Gelato Batch Log: Record-Keeping for a Consistent Lab
Table of contents
Every gelateria that makes consistently good product keeps records, and almost every one that cannot explain a bad batch does not. The batch log is the cheapest instrument in the lab. It costs a notebook and ninety seconds per production run, and it is the only thing that turns yesterday into evidence.
Ninety seconds per run, written at the machine, not remembered at the end of the day.
Why the log exists at all
There are two reasons and they pull in the same direction.
The first is legal. Record-keeping is not an optional refinement of a food safety system; it is one of its founding principles. The Codex Alimentarius HACCP guidelines set out seven principles, and the seventh is to establish documentation and record-keeping appropriate to the nature and size of the operation. In the European Union, Regulation (EC) No 852/2004 requires food business operators to put in place procedures based on HACCP principles, and Regulation (EC) No 178/2002 adds traceability in Article 18: operators must be able to identify who supplied them and, in turn, the businesses they supplied. That is the one-step-back, one-step-forward rule, and a batch log is where it lives for anything you made yourself. Operators in the United States face a comparable documentation regime under 21 CFR Part 117.
The second reason is that you cannot improve what you did not record. A complaint about a grainy pistachio three days ago is unanswerable without a log and nearly trivial with one. You look up the batch, see that the ageing time was four hours rather than the usual twelve, and you have both a cause and a fix. Without the entry, you are left adjusting a recipe that was never the problem.
Quick reference. The log has to be written at the machine, in the moment, by the person doing the work. A log reconstructed at closing time is a memory exercise, and memory is exactly the thing the log exists to replace.
Figure 1. Three blocks of fields. Identity makes a batch findable, process explains it, outcome makes it comparable.
The minimum viable row
Resist the urge to design something comprehensive. A log with forty fields is abandoned in a fortnight. The following fields fit on one line and answer nearly every question you will actually ask.
| Field | Why it earns its place |
|---|---|
| Batch code and date | The handle everything else hangs from |
| Flavour and recipe version | Distinguishes a recipe change from a process change |
| Batch size | Makes deviations comparable across runs |
| Lot numbers of key ingredients | The whole of your one-step-back traceability |
| Pasteurisation profile | Temperature reached and time held |
| Ageing start and end | The single most commonly skipped step |
| Draw temperature | What the machine actually delivered |
| Overrun or finished weight per pan | Your only in-line texture number |
| Operator initials | Not blame, context |
| Free-text note | Where the useful information usually is |
Batch codes should be boring and sortable. A date-plus-sequence format such as 2026-08-30-04 sorts correctly, tells you the day at a glance, and never collides. Whatever you choose, put the same code on the pan label so that a tub in the walk-in can be traced back to its row without guesswork; this is the same discipline described in lot traceability for gelato.
Two of those fields deserve emphasis. Draw temperature is what the batch freezer actually gave you, as opposed to what the recipe assumed, and it drifts as a machine ages. Finished weight per pan is a proxy for overrun that costs one weighing, and a slow change in it is often the first visible sign of a machine problem.
What the log will not do
A log is a record, not a control. It does not verify that your thermometer is telling the truth, so the numbers in it are only as good as your calibrated probes. It does not measure your mix, so a solids column filled in from the recipe sheet records an intention, not a fact; a bench reading from a moisture balance is what makes that column mean something.
It also will not survive being a chore. If the form asks for something nobody can answer at the machine, the entry gets left blank, then the next one does too, and within a month the log is fiction. Remove any field that has been blank three runs in a row. The log you actually fill in beats the log you designed.
Reading it back
A log that is only ever written is half an instrument. Once a month, spend twenty minutes reading it.
Sort by flavour and look down the draw temperature column. A stable number means a stable machine. A slow upward drift over eight weeks means a mantecatore losing refrigeration capacity, and finding that in a logbook is considerably cheaper than finding it during a July weekend. Do the same with finished weight per pan.
Then read the free-text notes on their own, ignoring everything else. This is where the pattern usually is, because people write down what surprised them. Three separate notes about a base being slow to cool, spread across a month, is a refrigeration problem nobody had yet named. That reading habit is also the natural input to your production timeline and to any change you are considering in the ageing step.
Keep the records long enough to be useful. Align retention with the shelf life of what you sell plus a comfortable margin, and check what your national authority requires; the underlying obligation to be able to trace product does not expire the day the tub is sold.
Paper or software
Paper wins at the machine. It survives wet hands, does not lock, needs no battery, and nobody has ever had to wait for it to load. Its weakness is that it cannot be sorted, which is precisely what you need for the monthly read-back.
The pragmatic answer for most labs is both: a clipboard at the batch freezer, transcribed weekly into a spreadsheet. The transcription is not wasted effort, because typing a week of rows is itself a review. Whatever you use, the entries must be attributable to a person and unaltered after the fact; correcting an entry means striking it through and writing the correction beside it, never overwriting. That is a requirement of any credible record, and it is also just good practice for a lab that intends to trust its own numbers. It sits alongside your HACCP setup, allergen labelling and cleaning protocols as part of one system rather than three.
Related Concepts
- HACCP Setup for a Gelateria is the framework this log documents.
- Lot Traceability for Gelato covers the ingredient side of the same obligation.
- Moisture Balance for Total Solids turns a calculated column into a measured one.
- Draw Temperature Explained is the field most worth trending.
- How to Measure Gelato Overrun gives you the texture number to log.
- Gelato Production Timeline is what the log lets you audit against reality.
- Cleaning Protocols for a Gelateria is the other record an inspector will ask for.
