, go-to-market) can and should do understand work.
  • 04Run understand work in parallel with execution so each sprint's learnings feed the next.
  • 05The mix shifts over time: more understand work when the space is unknown, less as confidence grows.
  • 06Slow down to speed up — fewer, better-understood ships beat high-volume, low-impact ones.

How to run it

  1. 1

    Separate what you know from what you must understand

    For each theme on the roadmap, honestly split it into what the data gives you confidence on versus what you're missing (research, instrumentation, insights, strategy).

    Pro tip Ask, in the planning session itself, 'what else do we need to understand to make this happen?' and include cross-functional partners so you see the real issues.

  2. 2

    Put understand work on the roadmap as owned items

    Write it down as an actual deliverable assigned to a person and function — a PM writing a strategy, a designer prototyping, data science pulling a funnel, engineering instrumenting logging — not something assumed to happen in the background.

    Pro tip Cap it: pick your top three or four understand projects for the roadmap so it doesn't sprawl.

    Watch out Without an affordance, everyone just sprints on execution and the understanding never gets done.

  3. 3

    Run understand work in parallel with execution

    Each sprint, execute on high-conviction, low-to-medium-effort/high-impact items while simultaneously running understand work on the unknowns. Sometimes understand work is a cheap test designed to learn as fast as possible.

    Pro tip Always ship something — maintain a portfolio of low-effort/high-impact and medium-effort/high-impact bets alongside the learning.

    Watch out Guard against endless understanding — the goal is still to ship; manage expectations that you ship the things you have more confidence in.

  4. 4

    Feed learnings into the next plan and let velocity compound

    At sprint end, combine what your launches taught you with insights from the understand work to plan the next sprint. Over successive sprints this parallel path produces more and better execution.

    Pro tip Tune the ratio to the team's understanding — e.g. 60/40 execution/understand when new to a space, shifting to ~80-85/15-20 as clarity grows.

In the wild

Instagram signup funnel had no logging

When Kaba joined Instagram in January 2016, the signup flow logged only how many people started and finished — eight steps in between were dark. The first understand work was instrumenting that funnel as fast as possible while simultaneously running tests on obviously-broken parts of the experience.

With growth marketing defining schemas, engineering adding logging, and data science building funnels, they could see where drops happened and fix them — a template he reused across teams.

YouTube live paid virtual goods

On joining, the team didn't fully understand the live ecosystem. Instead of shipping iterations to the current product, they treated mapping the full funnel (watchers, click-throughs, buyers) as multi-team understand work across engineering, data science, and PM.

Understanding where the gaps and opportunities actually were became the prerequisite for high-win-rate execution.

Common mistakes

Identify → Justify → Execute

Someone says 'this would be great to build,' you pull data to justify it, then sink ungodly time into making it work — and it launches flat because you never understood the real pain points, alternatives, or needed experience.

Assuming understanding happens in the background

If understand work isn't an explicit, owned roadmap item, the team just sprints on execution and the learning never lands.

Is it for you?

Best for

Product and growth leaders taking over a team or ambiguous problem space who want a higher ship win-rate rather than raw output.

Not ideal for

Well-understood, high-confidence problem spaces where the data is already clear and endless analysis would just delay obvious execution.

From the transcript

understand identify execute right so first you have to really understand from first principles what is actually going on

27:30

I call that identify justify execute right where you identify something someone says hey you know what this would be great to build and you…

27:00

it's a planned intentional time to the team

28:00

slow down and speed up

34:30

From the episode

Unorthodox frameworks for growing your product, career, and impact

Bangaly Kaba (YouTube, Instagram, Facebook, Instacart)