Compliance & traceability

Batch traceability and mock recall for a dairy or food plant, in four hours

Updated 8 August 202610 min readnaffo.tech manufacturing teamReviewed by naffo.tech implementation desk

How do you set up batch traceability and run a mock recall in a food plant?

Batch traceability needs seven unbroken links: supplier lot at GRN, incoming QC result on that lot, lot issued to a specific production batch, batch output with its own batch number, finished-goods QC release, batch allocated on each dispatch line, and the customer on that invoice. With those links, a mock recall runs in both directions — complaint back to supplier lot, and supplier lot forward to every affected customer. Target completion under four hours end to end, including customer notification lists.

Key takeaways

Traceability is not a report you buy, it is seven links you must not break. One manual step anywhere and the chain resolves to "ask Ramesh".
One-up, one-down is the legal minimum. A real recall needs one-up, one-down at every hop, which is what makes it bidirectional.
Batch allocation at dispatch is unrecoverable if skipped. It costs nothing at the time and cannot be reconstructed afterwards.
Run a mock recall quarterly and time it. Untested traceability is a claim, not a capability.
Rework is the link most often lost, because material from batch A ends up inside batch B without a record.
The recall decision is a business decision made under time pressure — pre-write the roles, thresholds and notification templates while nobody is panicking.

Key figures

7: Data links that must remain unbroken for traceabilityBasis: Supplier lot at GRN, incoming QC on lot, lot-to-batch issue, batch output identity, FG QC release, batch-to-dispatch-line allocation, dispatch-to-customer.
< 4 hours: Target time to complete a bidirectional mock recallBasis: Practical target for a single-plant SME with the seven links captured in one system, including building the affected-customer notification list.

What traceability actually means on a shop floor

Every food business says it has traceability. Very few can answer this in one sitting: "A customer in Rajkot has a complaint about a curd cup dispatched on 14 July. Which milk lots were in it, what were their incoming QC readings, which other customers received product from those same lots, and how much of it is still in the market?"

That question is the whole discipline. Answering it does not require expensive software — it requires that seven specific links were captured at the moment the work happened, because none of them can be reconstructed later.

The seven links that must never break

Each link, when it is captured, and how it usually breaks
#	Link	Captured at	How it breaks in practice
1	Supplier lot identity	GRN / milk reception	Material received without a lot number, or all of a day's intake merged into one nominal lot
2	Incoming QC result attached to that lot	Before put-away	QC recorded in a separate register with no lot reference
3	Lot issued to a specific batch	Material issue	Issue booked to a product or department instead of a batch
4	Batch output identity	Batch close	Output added to stock without its own batch number and manufacturing date
5	Finished-goods QC release	Before stock becomes sellable	Product dispatched while QC result is pending, released verbally
6	Batch allocated per dispatch line	Dispatch / invoice	Invoice shows quantity only; batch decided by whoever loaded the vehicle
7	Customer on that dispatch	Invoice	Cash sale or distributor sale with no consignee detail

One-up, one-down, at every hop

Regulators generally require one-up, one-down: you must know your immediate supplier and your immediate customer for any consignment. That is the floor. It is not enough for a real incident, because a real incident is bidirectional and multi-hop.

Backward trace (root cause). Complaint → customer invoice → dispatched batch → batch record → issued lots → incoming QC → supplier. This tells you what went wrong.
Forward trace (exposure). Suspect supplier lot → every batch that consumed it → every dispatch from those batches → every customer. This tells you how big it is.
The multiplier is rework. If output from batch A was reworked into batch B, then batch B inherits batch A's exposure. Systems that do not record rework as a traced input silently understate the recall scope — this is the single most common traceability defect in food plants.

Withdrawal versus recall — decide the language before the incident

Different situations, different responses. Agree the thresholds in advance.
Situation	Typical response	Trigger	Who decides
Product still under your control	Stock withdrawal / hold	QC failure found before dispatch, or in your own warehouse	QC in-charge
Product at distributor or retailer, no health risk	Market withdrawal	Labelling error, weight variance, quality complaint without safety implication	Plant head with owner informed
Product reached consumers, potential safety issue	Recall	Contamination, allergen, foreign matter, microbiological risk	Owner, with regulatory notification
Uncertain whether risk exists	Hold and investigate on a clock	Ambiguous complaint or borderline result	QC with a defined maximum decision window

The fourth row is the one plants handle worst. "We're looking into it" with no deadline is how a withdrawal becomes a recall. Set a maximum decision window — for example four hours for a safety-adjacent complaint — and name the person who must decide within it.

The mock recall drill

Traceability is a claim until you time it. Run this quarterly, on a different product each time, and never announce which batch you will pick.

01Pick a real batch dispatched 7–20 days ago
Old enough that people have moved on, recent enough that the product may still be in the market. Do not pick a convenient batch.
02Trace backward
Customer invoice → finished batch → every raw material and packaging lot issued → incoming QC readings → supplier and supplier's lot reference.
03Trace forward
Those lots → every other batch that consumed them, including via rework → every dispatch → every customer and consignee.
04Quantify exposure
Produced, dispatched, in your warehouse, at distributors, and estimated already consumed. Four numbers, on one page.
05Draft the actual notification
Customer name, contact, invoices, batch numbers, quantity, dispatch date, instruction. If you cannot produce this list, you do not have traceability — you have records.
06Record every break and time it
Every phone call, register lookup or estimate is a broken link. Log the total elapsed time; under four hours is a pass for a single-plant SME.
07Assign corrective actions with dates
One owner and one date per break. Re-run in 90 days on a different product and compare times.

Records to retain, and for how long

Batch manufacturing record per batch: recipe version, materials and lots issued, process parameters, output, by-product, rework, rejection with reasons, operator and shift.
Incoming QC records per supplier lot, including rejected and conditionally accepted consignments with rate adjustments.
Finished-goods QC and release records, with the name of the person who released the batch.
Dispatch records with batch allocation per invoice line, plus vehicle and e-way bill reference.
Cleaning, sanitation and calibration logs for the equipment used in the batch — these are what an investigation asks for second.
Complaint register with batch lookup, response time and corrective action closure.
Retention period: follow your applicable regulatory requirement, and as a practical rule keep records for at least shelf life plus a safety margin. Verify current requirements with FSSAI or your certification standard, since retention rules differ by product category and scheme.

How this looks when the chain is in one system

The reason plants fail the drill is almost never absent records. It is records living in six places — a reception register, a QC notebook, a production diary, a stock register, an invoice book and a Tally company — with no shared key between them.

The shared key

Supplier lot → material issue → production batch → FG batch → dispatch line → invoice → customer

One identifier chain. Every document references the previous one, so the trace is a query rather than an investigation.

This is what naffo.tech's manufacturing chain implements: incoming QC recorded against the supplier lot before put-away, FEFO-driven material issue tied to a specific batch, batch records carrying by-product, rework and reason-coded rejection, finished-goods QC release, batch stock with manufacturing and expiry dates, and batch allocation on dispatch lines that flows into the GST invoice and e-way bill. The mock recall then becomes two screens instead of two days.

The five failures that show up in almost every first drill

Merged intake. A whole day's milk from twelve farmers or two tankers recorded as one lot, so backward tracing stops at "14 July intake".
QC in a separate book. Results exist but carry no lot reference, so they cannot be attached to the batch under investigation.
Rework untracked. Batch B contains material from batch A with no record, so forward exposure is understated.
Dispatch without batch allocation. The invoice says 40 crates; nothing says which batch, so exposure becomes the whole week.
Distributor tier invisible. You know the distributor received it, and nothing about where it went next. Agree in your distributor contract that they will provide onward dispatch details within a stated time on request.

Fix these five and most plants move from "we could probably work it out in a couple of days" to a timed, evidenced four-hour capability — which is also, incidentally, the thing that gets you approved as a vendor by large institutional buyers.

Run a bidirectional mock recall drill

01Pick a real dispatched batch at random
Choose a batch dispatched 7–20 days ago. Do not pick a convenient one, and do not warn the team which batch you will use.
02Trace backward to supplier lots
From the customer invoice, identify the finished batch, then every raw-material and packaging lot issued to it, with the incoming QC results for each.
03Trace forward to all customers
From those supplier lots, list every other production batch that consumed them, and every customer and dispatch that received output from those batches.
04Quantify exposure
Total quantity produced, quantity dispatched, quantity still in your warehouse, quantity at distributors, and quantity likely already consumed.
05Build the notification list
Customer name, contact, invoice numbers, batch numbers, quantities and dispatch dates — the actual list you would send.
06Time it and record the breaks
Note every point where you had to phone someone, open a register or guess. Each is a broken link to fix before the next drill.
07Close out with corrective actions
Assign each break to an owner with a date, and re-run the drill in 90 days on a different product.

Frequently asked questions

What is one-up, one-down traceability?

It is the requirement to know your immediate supplier and immediate customer for every consignment — one step up and one step down your supply chain. It is the regulatory floor rather than an operational capability. A real incident needs bidirectional, multi-hop tracing: backward from a complaint to the supplier lot for root cause, and forward from that lot to every affected customer for exposure.

How long should a mock recall take?

For a single-plant SME with the seven data links captured in one system, under four hours end to end — including quantifying exposure and producing the actual customer notification list. If it takes days, the failure is almost always missing batch allocation at dispatch or QC records that carry no lot reference, not missing records overall.

What is the difference between a withdrawal and a recall?

A withdrawal removes product that has not reached consumers, typically for a quality or labelling issue with no safety implication. A recall retrieves product that may already have reached consumers because of a potential safety risk, and generally carries regulatory notification obligations. Agree the thresholds and decision-makers for each in advance, including a maximum decision window for ambiguous cases.

How do you keep traceability through rework?

Record rework as a traced input, not as an adjustment. When output from batch A is reworked into batch B, batch B's record must list batch A as a consumed input so that forward tracing from a suspect lot picks up both. Untracked rework is the most common reason a forward trace understates recall scope, and it is invisible until a drill exposes it.

How long must food manufacturing batch records be retained?

Retention depends on your product category, regulator and any certification scheme you hold, so verify the current requirement with FSSAI or your certification body. As a practical operating rule, retain batch manufacturing records, QC results and dispatch records for at least the product's shelf life plus a safety margin, and keep mock recall reports indefinitely as evidence of a working system.

Can traceability work if we sell through distributors?

Partly — you can trace to the distributor, but not beyond unless they cooperate. Put an obligation in the distributor agreement to provide onward dispatch details within a stated time on request, and include the distributor tier in your mock recall drill. Plants that skip this discover during a real incident that their trace ends at the warehouse door.

This page also answers

What is one-up one-down traceability?
How long should a mock recall take?
What records are needed for an FSSAI recall?
How do you trace a customer complaint back to a milk lot?
What is the difference between a withdrawal and a recall?
How do you keep traceability through rework?

References

Run this inside one system

naffo.tech is an all-in-one business management and manufacturing platform for Indian SMEs: versioned recipes and BOMs, batch-wise material issue under FEFO, yield, by-product and reason-coded wastage capture, in-process and finished-goods QC, batch traceability to dispatch, GST invoicing with e-invoice and e-way bill, and double-entry accounting that closes live.

Book a demo See the modules