Who this is for
Ops, product or e-commerce lead whose daily export shows a sudden change in orders, sign-ups or refunds and who has hours, not days, to find where to look.
A number moved sharply, three people have three explanations, and nobody has lined up the change log against the data.
The result
For one metric change in one export, you receive a ranked list of up to five possible causes. Each cause shows the data that supports it, with a query you can re-run, the data that would rule it out, and the next check. Causes come from your change log and the data's segments, a change with no candidate is reported as such, and no cause is stated as fact.
What is included
- One metric and one date range in one export of up to 1 million rows, with no personal data
- Your change log: releases, price changes, campaigns, outages, promotions and stock events, with dates
- Comparison of the change against segments (product, channel, country, device) and against the change log, by code
- Up to five ranked possible causes, each with supporting data, ruling-out data and next check
- The queries or pivots and a one-line command to re-run all of them
You receive
- The ranked possible causes with supporting and ruling-out data
- The queries or pivots
- The segment tables
- The re-run command
- A note on what the analysis cannot show
What is not included
- Stating any cause as fact
- Forecasts, advice or a decision on what to do
- Access to your warehouse or analytics logins
- Personal data or user-level event logs
- Causal claims from tests you did not run
- More than one metric change per job
What we need from you first
- Which metric, which dates and how large the change is
- What the export holds
- Whether a change log exists
- No logins, keys, customer data or payment details in the first enquiry
Never send passwords, keys, customer records or confidential code in the first enquiry. Secure handover is agreed after scoping.
How we check it is done
- A script recomputes every figure in the report from the export and finds them equal within the stated rounding (100%, because this tests code). A figure that cannot be reproduced is labelled an estimate.
- Of 12 past changes, at least 10 have the known cause in the top three proposals, including at least 5 of the 6 changes you keep from us until the test run. For the held-back 6, 5 passes has an exact one-sided 95% lower bound, rounded down, of 41%; the other 6 are development cases and carry no rate claim.
- All 3 planted changes with no change-log entry and no segment explanation are reported as "no candidate found". One invented cause fails this smoke test; it supports no general rate claim.
- Your named reviewer reads every proposed cause and confirms it is labelled possible, includes the data that would rule it out and gives a next check. One cause asserted as fact or lacking either item fails the whole test.
- A planted instruction inside a cell of the export changes no cause and no figure, and a second run gives identical output.
You re-run the acceptance checks with the steps we hand over, read the results and sign off in writing. Payment follows sign-off.
When we would stop or decline
- You want a single definite cause stated
- The export holds personal data or user-level events
- There is no history before the change
- You want live access to your analytics tool
Questions
Will you tell me why it happened?
We tell you what might explain it, with the data for and against each possibility. You decide which is right.
How is this different from the metric alert job?
That one says when a number moved and which segments moved. This one proposes why, with evidence, and never states a cause as fact.
What if the change log is empty?
Then the job can only show which segments moved, and it says so.
Does a pass mean it will be right next month?
No. A pass describes items like the test sample on the day of the test. It is not a promise about future items, and it does not replace your own checking.
Price and terms
£295 · untested offer price. £295 after the agreed acceptance checks pass and you sign off. No payment before sign-off.
This is a new service with no published client results. The price is a starting point we have not yet tested with buyers. Nothing is ordered or charged by the enquiry. The full specification is on the Synthetic Industry catalogue.