Methodology
Proved by the next reading
A recommendation is a hypothesis until something measures it. The second cycle re-reads the same store on the same rules against the baseline the first one recorded — and reports what moved, what did not, and what is still uncertain.
Operator-led · One cycle at a time · Nothing runs between cycles
The cycle
Six steps, repeated
- Diagnose
- Act
- Re-diagnose at ~30 days
- Verify
- Next constraint
- Act again
Each cycle is a separate, purchased diagnostic. Nothing runs in between: the 30-day point is when a second reading is taken, not a process that has been watching since the first.
What the second reading does
Five questions it answers
What was measured initially
The first Blueprint fixes the figures in writing — the margin, the refunds, the tax treatment, the product-level economics — together with the evidence classification behind each one. That record is the baseline. Without it there is nothing to verify against, and any later claim of improvement is an assertion.
What action the operator chose
The Blueprint names the binding constraint and what would address it. Which action is taken, and whether one is taken at all, is the merchant's decision. The re-diagnosis records what was actually done, because a figure that did not move because nothing changed is a different finding from one that did not move despite a change.
What should be measured again
The same figures, computed the same way, over a comparable period. Changing the definition between readings makes the comparison meaningless, so the second cycle applies the first cycle's rules even where a better rule now exists.
What changed
The difference is reported against the baseline, with the direction and the scope it holds over. If the figure did not move, the re-diagnosis says so plainly. A change that cannot be demonstrated is not written up as a win.
What remains uncertain
Some movements have more than one plausible explanation, and observational data cannot separate them. Where that is the case the report says which alternatives remain open rather than picking the flattering one.
The limit
Thirty days is a reading, not a promise
The 30-day point is when the store is read again. It is not a claim that every intervention produces a measurable result within exactly 30 days, and the method does not treat it as one.
Some changes surface almost immediately — a tagging fix that restores attribution coverage shows up within days. Others are portfolio averages that need a comparable trading period before a movement can be distinguished from ordinary variation, and a supplier renegotiation may not reach the orders at all until existing stock is sold through. Where the window is too short to support a conclusion, the re-diagnosis reports the figure as not yet established rather than reporting noise as a result.
Two other things can make a second reading uninterpretable, and both are checked before any comparison is drawn. If coverage moved between the readings, a margin can change purely because a different set of products became costed. And if several changes were made at once, the reading shows the combined effect and cannot apportion it — which is why the Blueprint names one constraint rather than a list.
What this is not
Autopsify does not act, monitor or test
Three limits, stated plainly, because the distinction is easy to blur:
Autopsify does not execute the action
Access to Shopify is read-only. Pricing, stock, cost data, supplier terms and the storefront are changed by the merchant, in the merchant's store.
Autopsify does not continuously monitor
There is no always-on watching, no alerting and nothing running between cycles. A second reading happens because a second cycle was commissioned.
Autopsify does not run experiments
There is no split testing and no automated test framework. Verification here means re-measuring the same business on the same rules — an observational comparison, not a controlled experiment, and the method does not claim the causal certainty a controlled experiment would give.
The commercial shape of this — what a cycle includes and what it costs — is on the verification page. The classifications that make two readings comparable are in the evidence model.
Afterwards
The next constraint
Clearing one constraint usually moves the binding one somewhere else. A store that fixes a unit-cost problem may find acquisition efficiency is now what limits it; one that fixes attribution coverage may find the channel figures it can finally see point somewhere unexpected.
So the next cycle starts from the store as it now is, not from the previous report. That is the sense in which the method is a cycle rather than a one-off audit: the output of verification is the input to the next diagnosis.
Methodology
The rest of the method
Start the first cycle
£497 for one diagnostic cycle. The re-diagnosis is a second cycle you choose to buy once you have acted — never an automatic charge.