The Method
Everything on this site rests on one question: can you trace this number back to something real? This section is the answer — how we measure, how we calibrate, how we record what we did, and what happens when we get it wrong.
What this section is
The method pages stand independently of any site measurement. A publication that grades things needs to hand you its rubric first, so you can judge the standard independently of whether you like a result.
None of it is complicated. It is a probe, a log, a stated set of conditions and a willingness to write down the batches that failed. It is only unusual because almost nobody in food publishing bothers.
The four rules
1. Measured claims need a test log. A page reports a measured result only with a test log naming who cooked it, when, on what equipment, what the probe read and how many times it was repeated. Without that log, the page stays a sourced method or a reader-run check.
2. No invented numbers, ever. A temperature is measured with a calibrated probe, or cited to a named source with a date. Never estimated, never interpolated, never carried across from a similar dish because it "should be about the same."
3. We publish the failure point. Every dish carries a doneness window in degrees and the temperature at which it stops working, along with what specifically fails. Finding that number means deliberately ruining a batch, which is part of the job.
4. Real photographs only. The photograph is the evidence. No AI-generated food imagery, no stock photography standing in for a result, and failure photographs published alongside the successful ones.
Provenance: the label on every number
Every figure on the site — in a Heat Spec, in a reference table, in a comparison — carries one of three labels.
| Label | What it means |
|---|---|
| Measured | We cooked it and read a calibrated instrument. The entry carries the tester's name, the date, the equipment, the ambient conditions and the number of runs. |
| Cited | The figure comes from a named external source with the date we retrieved it. It is styled differently from our own measurements so the two are never confused. |
| Reader check | The row is an observation prompt. Record your own ingredients, equipment, temperatures, and endpoint; do not read it as a site result. |
This is meant to be enforced mechanically rather than by editorial goodwill: a page with a number and no provenance record should not publish.
Where the method is written down
- How we test — the full protocol: instruments, calibration schedule, replicate counts, ambient logging, probe placement, and the log format we fill in for every run.
- Calibrate a thermometer — the ice-point and boiling-point checks, the altitude correction, and what to do about an offset you cannot adjust out.
- Test thermometers against ice and steam — how to check several probes against the ice point, against boiling water and against each other before you trust any of them.
- Corrections and failed runs — a number that changes is corrected on the page that carries it, with the date and the reason; nothing is silently edited. A batch that went wrong is logged with its diagnosis rather than discarded, because a site that only shows successes is not showing you its method.
Where the method does not apply
Food safety is the one area where we do not measure and do not conclude. Pasteurisation time and temperature, holding requirements, cooling rules and storage durations come from named authorities, quoted with retrieval dates on the page that uses them. We do not round them, average them, interpolate between them, or simplify a published table into a rule of thumb. Home canning and shelf-stable preserving are out of scope entirely.
What we do not know
Tables labelled Reader check are worksheets, not site measurements. A measured claim requires a dated log, named cook, instrument, endpoint, and replicate count. Otherwise the page cites an authority or makes no quantitative claim.
Where our sample is small — one oven, one hob, one pan — the page says so. Nothing here is offered as a universal constant.
Questions
Why publish the method before the numbers? Because a site that grades things should publish its rubric before it grades anything. If the protocol only appeared after the results, you would have no way of knowing whether it was written to fit them.
How many times do you cook something before publishing? As many as it takes to see whether the result is stable, and we publish the count. Some entries will be a single run, and when that is the case the page says so rather than implying more. A single honest run beats an unstated one.
What happens when two credible sources disagree? We publish both, with citations, and say which one our own measurement matched if we have taken one. Reconciling conflicting sources into a single tidy number would be inventing a figure that nobody actually published.
Do you use averages? Only where the underlying runs are published alongside. We never average across different equipment, different masses or different sources and present the result as a single fact.
What about food-safety figures? Those are never ours. Pasteurisation times, holding requirements, cooling rules and storage durations are reproduced from named authorities with retrieval dates, or they are omitted. We do not round them, interpolate between them, or simplify a table into a rule of thumb.
Is AI used to write any of this? It is permitted for drafting and editing, and forbidden for anything that would fabricate evidence: no generated temperatures, no generated photographs, no invented sources, no invented personas. A named human takes responsibility for every published page.
What if you get something wrong? The number is corrected on the page that carries it, with the date and an explanation in a visible note, and the retest is logged like any other run. Nothing is silently edited.
Run the method yourself
The test log fields are published in full on the protocol page, and the calibration procedure takes two minutes. Run the protocol in your own kitchen and record your conditions before you cook; your log is what makes your numbers comparable with anyone else's.