6 ways to automate food probe accuracy tests

#1 - Simple test

Who it's for: Single-site kitchens where the chef tests the probe themselves. No second checker, just a need to prove the probe reads true.

What it is: A food probe accuracy test is a record of a probe's reading in iced water and in boiling water. This version is the two readings that belong to one test: the reading in iced water and the reading in boiling water, in one box so each probe's test reads as one check. A probe should read 0°C in iced water and 100°C in boiling water, give or take a degree. Every cooked, chilled, and delivery temperature you record depends on the probe being right, so this is the check underneath all the others.

In practice: A single-site kitchen tests its probe each week. The chef fills a cup with iced water, reads 0°C, then boiling water, reads 99°C. Both within tolerance, logged in one box. If a probe reads 3°C in iced water, the log is what flags it before a week of cooked-food checks are taken with a faulty probe.

Why it works: Both readings live in one box, tied to a probe, so a test can never be logged half-done. The record proves the probe was sound on a given date, which is what backs up every other temperature record you keep.

Steps included:

• 1 grouped check (one box per probe) holding: reading in iced water (number), reading in boiling water (number)
• Duplicate the box for each probe

#2 - With guidance

Who it's for: Kitchens with rota staff who don't all know the expected readings and tolerance.

What it is: The simple test plus a guidance note. The note explains that the probe should read 0°C in iced water and 100°C in boiling water within about a degree, and that a probe outside that should be recalibrated or taken out of use because it can no longer be trusted.

In practice: A high-turnover kitchen runs this. A new chef de partie reads 0°C and 102°C, sees the note saying that is outside tolerance, and takes the probe for recalibration instead of carrying on with it. The faulty probe never gets used for a real cooked-food check.

What it adds to the basic test:

1. The expected readings and tolerance are on screen at the moment of the test
2. New staff know what counts as a fail without asking
3. The "recalibrate or remove from use" action is in front of them

Why it works: The guidance sits in the same box as the readings, so staff see it as they test. It turns the standard into a prompt that is always on screen.

Steps included:

• 1 grouped check (one box per probe) holding: reading in iced water (number), reading in boiling water (number)
• 1 guidance note (expected readings and tolerance)
• Duplicate the box for each probe

#3 - With a signature

Who it's for: Multi-site groups where someone signs off the probe tests and the records need a named sign-off for the audit trail.

What it is: The simple test plus a signature. The person running the test signs to confirm the probe reads within tolerance. For a group, that signature makes each site accountable for the calibration that all its other temperature records rest on.

In practice: A 25-site restaurant group runs a weekly probe test at every site, signed off by the duty manager. The food safety lead can confirm, across the estate, that every probe behind every cooked-food record was calibrated, all timestamped and signed.

What it adds to the basic test:

1. A signature confirming the probe reads within tolerance
2. Named accountability for each site's calibration
3. A record someone has put their name to, not just numbers on screen

Why it works: A signature turns a set of readings into a record someone has put their name to. When something goes wrong with a temperature check later, the calibration sign-off is there showing who confirmed the probe was sound.

Steps included:

• 1 grouped check (one box per probe) holding: reading in iced water (number), reading in boiling water (number)
• 1 signature (sign-off)
• Duplicate the box for each probe

#4 - With photo evidence

Who it's for: Kitchens under EHO scrutiny that want to show their probes are calibrated, not just claim it.

What it is: The simple test plus a photo of the probe reading in iced water. A photo of the probe held next to the ice, showing it reading close to 0°C, is proof the test was run and passed, which underpins every other temperature record you show an inspector.

In practice: A care-home kitchen photographs its weekly probe test by holding the probe next to the iced water. When an inspector questioned whether the cooked-food temperatures could be trusted, the calibration photo showed the probe reading true at that moment.

What it adds to the basic test:

1. A photo of the probe reading in iced water, captured at the time
2. Proof the probe was calibrated, which underpins every other check
3. A visual record that ties the test to a real probe

Why it works: Evidence taken in the moment is far stronger than numbers recalled later. A photo of the probe in the iced water ties the calibration to the actual probe and time, so the temperature records that depend on it hold up too.

Steps included:

• 1 grouped check (one box per probe) holding: reading in iced water (number), reading in boiling water (number)
• 1 photo of the probe reading in iced water
• Duplicate the box for each probe

#5 - With Poppi checking the photo

Who it's for: Kitchens where the calibration photo gets taken but nobody looks at it. Multi-site groups where head office can't review every site's probe tests.

What it is: A photo-checked test is the simple test plus a photo of the probe reading in iced water that Poppi (AI) reviews the moment it's saved. Poppi answers one question about that photo, set by you: does the probe read close to 0°C in the iced water, within 1 degree? A single named spot like the probe in iced water is something an AI can actually judge, where a wide shot cannot. If the answer is no, Poppi posts what it spotted to the team chat, so the problem gets fixed before the cooked-food checks start.

In practice: A three-site kitchen group photographs the probe test each morning. The opening shift takes the photo of the probe in iced water. Poppi reads the photo: reads 0°C, within range, no message. On a rushed Monday the photo shows the probe reading 3°C in iced water. Poppi answers no and posts the reason to the team chat ("The photo shows the probe reading 3°C in iced water, outside the acceptable tolerance"). The kitchen takes the probe for recalibration before service starts.

What it adds to the basic test:

1. A photo that gets checked the moment it's saved, not just stored
2. A team chat message with Poppi's reason the moment a photo fails the check
3. The manager stops being the only person who ever looks at probe photos

Why it works: The check happens in the seconds between the photo being taken and the cook starting service. That's the only moment the problem is still cheap to fix. A manager reviewing photos at the end of the day can only record that the probe was wrong; Poppi catching it at 6:30am gets it recalibrated before breakfast.

Steps included:

• 1 grouped check (one box per probe) holding: reading in iced water (number), reading in boiling water (number)
• 1 photo of the probe reading in iced water
• 1 Poppi decision (judges the photo against your question)
• 1 Poppi action (posts to the team chat if the photo fails the check)
• Duplicate the box for each probe

#6 - With a team alert if the test is overdue

Who it's for: Multi-site groups where the daily probe test has to be run at a set time, and sometimes it gets skipped or forgotten on a busy shift.

What it is: An overdue-test alert is the simple test plus a Poppi (AI) action set to the test's scheduled time. If the probe test isn't finished by when it should have been run, Poppi posts a message in the Pilla team chat so the outstanding test gets done before the cooked-food checks start. It watches the scheduled time, so it catches the test that quietly got skipped, not just the one that was never attempted.

In practice: A 12-site cafe group schedules the daily probe test for 6:30am before service starts. On a busy Saturday one site is short-staffed and the opener skips it to get service going faster. With this version, if the test isn't finished by 6:30am, the team chat gets a message, so the shift lead catches the skipped test that morning rather than finding out at an audit that one site ran its temperature checks on an uncalibrated probe.

What it adds to the basic test:

1. A message in the team chat if the test isn't finished on time
2. A catch for the test that got skipped, not just forgotten
3. A record of when the test should have been run, next to when it was
4. The manager finds out the same day, not from a compliance review

Why it works: The alert is tied to the scheduled test time, so an overdue test raises its own hand. Nobody has to notice it's missing; the deadline does.

Steps included:

• 1 grouped check (one box per probe) holding: reading in iced water (number), reading in boiling water (number)
• 1 Poppi action (posts to the team chat if the test isn't finished by its scheduled time)
• Duplicate the box for each probe

How to pick the right version

You don't need to know our product to choose. Every version here is the simple test plus one addition, so pick the additions your kitchen actually needs.

Do other people run the test?

If you test it yourself and know the tolerance, the plain list is enough: #1. The moment rota staff do it, the tolerance needs to be on the screen: #2.

Does the test need a name against it?

If knowing it was done is enough, skip this one. If you want who signed it off on the record, #3 adds a signature.

Do you need photo proof?

A logged test says the work was done; a photo shows it. If you want visual proof, #4 adds a photo of the probe reading in iced water.

Does anyone actually look at the photos?

If a manager genuinely reviews every photo, #4's record is enough. If photos get taken and filed unseen, #5 has Poppi (AI) check each one as it's saved, and tell the team chat when something's wrong.

Does the test have to be finished by a set time?

If a busy team sometimes skips the test to get service going, #6 posts a message to the team chat when the deadline passes with the test unfinished.

Need more than one addition? Open the version with the addition that matters most in the playground and add the others as steps. That's how the product works anyway: every option here is one step added to the same test.

Related workflows

• Cooked food temperature check - the check this calibration underpins
• Fridge temperature check - another reading only as good as the probe behind it
• Food cooling temperature check - where a true-reading probe proves food cooled in time

Conclusion

Every temperature you record is only as trustworthy as the probe that took it. A logged accuracy test is what lets you stand behind all the others. Every version above is the same simple test plus one addition: guidance, a signature, a photo, an AI check on the photo, or a deadline alert. Pick the ones your kitchen needs and combine them in the playground.

More additions are coming in future refreshes, like pulling every site's calibration records into one report. Those need more review time and will land separately.

→ Build your own food probe accuracy test on Pilla.