Food Safety
Sampling during a food safety audit, the Battleship approach
by Erasmo Salazar • January 04, 2026 • 4 min read
“In the field of observation, chance favors only the prepared mind.” Louis Pasteur
Two auditors can walk the same site and reach different conclusions. One leaves with a report that is mostly descriptive, the other leaves with findings that genuinely help the organization. Same facility, same system, same people, same audit duration, different mindset.
Finding systemic nonconformities is not luck, it is your ability to sample with intent, pivot when the evidence shifts, and stay disciplined about collecting evidence.
Sampling during audits goes wrong in two common ways.
A. Random sampling becomes lazy sampling. You pull a few documents that are easy to access, you confirm the organization knows the requirements, and you mistake that for assurance and move on. (You see this on auditors that are only focused on their report).
B. Sampling becomes unlinked. You sample a procedure here, a record there, a training sign-in sheet somewhere else, but nothing connects. Samples need to be linked and come from the same trail, otherwise you cannot verify the process is working. (You see this on auditors who did not prepare well for the audit).
Battleship is a useful metaphor because it forces you to think in probabilities, patterns, and linked “shots,” not in random paperwork.
In Battleship, you are staring at a grid. The ships are hidden. Every move gives you one of two outcomes, hit or miss. The game is won by switching modes at the right time.
Once you get a hit, you stop hunting and you start targeting. You test adjacent squares to determine the ship’s orientation, then you concentrate until you sink it.
These are the steps of the Battleship approach.
- Build your grid before you fire
Decide what population you are sampling, and what the grid axes are, for example time, shift, line, product family, supplier, or lot. - Start with structured hunting
Maybe you saw a clue when you were reviewing the last management review record or the most recent food safety team meeting minute, and you want to investigate. Check what you want to check, not what the auditee prepared for you. - Treat the first hit as a hypothesis, then switch to targeting
When you spot a pattern, pivot fast. Targeting means narrowing the population, increasing density, and sampling adjacent points on the grid, same shift, adjacent dates, adjacent lots, same operator, same line, same ingredients, same control point. - Keep samples linked, follow the audit trail
This is where Battleship and audit trail thinking merge. Your best samples are linked because they prove, or disprove, how the process actually worked for a specific product, batch, day, and shift. - “Sink the ship” with triangulation
Before you call it a finding, you verify. You connect records, observation, and conversation, and you confirm that what is in front of you is a nonconformity.
Now an example: from “hunting” to “targeting” in an audit.
Let’s say you are auditing a food manufacturing process, and you are reviewing batch logs that include ingredient traceability information.
You begin in hunting mode. You sample records from last week, then the week before, across multiple shifts. You are not trying to prove a problem, you are trying to understand how the system behaves across time.
Then you notice a pattern: On the afternoon shift, two high-volume ingredients always show the same lot number day after day.
At that moment, stop hunting and start targeting.
You ask for records from three weeks ago, but only the afternoon shift, and you focus only on those two ingredients. If the same pattern repeats, you have enough to move to the next step which is asking questions about the pattern.
Why do these two high-volume ingredients always have the same lot number on shift 2 when the consumption rate is high?
For these two ingredients we sometimes see the same lot recorded on shift 3 and then again on shift 1 the next day, but we rarely, or never, see the same lot carry over from shift 1 into shift 2 on the same day. Can you explain this pattern?
Those questions often uncover the problem fast. Maybe an operator is recording an ingredient item code instead of the lot number. Maybe they are scanning a pallet code, a staging label, or a different barcode than the one tied to the true lot. Maybe shift handovers are bypassing a check, and the record looks “complete” while the control is weak.
Once you have confirmed the deficiency and captured the evidence clearly in your notes, you can explain the auditee you have a nonconformity, explain why and continue your audit with the same discipline.
Great audits do not add value by firing more shots. They add value by firing smarter shots.