A metric dropped. Know where to look in minutes.
Tell Doe "activation dropped 12% this week" and it decomposes the metric across every dimension in your database until it identifies which segment is driving the decline. Every decomposition query is visible. You get the "where," your team investigates the "why."
When a metric moves unexpectedly, Doe queries your database and decomposes it across every available dimension: geography, plan type, acquisition channel, device, signup cohort. Connects to Neon, Supabase, and PlanetScale. It identifies which segment is driving the change so your team knows exactly where to investigate. Every decomposition query is visible and verifiable.
What changes
| Dimension | Before | With Doe |
|---|---|---|
| Time to isolate the segment | Half a day to a full day of manual investigation | Doe checks every dimension and reports which segment is driving the change |
| Dimensions checked | Whatever the analyst thinks to check, one at a time | Every available dimension in your schema tested |
| Traceability | An answer with no way to verify it | Every decomposition query and result attached to the findings |
| Analyst time | A full day on one question, everything else blocked | The segment identification is automated; the analyst focuses on the fix |
How Doe investigates metric changes
Doe identifies the metric, timeframe, and magnitude, then maps every available dimension in your schema for systematic testing.
SupabaseActivation rate computed by plan type, geography, channel, device, browser, and signup week. Each dimension tested for whether it explains the decline. SQL logged.
Doe pinpoints which segment accounts for most of the decline — or reports that the drop is broad-based. Intersections like "mobile AND free-tier" are tested too.
SlackDoe delivers the driving segment, its contribution magnitude, and when the drop started — with every decomposition query attached for verification.
"Activation is down 12%. Can you look into it?"
You open a notebook, pull activation by day to find when the drop started, then segment by plan type, geography, channel, device, browser, signup cohort. Each cut takes 20 minutes. Most show nothing. Four hours later you've narrowed it down: the drop is concentrated in free-tier signups from organic search on mobile.
That tells you where the problem is, but not what caused it. You still need to check deploy logs, talk to engineering, review recent changes. The data investigation alone took a full day. The actual fix often takes 30 minutes once someone knows where to look.
Get started in under 10 minutes
Connect your tools
One-click OAuth for each integration. No API keys, no engineering.
Describe what you need
“Daily active users dropped 12% this week. Check every dimension — plan tier, geography, device type, signup cohort — and tell me which segments drove the decline.”
It runs on schedule
On demand. Ask whenever a number moves and the answer lands in minutes.
Metric Root Cause Analysis FAQ
Doe identifies which segment is driving the metric change: which plan type, geography, channel, device, etc. This narrows the investigation from "why did activation drop?" to "why did activation drop for free-tier mobile users from organic search?" Your team then investigates the specific cause (a code change, a campaign, a third-party issue).
Stop doing the work your tools should do for you.
Set it up once. Doe runs it every time.
