Everything That Can Fail a GACP Audit — From Tiny to Terminal

GrowRoom · July 2026 · 7 min read

Farms rarely fail a GACP audit because of the plants. They fail on findings — and findings come in sizes. Here's the honest ladder, from the ones that cost you a remark to the ones that cost you the certificate, and why most rungs simply vanish when your records keep themselves.

The small stuff (that adds up)

• A missing initial or date on one log line. Trivial alone; a pattern of them tells the auditor your records are written after the fact.
• Correction fluid or scribbled-out entries. Data-integrity rules want corrections visible: original value, new value, who, when, why.
• Inconsistent formats. Three notebooks, two spreadsheets and a wall calendar for the same batch invites transcription errors — and questions.
• Illegible handwriting. If the inspector can't read it, it may as well not exist.

The medium stuff (findings you must answer)

• Gaps in the daily record. The long weekend nobody logged. Songkran. The week the intern left. Auditors look for continuity first because it's the easiest thing to check.
• Overdue calibration. An EC pen with no calibration log casts doubt on every reading it ever produced.
• Inputs without lot numbers. "We fed nutrients" isn't traceability. Which product, which lot, how much, applied how, by whom?
• Waste with no weight or witness. Cannabis waste is controlled material — an unwitnessed disposal line is a diversion question waiting to be asked.
• Photos with no dates or batch link. Evidence that can't be placed in time or tied to a lot proves very little.
• SOPs that exist but show no completions. A beautiful IPM plan with no scouting log is worse than no plan — it documents your own non-compliance.

The big stuff (certificate at risk)

• An environmental excursion with no record and no response. The humidity spike happened; your notes are silent. If mold shows up in the COA later, the silent record becomes the story.
• Deviations without CAPA. Problems are acceptable — undocumented problems are not. No root cause, no corrective action, no close-out: major finding.
• Yield that doesn't reconcile. Wet weight minus dry weight minus recorded waste should roughly close. A few hundred grams unaccounted for is a traceability failure — and in a controlled crop, a serious one.
• A broken chain of custody. If a lot can't be followed seed → room → harvest → dry → COA → release, the batch record has failed at its one job.
• Signs of backdating. The fastest way to convert every small finding into a fatal one: once integrity is in doubt, nothing in the binder counts.
• A failed COA with no quarantine trail. Failing a lab test is survivable. Shipping — or being unable to prove you didn't ship — before the retest is not.

Now notice what almost all of these have in common

They're not growing failures. They're recording failures: things that weren't written down, weren't dated, weren't attributed, weren't linked, or can't be trusted. Which is exactly why they respond so well to automation.

How this list collapses when records keep themselves

With GrowRoom, the environmental record is continuous by construction — there are no gap days, because sensors don't take Songkran off, and an excursion generates an alert and a visible spike in the batch record instead of a silence. Every note, input, waste entry and SOP completion is timestamped and attributed the moment it's entered; corrections live in an append-only audit trail, so there is nothing to scribble out and no way to backdate. The forms themselves ask for what auditors ask for — lot numbers on inputs, weights and witnesses on waste, categories on notes — so the small stuff is captured right the first time. Calibration logs carry their own next-due dates. Photos attach to the batch with dates. Deviations have root-cause and CAPA fields built in, and a failed COA sits in the record next to its quarantine notes and its passing retest.

Then the batch dossier assembles all of it — reconciled yield included — into one document, on demand. The audit stops being a test of your memory and becomes a printout of your season.

Want to see what a batch that hit real problems looks like when it's documented properly? The live demo includes one: mites, a humidity excursion, a failed microbial COA — deviation-managed, remediated, retested and released, with every step in the dossier.

Better yet, read the paperwork itself. These are the actual generated dossiers from the demo: DEMO-A-LOOSE (18 pages) — the batch that hit every problem above and still holds together as an audit record, because each failure has its deviation, CAPA, waste log and retest — and DEMO-B-TIGHT (14 pages), the clean first-pass batch. Notice what you don't see in either: gaps, scribbles, or anything reconstructed from memory.