Validation answers one question about a control point: if we do what the plan says, will it control the hazard? It's done before you rely on the plan and again when something changes. It has two halves. The first is evidence that the limit itself is right: a regulation, a study or a process authority's letter. The second is evidence that your kitchen or plant can meet that limit with its own equipment, on its worst day. A validation record puts both on one page with a conclusion and a signature. It isn't the same as verification, which asks whether the plan is being followed day to day. Below are the rules, the ten parts of a record with an example, and a file.
The short version
- Validation: can the plan work? Verification: is it being followed?
- Two kinds of evidence: the science behind the limit, and your own measurements.
- Test the hardest case, not the easiest.
- Write a conclusion that says exactly what was validated, and what wasn't.
- Do it again when the product, the equipment or the method changes.
- TaroStack keeps every batch's readings, which become the data for the next validation. More below.
What the rules ask
| FDA-regulated food (21 CFR 117.160) | Meat and poultry plants (9 CFR 417.4) | |
|---|---|---|
| What's validated | Preventive controls, as fits their role | The HACCP plan's adequacy |
| Who | A preventive controls qualified individual, or under one | The establishment |
| When | Before the plan is implemented, or within 90 calendar days after production first begins; on a change; when a reanalysis calls for it | On completing the plan; reassessed at least annually and on any change |
| How | Scientific and technical evidence, or studies where evidence is lacking | Repeatedly testing the CCPs, critical limits, monitoring, records and corrective actions |
| Not needed for | Allergen controls, sanitation controls, the recall plan, the supply-chain program |
A restaurant working under the Food Code meets the same idea from another side: a plan filed for a variance has to show the method is safe, and the health department reviews it.
The ten parts of a record
A made-up kitchen validating its cooling step.
| Part | Example: cooling chili in one-gallon bags |
|---|---|
| Control point and hazard | CCP 2 Cooling. Spore-forming bacteria growing as cooked chili cools |
| Critical limit | 135°F to 70°F within 2 hours; to 41°F within 6 hours in all |
| The question | Does our method (one-gallon bags, laid flat in an ice bath) reliably meet the limit? |
| Scientific or technical support | FDA Food Code 2026, 3-501.14. The limit is the code's own |
| In-plant demonstration | 10 batches, Oct 5 to Oct 30, 2026. Probe in the center of the thickest bag, at 2 and 6 hours |
| Worst case covered | Largest batch (20 gallons, 20 bags), ice bath at the end of a shift, warmest day of the month |
| Results | At 2 hours: 58°F to 68°F. At 6 hours: 36°F to 40°F. All 10 inside the limits |
| Conclusion | Validated for one-gallon bags laid flat. Not validated for two-gallon bags, or for bags stacked |
| Who and when | R. Paek, chef; November 2, 2026 |
| Do it again when | Bag size, batch size, ice bath, recipe thickness or kettle changes; a cooling failure; the yearly review |
Three things that make a validation hold up
The worst case. Ten easy batches prove nothing about the hard one. The slowest reading here was 68°F at two hours, two degrees inside the limit, and it came from the 20-gallon batch on a warm day. That's the number that matters, and it says this method has little room to spare.
A narrow conclusion. The record says what was tested and stops there. One-gallon bags, flat. When someone buys two-gallon bags to save time, the record makes it obvious that nobody has shown those work.
The data, attached. A conclusion without readings is an opinion. Keep the sheet of 20 numbers with the record.
When the limit isn't from a rule
The example is the easy kind: the limit is the Food Code's, so the science is settled and only the in-plant half needs doing. A roast on cocoa beans, a ferment's pH and time, or a dried sausage are harder. There the first half needs a real source: a published study that matches your process, a guideline from the agency, or a process authority who'll put it in writing. If you can't find one, the rule says to do a study, and that usually means paying someone qualified.
Validation, verification, reassessment
| Question | Examples | |
|---|---|---|
| Validation | Can the plan work? | This record |
| Verification | Is it being followed? | Calibrating thermometers; watching a reading being taken; reviewing logs |
| Reassessment | Is it still the right plan? | A yearly review; after a new supplier, recipe or machine |
The first two are compared in validation vs verification. When a limit is missed, the record for that is a corrective action form.
This is general information, not legal advice. Your inspector, auditor or health department has the last word on what evidence they'll accept.
Get the template
Download the HACCP validation template. The ten parts, with the cooling example beside a column for yours, and a read-me. One sheet for each control you validate. It's an ordinary Excel file with no macros, and it works in Google Sheets, LibreOffice and Numbers. The limits most kitchens validate against are in critical limits for CCPs.
Where this stops working
The hard part of a validation is the data: ten batches, twenty readings, the batch sizes, the dates. If the readings were written on a clipboard, someone has to find the sheets and copy them out. A year later, at reassessment, they do it again. And when the bags change in March, nobody connects that to a record signed in November.
How TaroStack does it
TaroStack already has the readings, because they were recorded with each batch.
Readings are fields on the batch record. Cooking temperature, the 2-hour and 6-hour readings, pH: whatever your plan monitors, recorded with the batch, its size and its date.
Every batch is kept. Batches by product and date, each with its readings. Nothing is deleted, so the ten batches for a validation, or the year's for a reassessment, are a search.
Recipes have versions. A change to the recipe is a new version with a date, which is the prompt to ask whether the validation still holds.
Recipe truth. From your own batch history, TaroStack shows whether a recipe yields what it claims: an early sign that a process has drifted.
| The job | Clipboards | In TaroStack |
|---|---|---|
| Ten batches of readings | Find the sheets and copy | Search by product and date |
| Batch size for each reading | Often not written | On the batch record |
| When did the recipe change? | Memory | The version's date |
| A year of data for reassessment | A box of paper | The same search |
TaroStack doesn't validate a plan or judge the evidence. Recipes and batches are on every plan, from $9.99 a month (5 recipes) or $39 a month (50 recipes). The first 14 days are free, with no credit card. If you'd like a hand setting up, ask, and we'll do it with you.
