TL;DR. A dairy line spends hours each week changing over, and most of that time is non-negotiable. The loss is the part beyond the agreed standard, plus the waiting around it. Split required time from excess, and you stop chasing hygiene and start chasing waits.
Why does changeover distort loss reports?
If every changeover minute is counted as downtime, the numbers look catastrophic and nobody believes them. If every minute is treated as planned, the numbers look fine and nothing improves.
Both versions are wrong. A changeover contains some time you must spend and some you do not.
The honest approach is to record it in parts.
What is the required time?
Required time is whatever food safety, quality and legislation demand. Cleaning cycles, rinses, allergen checks and verification sign-offs fall here.
In dairy, this is not negotiable and should never be treated as a target for cutting. Shortening a cleaning step to improve a metric is a quality incident waiting to happen. Validated cleaning parameters change only through formal change control, never to improve a metric.
Set the required time with the quality team, write it down per changeover type, and treat it as fixed until the process itself changes.
What is the recoverable part?
Illustrative example: a fictional dairy has a standard of 90 minutes to change a line from plain to flavoured yoghurt. Actual changeovers take 90 to 150 minutes, with an average of 118.
The 28 minutes of average overrun is where the opportunity sits. It is often not the cleaning itself but the waiting around it.
- Waiting for quality to be told cleaning has finished, or for a quality person to be free. The verification itself is required time.
- Waiting for the next product’s packaging.
- Waiting for a person who is on another line.
- Searching for a tool, a gasket or a hose.
- Delayed start because the previous run finished late.
None of these touches hygiene. All of them can be fixed with planning.
How much can this be worth?
Illustrative example: 14 changeovers a week, an average overrun of 28 minutes, 50 working weeks. That is 14 × 28 × 50 = 19,600 minutes in a year.
At an assumed $20 per minute of line contribution, that is $392,000 a year of gross exposure, not a saving. Treat it as an estimate. A fix will not recover all of it, and finance will want it confirmed on comparable changeovers.
The point of the example is the shape: a modest overrun, repeated often, becomes a large number. Read three tiers of money before you quote it.
How should you record a changeover?
Split it into at least three timestamps.
- Production stopped.
- Cleaning completed and handed to quality.
- Quality release given and production restarted.
Add a reason code for any wait that exceeds a few minutes. That gives you the required time, the wait for quality to attend and the restart delay, each with its own owner.
Many plants record only the first and last. That hides everything useful in the middle.
Who owns each wait?
This is where changeover improvement usually succeeds or fails. A wait with no owner persists.
Any wait for quality to attend belongs to the quality lead. The verification step itself stays as required time. The packaging wait belongs to the planner. The late finish of the prior run belongs to the line supervisor. Name them.
A single changeover owner for the whole plant can coordinate, but each wait still needs someone who can act on it.
How do you check an improvement is real?
By comparing like with like. A changeover between plain and flavoured is not the same as one between nut-containing and nut-free product, which is far longer.
Compare similar changeover types, not monthly averages. If the mix of types changed, the average can improve without any real gain.
What to do on Monday
- List your changeover types and agree a standard time with quality for each.
- Start recording the three timestamps on your busiest line.
- Add a short list of wait reasons.
- Name an owner for the largest wait.
- Review overrun by type, not by month, in four weeks.