Temperature monitoring software for food operators.

I am Andrew Langevin. I run a CFIA-licensed (Canadian Food Inspection Agency) mushroom production facility in Brantford, Ontario, under the SFCR (Safe Food for Canadians Regulations). I have lived through a walk-in failure on a Monday morning, a 25-minute calibration binder hunt during a CFIA inspection, and the quote for a $189-per-month-per-location sensor bundle that would have eaten my net margin. The temperature monitoring inside HACCPlan exists because none of the options I priced fit a real small operator. Here is what the software does, what it does not do, and why it ended up the shape it did.

01The gap

The reading is not the record.

Every food business already has thermometers. The pain isn't getting a number on a screen. The pain is the gap between the number on the screen and a defensible record an inspector will accept. That gap is what kills people.

A walk-in that read 51 degrees Fahrenheit at 6 a.m. Monday morning isn't a software question. The compressor failed Friday night, the unit warmed slowly through Saturday and Sunday, and nobody knew until my production lead opened the door for the morning shift. We lost about $3,800 of product and spent the next four hours on the phone with customers. That weekend cost more than five years of monitoring software.

The cooler that reads 38 degrees Fahrenheit and gets logged at the same number every day for two weeks in the same handwriting? That's a different version of the same problem. The cook filled it in Friday afternoon for the whole week. The inspector reads the binder, notices the pattern, and writes you up. The reading exists. The record doesn't.

What temperature monitoring software is for, in one sentence, is connecting the reading to the record — at the moment the reading happens, with the person who took it, against the threshold that applies, with a workflow when something is wrong.

02The regulations

What an inspector actually checks.

The temperature rules are specific. "We check it sometimes" does not survive an inspection. The sections that drive most write-ups, in the order an inspector usually reads them, are these.

1. 01

   Hot and cold holding — FDA Food Code §3-501.16

   Cold-held TCS (Time/temperature Control for Safety) food at or below 41 degrees Fahrenheit (5 Celsius). Hot-held at or above 135 degrees Fahrenheit (57 Celsius). The range in between is the danger zone — where bacteria multiply fastest. This is the single most-cited family of temperature provisions in open city inspection databases.

2. 02

   Two-stage cooling — §3-501.14

   Cooked TCS food from 135 degrees Fahrenheit to 70 degrees Fahrenheit within 2 hours, then from 70 degrees Fahrenheit to 41 degrees Fahrenheit within an additional 4 hours. Total cooling not to exceed 6 hours. The first 2-hour window is the most critical because food is passing through the fastest-bacterial-growth range. The two-stage cooling log is the most-failed log in the Food Code — every inspector I have spoken to puts it first.

3. 03

   Cooking — §3-401.11

   Poultry and stuffed items 165 degrees Fahrenheit (74 Celsius) for 15 seconds. Ground meat 155 degrees Fahrenheit (68 Celsius) for 17 seconds. Whole-muscle meats, fish, and shell eggs cooked to order 145 degrees Fahrenheit (63 Celsius) for 15 seconds. Internal temperature, probed at the thickest part.

4. 04

   Date marking — §3-501.17

   Refrigerated ready-to-eat TCS food held more than 24 hours marked with the date or day prepared and a discard date no more than 7 days later at or below 41 degrees Fahrenheit. Temperature monitoring and date marking are usually checked together because they answer the same question — is the food still safe to serve.

5. 05

   Monitoring, corrective action, verification — 21 CFR §117.140

   For FSMA-regulated manufacturers, the law requires the trio: monitoring (you measured the value), corrective action (you did something when the value went out of range), verification (someone reviewed the monitoring records within 7 working days). All three documented. Missing any one is a 483 observation. (Source: eCFR §117.140.)

6. 06

   Process controls — SFCR §86 (Canada)

   For CFIA-licensed operators, the Preventive Control Plan (PCP) must demonstrate that hazards — including temperature-related hazards during transport, storage, and processing — are prevented, eliminated, or reduced to an acceptable level. Records under §95 retained for 2 years. This is the section my own inspector audits against every six months at the Brantford facility.

03Manual + sensor

Manual logs and sensor data are not a choice — you need both.

Most platforms force you to pick one. The honest answer is that any operation past a single shift needs both, because they answer different questions.

A sensor in a walk-in cooler runs continuously. It catches the failures that happen at 11 p.m. Friday when nobody is in the building. It does not tell you whether the chicken breast you just pulled out of the oven hit 165 degrees Fahrenheit. You cannot strap a wireless probe to a chicken breast, and even if you could, it would not survive the next service.

A manual log with a hand-held probe is the only honest record of cooking temperatures, cooling-curve milestones at the 2-hour and 6-hour marks, receiving checks at the truck, and hot-line spot checks during service. It is also the only thing that gets signed by the person who took the reading — which is what the inspector wants on the page.

The two layers work together. Sensors handle the ambient environment of equipment you cannot stare at for eight hours. Manual logs handle the product. HACCPlan ships both as first-class workflows, not one as an afterthought added to the other.

04Probe-photo

The probe-photo capture, and why it exists.

If you have ever asked a cook to stop, find a tablet, open an app, type "walk-in three, 38 degrees Fahrenheit," and sign — you already know the answer. They will not do it during service. They will do it at 4 p.m. for the whole day.

The probe-photo capture in HACCPlan is the workaround. The cook reads the thermometer, holds her phone or the line tablet up, takes a photo of the LCD screen or the analog dial, and taps once to confirm. The value is read off the photo by Claude vision, pre-filled into the log against the equipment she selected, time-stamped from the device clock, and signed under her account. The whole interaction is about three seconds. The photo itself becomes the evidence — if an inspector ever questions whether the reading is real, the picture of the thermometer at that timestamp is in the log.

The feature exists because I was looking at a stack of Bluetooth probes one morning and realized none of my staff was going to type the numbers in. Not because they were lazy. Because they were cooking. The right tool is the one that captures the reading in the same motion as taking it.

“

Just the picture? That's it?

”

My production lead, the day we rolled out probe-photo capture

A few honest limits. Probe-photo capture speeds up the data entry — it does not replace the human looking at the value. If the cook photographs a thermometer that's broken and reading 40 degrees Fahrenheit when the food is actually 50, the photo still says 40. The reading discipline still belongs to the person. What the photo does is remove the friction so the reading actually gets logged at the right moment, with the right equipment association, against the right threshold.

05Sensor-agnostic

Bring the sensors you already own.

HACCPlan does not sell sensors. That's a deliberate choice. Most operators evaluating a monitoring platform already have hardware installed — a few Monnit sensors in the walk-ins, a Dickson logger in the warehouse, a Comark handheld probe on the line, an old Sensaphone box wired to the freezer. The math of a rip-and-replace doesn't work for an SME operator, and the lease contracts that come bundled with hardware-included subscriptions are how monitoring budgets explode.

The sensors HACCPlan reads, in plain terms:

1. 01

   Monnit (ALTA wireless family)

   Industrial-grade wireless sensors, $70 to $150 each, 21 CFR Part 11(B) compliant, NIST-traceable certificate at ship. Strong in warehouse and cold storage. HACCPlan reads the iMonnit cloud feed via API. You keep your existing sensors; the data flows into HACCP CCP records.

2. 02

   Dickson Data (DicksonOne)

   Display Loggers, wireless temp-and-humidity loggers, differential-pressure loggers. Every device ships with a NIST-traceable calibration certificate that you can attach to the probe record in HACCPlan. Strong in food and pharma. Read via DicksonOne API.

3. 03

   Comark Instruments (Diligence WiFi)

   Sixty-plus years in temperature. Diligence WiFi sensors and the handheld HACCP-checked probes that are industry standard in catering. Read via Comark Cloud. Common in school catering and contract foodservice.

4. 04

   SmartSense by Digi

   Cellular-gateway monitoring built for grocery chains and multi-unit QSR. Open API. If you already pay for SmartSense at one site and need to consolidate records across sites that don't, HACCPlan reads the SmartSense feed.

5. 05

   Sensaphone

   Alarm-and-monitoring boxes that phone-call contacts when thresholds break. Widely used in dairy, ice cream, and frozen storage. Read via webhook.

6. 06

   Bluetooth probes (ThermoWorks, Cooper-Atkins, Inkbird, Thermapen)

   Direct probe-to-app pairing for hand-held use. The reading goes straight into the cooking or cooling log on the tablet — no transcription, no back-fill.

7. 07

   Anything else

   Generic webhook ingestion and CSV import for any brand not on the native-connector list. If the sensor can post a JSON payload to a URL, HACCPlan can read it.

The principle is the same as the rest of the product. The HACCP layer is where the value is. The sensors are commodity inputs. Your existing hardware should not become a hostage to your software choice.

06Alerts

Alerts that don't burn out the manager.

A walk-in that crosses 42 degrees Fahrenheit for 30 seconds while the door is open during a delivery is not a crisis. A walk-in that holds at 44 degrees Fahrenheit for 4 hours is the difference between a $4,000 product loss and a $0 product loss. The difference between a useful alert and a useless one is whether the alert ladder matches the operational reality.

HACCPlan ships threshold templates with the FDA Food Code values pre-loaded, plus a tiered alert ladder you configure per piece of equipment:

1. 01

   Warning band (e.g., 30-minute excursion)

   In-app notification. Goes to the person on duty. Most of these resolve when somebody closes a door, finishes a delivery, or repositions a fan.

2. 02

   Manager escalation (e.g., 2-hour excursion)

   Text and email to the manager on duty. Now a human has to look at it. Product status decision: stay, move to backup unit, or hold for evaluation.

3. 03

   Owner escalation (e.g., 4-hour excursion or sustained drift)

   Text, email, and Slack or Teams to the owner or director. At this point the question is whether to call a refrigeration tech and whether the affected product can still enter commerce. (For TCS food held above 41 degrees Fahrenheit for more than 4 hours, the answer to the second question under §3-501.16 is generally no.)

07CCP linkage

Every reading attached to a CCP — when one applies.

A reading without context is a number on a screen. A reading attached to the equipment it monitors, the CCP it supports, the threshold it's measured against, and the person who took it is an audit-defensible record. HACCPlan links readings to all four automatically.

When a deviation occurs, the platform does the work most apps skip. The reading triggers a draft corrective action record (CAR) pre-populated with the equipment, time, value, excursion duration, and affected product. The operator confirms or amends the suggested actions — move product, adjust thermostat, call refrigeration tech, discard if more than 4 hours in the danger zone. The completed CAR links back to the original reading. The supervisor signs off as part of the 7-working-day verification window under 21 CFR §117.165.

A note on the over-CCP-ing trap. Not every temperature reading needs to be tagged to a CCP. The 2023 Codex CCP decision tree first asks whether the hazard can be controlled by prerequisite programs alone. For a lot of cold-holding scenarios, the answer is that ambient temperature is a prerequisite program control, not a CCP. The first consultant draft of my Brantford facility's PCP labeled a cooling step as a CCP that ran the team ragged on 30-minute monitoring intervals. Running the 2023 tree honestly dropped it to a prerequisite-program control with 2-hour intervals plus sensor backup. The hazard is still controlled. The team got nine hours a week back.

08Calibration

The part operators forget until the inspector asks.

The first question a serious auditor asks about temperature monitoring is rarely about the readings themselves. It's about the thermometers. When was this probe last calibrated? Against what reference? Who did it? What was the result? At my facility's first CFIA inspection, that line of questioning was a 25-minute binder hunt that ended with "sometime in March." That hunt is the reason HACCPlan has a calibration tracker module.

Both FSMA (21 CFR §117.40(b) and §117.165(a)(4)) and USDA FSIS (9 CFR §417.4 ongoing verification) explicitly require calibration of temperature-monitoring instruments. The two acceptable field methods:

1. 01

   Ice-point check (cold-side accuracy)

   Crushed ice plus a little cold water in a glass. The slurry stabilizes at 32 degrees Fahrenheit (0 Celsius). Insert the probe to the marked depth, do not touch the glass, wait 30 seconds, read. Acceptable accuracy: plus or minus 1 degree Fahrenheit. Run weekly at minimum; daily for high-use probes; always after a probe is dropped, frozen, or out of range during a check.

2. 02

   Boiling-point check (hot-side accuracy)

   Rolling boil at sea level reads 212 degrees Fahrenheit (100 Celsius). Adjust about 1 degree Fahrenheit lower per 550 feet of elevation. Insert the probe, do not touch sides or bottom, read. Acceptable accuracy: plus or minus 2 degrees Fahrenheit.

3. 03

   NIST-traceable reference

   An annual external calibration of your reference thermometer against an ISO/IEC 17025-accredited lab standard. Most sensor vendors ship sensors with a NIST-traceable certificate that satisfies this for the sensor itself. The reference thermometer you use to check your other probes is the one you should send out annually.

HACCPlan's calibration tracker stores per-thermometer: ID, manufacturer, model, serial number, reference standard used, last calibration date, next calibration due date, calibrator name, calibration result (pass, fail, out-of-range, adjusted). It auto-reminds you when a probe is due. It auto-locks logged values from an overdue thermometer — "this probe is overdue for calibration; recalibrate before this reading can be saved" — so the audit never finds a reading produced by an uncalibrated probe.

You can attach the NIST certificate PDF directly to the probe record. The inspector asks when each probe was last calibrated; one tap shows him the register with the certificate behind every row.

09Records + export

Records that survive an inspection.

The format of a defensible record matters as much as the readings themselves. Every reading captured in HACCPlan carries:

1. 01

   Timestamp from the device clock

   Not a typed time that can be back-filled. The clock the entry came from is captured against the entry.

2. 02

   Person or sensor ID

   The user account that signed the entry or the sensor that produced it. No anonymous readings.

3. 03

   Equipment ID

   Which unit. Walk-in 1, Freezer 3, Hot well A. Inspectors need to be able to trace a deviation to a specific piece of equipment.

4. 04

   Value plus unit

   The actual reading. Fahrenheit or Celsius, your choice; both displayed where it helps.

5. 05

   Threshold and citation

   The applicable threshold (e.g., ≤41 degrees Fahrenheit) plus the citation (e.g., §3-501.16 or SFCR §86). When a reading is logged, the threshold it's checked against is captured with it.

6. 06

   Photo evidence (when probe-photo capture was used)

   The image of the thermometer attached to the entry.

7. 07

   Corrective action link (when a deviation occurred)

   The CAR record linked to the original reading.

8. 08

   Supervisor verification (within 7 working days for FSMA-regulated operations)

   A second signature confirming the supervisor reviewed the entry and any linked corrective actions.

Records are retained for a minimum of 2 years (configurable longer) — satisfying 21 CFR §117.330 for FSMA operators, 9 CFR §417.5 for USDA FSIS operators (6 months on-site, 24-hour accessibility for the remainder), and SFCR §95 for CFIA-licensed Canadian operators. The audit trail is immutable: entries cannot be deleted, only marked corrected with a reason and an audit-trail entry.

Export options match the audiences who ask for records:

1. 01

   PDF for inspectors

   Per-equipment time series, deviations highlighted, linked CARs, calibration log, supervisor sign-off — assembled as one document in the order an inspector typically reads it. The same export shape that came out of my own facility's PCP record review.

2. 02

   CSV for analysts

   Raw data for the operator who wants to do their own trending in Excel or Sheets, or hand to a corporate QA team.

3. 03

   JSON for developers

   Structured payload for any operator integrating with an ERP, an MRP, or a custom dashboard.

10Where it fits, where it doesn't

Honest positioning.

The places HACCPlan's temperature monitoring is the right answer:

• A multi-location restaurant operator, 1 to 25 locations, who has been written up on cold-holding logs or lost product to an overnight equipment failure.
• A small-to-mid food manufacturer ($1M to $50M revenue) treating temperature as a CCP under HACCP, FSMA Preventive Controls, or SFCR — who needs defensible records for SQF or BRC, traceable calibration, and deviation-to-corrective-action linking.
• A CFIA-licensed Canadian operator under SFCR §86 needing electronic temperature records as part of a PCP.
• Any operator who already has Monnit, Dickson, Comark, SmartSense, or Bluetooth probes installed and does not want a rip-and-replace.

The places it is not the right answer:

• A 500-plant industrial manufacturer with PLC-and-SCADA temperature feeds tied into a real-time CCP dashboard. SafetyChain serves that scale honestly. At their size, that's the right call.
• A one-person operation with a single home-style fridge and no inspection regime to satisfy. The free temperature log templates further down this page are enough.
• An operator who genuinely wants the vendor to install and own every sensor in the facility. SmartSense by Digi is built for that model. If a bundled lease is what you want, they're the answer.

The middle — sensor-agnostic, manual-and-sensor both first-class, HACCP CCP integration, AI probe-photo capture, calibration tracker, published price — is the gap HACCPlan was built to fill.

11Templates first

Start with the free templates if you're not ready for software.

If you're still on paper and want to upgrade your logs without committing to software yet, start here. Every template below has the threshold printed in the column header with the Food Code citation, a corrective-action column, and a supervisor-verification column. The sibling article on the temperature log template hub goes deeper on why those three columns matter.

Most operators run on the free templates for a few weeks before deciding whether the software fits. That is the right order. Get the logs filled out reliably on paper first; the software is the version that keeps the records integrated, time-stamped, and inspector-ready.

12What this week looks like

A realistic first week.

If you decide to move from paper to the platform, the first week of onboarding looks roughly like this for a single-location operation. Multi-location chains add a corporate-setup phase up front.

1. 01

   Days 1 to 2 — inventory your equipment and your probes

   List every cooler, freezer, walk-in, hot well, and ambient-controlled storage area that needs monitoring. List every probe and reference thermometer. Note the manufacturer, model, and serial of each. This is the foundation of both the equipment list and the calibration register.

2. 02

   Days 3 to 4 — connect your existing sensors

   If you have Monnit, Dickson, Comark, SmartSense, or Sensaphone hardware installed, connect the API credentials. The historic data backfills into the platform so you don't lose your past month's record. If you're using Bluetooth probes, pair them.

3. 03

   Day 5 — set the thresholds and the alert ladder

   Per-equipment critical limits seeded from the FDA Food Code templates, plus the consecutive-excursion durations for your three-tier alert ladder. Decide which manager gets paged at which level. Test the alert flow with a manufactured deviation.

4. 04

   Day 6 — go live on the manual logs

   Switch the cooking, cooling, hot-cold holding, and receiving logs from paper to the tablet. Spend one shift per station shoulder-to-shoulder with the cook making entries until the rhythm clicks. Keep paper backups for the transition month.

5. 05

   Day 7 — run the first calibration round

   Ice-point check on every probe. Record results in the calibration tracker. Set the next-due dates. Any probe that fails the check goes out of service until adjusted or replaced. Sign off.

By the end of the first week, the platform should be producing readings, the calibration tracker should be current, the alert ladder should have been tested, and the inspection-day export should be running against a real two-week-old data set. If it is, the system is working.

Footnotes

1.FDA Food Code 2022 (full PDF, §3-401.11 cooking, §3-501.14 cooling, §3-501.16 hot/cold holding, §3-501.17 date marking) — fda.gov

2.FDA Cooling Cooked TCS Foods guidance (§3-501.14 explainer) — fda.gov

3.21 CFR §117.140 (FSMA preventive controls: monitoring, corrective action, verification) — ecfr.gov

4.21 CFR §117.165 (FSMA verification, including calibration and 7-day record review) — ecfr.gov

5.9 CFR §417.4 (USDA-FSIS HACCP verification, including calibration) — ecfr.gov

6.CFIA — Preventive Control Plan regulatory requirements (SFCR §86) — inspection.canada.ca

7.CDC — Foodborne Outbreak Surveillance (NORS), contributing-factor data on time-temperature abuse — cdc.gov

Andrew Langevin·CFIA-licensed facility, Brantford ON· Published 2026-06-04· 11 min read· Wikidata Q139112497