Food systems software: traceability without the enterprise theatre
Traceability fails when it is bolted on at the end. For processors and agrifood SMEs, it works when it matches how product actually moves from field to buyer.
Vinerals TechnologiesWorkshop notes
Traceability in food systems sounds like something only large exporters need. Then a buyer asks where a batch came from, a regulator asks for records you cannot produce in one afternoon, or a recall forces you to trace a lot through three spreadsheets and a WhatsApp thread.
At that point, traceability stops being a buzzword and becomes a Tuesday. The question is not whether you need it. The question is whether your system matches how product actually moves through your operation, or whether you are trying to tape an enterprise portal onto a process that still runs on paper and memory.
This piece is for food processors, co-ops, and agrifood SMEs who need practical chain-of-custody without buying a platform built for a company ten times their size.
What traceability means in practice for an SME
Traceability is the ability to follow a product, or a lot, backward and forward through your chain: where inputs came from, what happened to them on your floor, and where shipments went. That is it. Everything else is packaging.
For a small or mid-sized operator, the useful version is usually narrower than the enterprise brochure suggests.
One step back, one step forward.
You can name your immediate supplier and your immediate customer for any lot you shipped. For many SMEs, that is the regulatory and commercial baseline.
Batch or lot identity that holds together.
A code on the label ties to a record in your system. Mixing, splitting, and repacking create new lots with clear parentage, not mystery blends.
Events, not just inventory counts.
Received, washed, cut, packed, held, shipped. The story matters as much as the quantity.
Records you can pull in hours, not weeks.
When a buyer or inspector asks, you are not reconstructing the month from three people’s notebooks.
Export rules, buyer audits, and sector programmes may ask for more depth over time. Start with the chain you actually run today. Software that does not match that chain will be fought every day until someone goes back to the spreadsheet.
Data you already have vs data you invent
Most agrifood SMEs already generate traceability data. It just lives in the wrong shape: harvest tags, receiving slips, production logs, cold-room sheets, packing lists, invoices. The work is often collecting and linking, not inventing information from nothing.
Data you probably already have
Receiving.
Supplier name, delivery date, quantity, sometimes a grower or field code on the tag.
Production.
What went in, what came out, on which day, in which run. Even handwritten shift logs count if they are consistent.
Shipping.
Customer, lot or product code, quantity, ship date. Often half in the ERP and half on the loading dock clipboard.
Quality holds and corrections.
The batch that sat in quarantine, the relabel, the rework. These are traceability events too.
Data people invent because the process is broken
Retroactive lot numbers.
Assigned after the fact because nobody recorded them at receiving.
Generic “mixed lot” buckets.
Used so often they stop meaning anything.
Duplicate master data.
The same supplier under four spellings, the same product with two codes.
Narrative traceability.
“Ask Marie, she was on shift that week.” Marie is invaluable. She is not a system.
Good software starts by mirroring the honest path product takes, then makes that path easier to record. Bad software asks you to describe an ideal process you do not run, then blames the floor when adoption fails.
Common failure modes
We see the same patterns in food-sector projects, whether the client came to us or arrived after a failed rollout elsewhere.
The spreadsheet cathedral.
Years of tabs, macros, and colour codes only one person understands. It works until that person is on vacation or the file corrupts. Then traceability is whatever can be reconstructed.
The buyer portal bolt-on.
A large customer demands their format. Someone builds a one-off export or a separate portal just for them. Then a second buyer wants a different format. You are maintaining three truths.
Enterprise software, SME operation.
A platform built for global traceability, sold with a demo that skips the part where someone scans every crate on a wet loading dock at 6am. Licence fees match the brochure. Adoption matches reality.
Barcode theatre.
Labels printed, scanners bought, then used for the first week because the workflow around them is slower than pen and paper.
Traceability as a reporting layer only.
Pretty dashboards built on data nobody enters at the source. The report exists. The chain does not.
The failure is rarely technology alone. It is a mismatch between how product moves and how the system expects people to record it. Fix the match first. Tools second.
What to build first
A sensible first version for most SMEs is smaller than the RFP template suggests, and more useful than a pilot that only works in the office.
Lot identity at the moments that matter.
Receiving, transformation, and shipping. Capture there, not in a separate “traceability module” at the end of the day.
Simple parent-child rules.
When lots combine or split, the system records what went into what. No black-box blending.
One export shape that answers most questions.
A standard report or file you can send to buyers and regulators, with filters by date, lot, and customer.
Search that works for ops, not only for auditors.
“Where did this lot go?” and “What came in on Tuesday?” answered in under a minute.
Offline or low-friction capture where the work happens.
Cold rooms, fields, loading bays. If recording is harder than the task, the task wins.
Defer the glossy customer portal, the blockchain slide, and the multi-year integration map until the internal chain is solid. Buyers care that you can answer accurately, not that you bought the most impressive noun.
Where AI helps later, not first
Food traceability is a tempting place to put AI because the documents are messy: handwritten tags, PDF certificates, photos, emails. Models can help read those. They cannot replace a chain that was never recorded.
Useful later.
Pulling fields off supplier certificates into structured records, with a human checking the uncertain ones. Flagging when a scanned tag does not match the lot in the system.
Not a starting point.
Predicting contamination risk, automating recall decisions, or running customer chat over traceability data you do not have yet.
The order of operations.
First: one reliable internal record per lot. Second: clean enough exports for the buyers you serve. Third: assistive reading and search where volume justifies it.
If your traceability problem is really a data capture problem, AI will automate confusion. Fix capture first. Models are seasoning, not the meal.
How we approach food-systems work
We are a Montréal solidarity cooperative that builds software by hand. We have done work in agrifood and adjacent sectors. We are not a traceability vendor with a single product to fit every chain. We are a shop that listens to how product moves, then builds the smallest honest system that makes that movement legible.
Walk the floor first.
Receiving, processing, packing, shipping. We want to see where labels are applied, where data is invented, and where people already cheat the official process.
Start from existing records.
Tags, forms, exports. Software should reduce double entry, not add a parallel universe.
Ship something the ops team will use in week one.
Traceability that only exists for audits is traceability that fails audits under pressure.
Leave room for buyer-specific formats without forking the truth.
One internal chain, many export views.
Say when custom is the wrong answer.
Sometimes a sector platform or a tightened spreadsheet plus training is enough for the next two years. We would rather say that early.
We do not promise compliance on your behalf. Rules change, buyers differ, and inspectors ask their own questions. We promise software that makes your actual chain easier to show, without enterprise theatre you have to hire a team to maintain.
A short checklist before you buy or build
Run this internally before you sign with anyone, us included.
1. Can we trace one real lot today, on paper, end to end?
If not, software will not fix the gap. Process might.
2. Where are lots born, split, and buried in our operation?
Those moments are where capture must live.
3. What do our top three buyers actually ask for?
Build to that, not to a generic global standard.
4. Who will enter data at the source, and what will we remove to make room?
Adoption is a headcount and habit question.
5. What is v1 success?
“Pull a chain-of-custody report in ten minutes” beats “full digital transformation.”
Traceability is not a luxury feature for food-sector SMEs anymore. It is part of selling, operating, and sleeping at night when something goes wrong. It does not require an enterprise platform. It requires a chain that matches reality, recorded where the work happens.
If you are in food systems and trying to figure out whether to fix the spreadsheet, buy a platform, or build something that fits your chain, we are glad to walk through it with you. Bring one lot you shipped last month and how you traced it. We will start there.
