A food plant can pass every hygiene audit and still ship a lot that is underweight, too salty or slightly off in flavour. Quality control is the chain that stops that: a written spec for every material and product, checks at the dock, checks on the line, tests on the finished lot and a release decision that someone signs. When those links sit on paper, in spreadsheets and in the lab's inbox, release gets slow or blind. See one connected lot record in a short quality release demo.
Every Check Against the Spec, Every Lot Released on Evidence
iFactory Food Quality Pack keeps specs, voice-logged in-process checks and lab results in one record per lot. Positive release workflows hold each lot until the criteria in your SQF or BRCGS system are met, then route it to QA for sign-off.
Five Links From Spec to Release
Food safety plans control hazards. Quality control makes sure each lot is what the spec says it is: the right weight, recipe, taste and texture, in the right pack with the right code. The work follows the product through the plant, and each link hands evidence to the next. If one link is missing, the release decision rests on memory. Map your own chain with our food quality team.
Three Kinds of Spec, One Version Number
Both main GFSI schemes expect written specs. SQF Code Edition 9 covers them in 2.3.2, spanning raw materials, packaging, finished product and contract services, and BRCGS Food Safety Issue 9 sets out specification requirements in clause 3.6. A spec only works if the line and the lab use the same version, so every check should record which version it was measured against. SQF Edition 10 was published in March 2026, with audits expected no earlier than January 2027, so recheck clause numbers when you move over. Build your spec library in a spec setup session.
A COA Is a Claim. The Receiving Check Tests It.
A Certificate of Analysis reports the supplier's results for one lot. It is useful evidence, but only as good as the supplier's lab and its match to your spec. Good receiving practice compares every COA field with the current spec, matches the lot number to the delivery, adds quick physical checks at the dock and runs your own verification tests on a risk-based schedule. Materials stay on hold until that review is done. Set up COA review rules with our support team.
What the Line Checks, and Why
In-process checks catch drift while it can still be corrected. Each one needs a target, limits, a frequency, a method and a reaction plan, and each result should be logged against the lot and spec version. In the US, some of these checks also sit in the food safety plan as preventive controls under 21 CFR 117. Keep them in one place, but mark which are food safety controls and which are quality checks, because the response to a failure differs. Walk through your check plan in a line check review.
US packers check fill against NIST Handbook 133: the lot average must at least equal the label, and no pack may fall short by more than the Maximum Allowable Variation. In the EU and UK, the average quantity system sets three packers' rules, and packs from 5 g to 10 kg that meet them may carry the e-mark.
Speak the Check, See It Against the Spec
Operators say the reading. iFactory logs it to the lot, compares it with the current spec version and reads the result back. Anything outside the limits raises a hold before the next case is packed.
Four Test Families for the Finished Lot
Finished product tests confirm what the line checks suggest. Which tests you run, and how often, should come from the spec and your risk assessment. BRCGS Issue 9 covers this in clause 5.6, product inspection, on-site product testing and laboratory analysis, and expects results and trends to be understood and acted on. SQF Edition 9 covers product sampling, inspection and analysis in 2.4.4. Plan your sampling with our lab specialists.
Grade Against a Gold Standard, Prove Differences With a Triangle Test
Routine sensory checks compare each lot with a gold standard: an approved reference sample or description, stored under control and renewed on schedule. Trained assessors grade the lot on a short scale with clear anchors. When an ingredient, supplier or process changes, a triangle test shows whether a difference can be detected. ISO 4120:2021 describes it: each assessor gets three coded samples, two the same and one different, and picks the odd one, so a guess is right one time in three. Set up grading scales in a sensory setup session.
Positive Release or Release on Records?
Every lot needs a release decision, and sites make it in two ways. Under positive release, the lot stays on hold until all required results are back and acceptable. SQF Edition 9 clause 2.4.7.2 says that where a site uses positive release based on pathogen or chemical testing, a procedure must ensure product is not released until acceptable results are received. BRCGS Issue 9 sets product release requirements in clause 5.7. Other lots are released once the production records are complete and in spec, with tests run as verification. Choose per product, based on risk, with our food quality team.
- Lot held until every required result is in
- Used for pathogen, allergen or customer-required tests
- Needs hold space and shelf-life time
- Clear evidence trail behind every shipped lot
- Lot released when records are complete and in spec
- Tests verify the process over time
- Faster, with less storage needed
- A late failing result can mean a withdrawal
A Lot Release Checklist You Can Use Today
Use this as a starting point and tick each item as you confirm it. Adjust the items to your own specs, food safety plan and customer requirements. Get an editable version in a release workflow session.
Trends Across Lots, Not Just Pass or Fail
A check can pass every time and still drift toward a limit. When every result lives in one record, AI trending shows which checks are moving, on which line and shift, before a lot fails. The chart below is an illustration, not customer data. See trending on sample data with our support team.
Three Release Decisions, Illustrated
These are illustrations of typical situations, not case studies. Each shows how a connected record changes what happens next. Walk through your own recent holds in a release scenario review.
An Implementation Guide in Five Steps
You don't need to change everything at once. Start with the spec library, because every other link depends on it, then connect checks and results to lots one line at a time. Get help sequencing the work from our quality specialists.
Food Quality Pack and NVIDIA AI Server, Delivered Ready
iFactory delivers the Food Quality Pack as a turnkey package. A pre-configured NVIDIA AI server ships racked and ready with spec management, voice capture and trending models pre-loaded, so line checks and lab results stay on your own network. You rack it, plug in power and Ethernet, and the AI is live. To scope it for your plant, request a turnkey quote.
- Pre-configured NVIDIA AI edge server, racked and ready
- Spec library loaded from your current specs
- Voice capture for line checks on handhelds and headsets
- Lab result import from your LIMS or instruments
- Positive release workflows mapped to SQF or BRCGS
- AI trending of checks, holds and COA results
- 24×7 remote monitoring of models and hardware
Specs loaded, check plans written, server installed and linked to lines and lab.
Voice checks and lab results run on one line next to the current paper records.
Release workflows switched on, QA trained, weekly trend review started.
One Record From Spec to Signed Release
iFactory connects the links food plants usually keep apart, so QA can release a lot by reviewing one record instead of chasing five. Our support team can walk you through a pilot plan.
What Food QA Leaders Ask About Quality Control
Release Every Lot on Evidence, Not Memory
Keep specs, voice-logged checks and lab results in one lot record, and release only when every criterion is met. The full Food Quality Pack is delivered turnkey and goes live in 6 to 12 weeks.
