Case studies

Case study · Nimbus, a company the team generated for demos

The apparent success

Almost every account sets up. Why do only a third subscribe?

Nimbus’s signups rose from the first half of to the second. More people trying the product: good news, on the face of it. So we followed them from setting up to paying.

The contradiction

But only a third of them subscribed.

Of accounts that started onboarding, subscribed, in the product’s own funnel: every account set up, explored the core features, reached “Onboarding Complete” and used the product within two weeks.

Each stage is counted on its own, not as one account’s path: that is why more used the product than completed onboarding.

The discovery

The leak is the same for every plan.

The biggest leak, as Flow finds it, is from using the product to subscribing: make that step ( of ). Split by plan, it is within five points for each: Pro , Enterprise , Starter .

The plans differ earlier: Enterprise reaches the later onboarding steps more often (, against Starter’s ).

The check

The rise in signups holds. The funnel needs careful reading.

Signups’ 95% interval runs from to , all of it above no change.

  • The stages are the product’s AI’s. Its “Onboarding Complete” is any step from the sixth on; only of accounts complete the step named Complete Onboarding.
  • Counted on its own. Each stage counts the accounts that meet it, whatever came before.
  • One of three flows. Of the AI’s three suggested flows, one measured against today’s date and one timed out.
  • Rates only, by plan. No counts or intervals, and accounts with no plan are left out.

And no person has checked what Signups counts.

On Nimbus’s Living Map, Signups sits two levels under New MRR. The machine matched what it counts, and no person has reviewed it.

What we’d decide

Proposed decision

Fix the step from using the product to paying once, for every plan, before buying more signups.

In Nimbus’s data, which the team generated for demos, signups grew, but fewer than half of the accounts that use the product go on to pay, whatever their plan.

Next: Find what engaged accounts that don’t subscribe have in common.

Provisional · further evidence required

The strategy, from what we know and what we don’t

  1. Find what engaged accounts that don’t subscribe have in common.

    Because: of the accounts that used the product didn’t subscribe, whatever their plan.

    Not known: What they share. These runs cut by plan only.

  2. Help Starter and Pro reach the later onboarding steps.

    Because: Enterprise reaches them of the time, Pro , Starter .

    Not known: Whether reaching them leads to paying: the funnel counts each stage on its own.

  3. Hold paid acquisition until its rise holds.

    Because: Paid signups rose but might be zero; organic’s holds.

    Not known: Whether paid signups pay as often as organic ones.

What would change it: If the leak differs by source or region, which these runs didn’t cut, fix it where it is worst first.

The full analysis the definition, the sources, the funnel, the split by plan, what ran, the query

The definition

Signups: COUNT(DISTINCT account_id) on dim_accounts, by the day each account was created. Matched by the machine; no person has reviewed it.

On Nimbus’s Living Map, Signups sits two levels under New MRR, past a step with no metric of its own.

The rise, source by source

SourceChange a quarter95% intervalOn the first half

Shares of the rise: . The parts add up to the rise.

The funnel

Flow Analysis, on the stages the product’s AI suggested (, entity ).

StageAccountsOf the firstThe AI’s condition
StepIn both stagesRate

Each stage counts the accounts that meet it, whatever came before; a step’s rate counts the accounts in both stages. Time in stage reads zero (the flow’s clock measures every stage from one event), so it isn’t used. The step named Complete Onboarding, counted with a stage we wrote in the product’s form: of .

The funnel by plan

Flow with Cohort: it runs in MindPalace’s engine, but isn’t on the app’s screens yet.

StepEnterpriseProStarter

Rates only: the product gives no counts or intervals here, and it leaves out accounts with no plan ( of signups).

The flows the AI suggested

What ran

Signups: KPI Drill-Down, then Deep Analysis in Trend mode, by quarter, January 2024 to December 2025: rows; the two agree quarter for quarter. Flow Analysis and Flow with Cohort: 11 November 2026, on the app with fixes not yet released. The product rates its confidence in the signups analysis . How they work, on Product.

What the product’s AI wrote, and why it isn’t evidence

  1. ""
  2. ""
  3. ""

Written by the product’s AI after the numbers and quoted as it wrote them: none of it is evidence.

Who decided what

No person has reviewed the definition the signups run used. The decision above is proposed, not taken. Nobody has acted on any of it: Nimbus is a company the team generated for demos, and none of this is a customer’s result.

About the data

Nimbus is a company the team generated for demos, so what holds here holds in its data, not in a market. Every figure is the product’s own output, saved by its runs, except the share of engaged accounts that didn’t subscribe, the share of signups with no plan and the one stage we wrote, which we worked out from them. How we check a case.

The query behind the signups

The engine wrote this from the definition. No AI touches it; the same question gives the same query.

, …  the joins for each factor follow:  lines in all

Design partners

Bring your own question.

Design partners take the question their teams keep asking through their own map, with us.

Become a design partner