← All Resources

Scaling & Production

A Bad Production Run: The First 48 Hours and Finding the Cause

Molly Mills||11 min read
A shrink-wrapped pallet of plain cases set apart inside a square of yellow floor tape, with a blank orange tag hanging from the wrap and stainless kettles and a filling line behind it.

Put the product on hold first. Before anyone asks why the run came back wrong, make sure none of it can ship while you find out: the pallets still at the plant, anything at your own or a third-party warehouse, and any orders that have not left the dock. A held run costs you storage and patience. A run that ships while you are still deciding can cost a great deal more, and it turns a quality question into a traceability one.

You might hear about it from the plant's QC lead, from your own sample of the batch, or from a buyer who opened the first case. However it reaches you, the next two days decide whether you find a real cause or settle for a guess. Here is how I would spend them, how to tell a formula problem from an ingredient or process problem, what the fix looks like for each, and how to tighten the spec so the same run is less likely to happen again.

Hour One: Is This a Quality Miss or a Safety Question?

Some bad runs are quality misses: the color is off, the texture is thin, the flavor does not match the gold standard. Others are something else entirely. If any of the following is true, this is a food safety matter before it is anything else, and it goes to your co-packer's food safety team and your process authority now:

  • A pH reading above the limit in your process authority's letter or scheduled process, or a required reading that was never taken.
  • A cook temperature, fill temperature, or hold time that missed what the scheduled process requires.
  • Seal or vacuum failures, swelling, leaking, or off odors.
  • Foreign material found in product or on the line.
  • A wrong or missing ingredient, or a label that may not match what is in the jar, especially anywhere an allergen could be involved.

The plant will have its own procedure for a process deviation, and the process authority evaluates whether the product is safe and what, if anything, can be done with it. If any of that run has already shipped, bring in regulatory counsel too, have your traceability records ready to show where every case went, and let counsel and the food safety team guide what happens next. If those records turn out to be thin, running a mock recall before you need one is the fix for afterward.

None of that is formula development, and none of it is mine to run. Once the safety side is settled, the specialists may still conclude that the formula or spec contributed, and that is the part I help untangle. Everything below assumes a quality failure, or a safety question that has already been handled.

The First 48 Hours

Set the tone of the first call

Open with facts and an invitation, not a verdict: "The run from Tuesday doesn't match the gold standard on color and texture. We've put it on hold. Can we go through the records together?" If the plant's QC team flagged the problem before you did, thank them. That is the system working. A post-mortem that starts with blame tends to get defensive answers and thin records. One that starts with shared curiosity is far more likely to get you the operator's notes, the mid-fill readings, and the detail nobody thought to write down.

Gather the evidence while it still exists

Records get filed, retained samples get cleared out, and ingredient lots get used up. Ask early, and keep a copy in your own file:

  • The batch record for the bad run: actual ingredient weights, ingredient lot numbers, the kettle or line, cook and hold times and temperatures, order of addition, any rework or water added, and operator or shift notes.
  • The plant's QC data for that run: pH, Brix (a measure of dissolved solids, mostly sugar), viscosity, fill weights, and when during the run each reading was taken.
  • The same records for the last good run. The comparison tells you more than either record alone.
  • Certificates of analysis (the supplier's test results for each ingredient lot) for both runs, with any new supplier or lot flagged. How to read a certificate of analysis covers which numbers move a finished product.
  • Retained samples from the bad run and the last good run, plus your gold standard. If you have been getting a sample from every batch, pull yours too.
  • The spec version the plant actually ran. A perfect run of an outdated version is still a bad run for you, and it calls for a very different fix.

Taste and measure side by side, at the same age

Put the bad run, the last good run, and the gold standard on one bench at the same temperature, and measure pH, Brix, viscosity, salt, and color on all three. Taste them blind if someone else can pour, and taste the bad run only after the plant's food safety team has ruled out a safety question. Write down differences in specific words: "less garlic," "sharper acid," "thinner, with oil at the edge," "darker." "It's off" leads nowhere. If you do not have a gold standard yet, setting a gold standard and keeping retained samples is the place to start, because this comparison is weak without one.

Mind the age of what you are comparing. These products keep changing for about two weeks after they are made, and the settled version is the real one. A day-old run beside a gold standard approved at week two can look wrong when it is only young, and a run that looks close on day one can drift off target by week two. Keep it on hold and re-check at matched ages before you decide. That patience never applies to a safety reading. An out-of-limit pH is reported to the plant's safety staff and your process authority right away, not in two weeks.

Separating Formula, Ingredient, and Process Causes

Inside the product itself, a bad run usually traces back to one of three places, sometimes two at once. The records you just gathered are how you tell them apart.

Signs it is the ingredient

  • The bad run used a new ingredient lot, supplier, or crop, and the good run did not.
  • A certificate shows a number that moved (a paste at a different Brix, a vinegar at a different strength, a pepper lot with different heat), even inside the supplier's own range.
  • The defect is the kind an ingredient carries: heat, color, sweetness, acidity, aroma.
  • Making the formula on the bench with the suspect lot and the previous lot, if the plant still has some of each, reproduces the difference.

Signs it is the process

  • The batch record differs from the last good run: a longer hold, a different kettle, a new order of addition, rework blended in, water added to adjust, a stoppage mid-fill.
  • The QC readings drift during the run rather than sitting off from the start.
  • The defect is the kind a process creates: scorched or overcooked notes, over-reduction (Brix and viscosity both high), a broken emulsion, shredded or mushy particulates, a texture that thinned under extra mixing or pumping.

Before you call a process difference a plant mistake, read your own process notes. If they never said when the garlic goes in, how long a batch can hold before filling, or which kettle to use, the operator made a reasonable call on something you left open. That is a spec gap, not a plant failure. The plant fills those gaps in whatever way works on its line, which is reasonable. If a step matters to your product, it has to be written down.

Signs it is the formula

  • Nothing obvious changed: similar ingredient lots, a clean batch record, and the run still came out wrong.
  • Every number on the QC sheet was inside spec and the product is still wrong. Sit with that one. If every number passed and the product failed, the spec is measuring the wrong things or allowing ranges that are too wide.
  • The formula sits close to an edge: a pH near the top of what the process allows, a thickener near the limit of the heat and shear it can take, a heat level that depends on one variable ingredient. A formula like that runs fine while everything lines up and can fail when one small thing moves.

That last sign often means the honest answer is two causes at once. An ingredient moved a little, and the formula had no margin to absorb it. Fix only the ingredient and you are waiting for the next small move.

What the Fix Looks Like for Each Cause

Ingredient fixes

Specify the ingredient by what it has to do and what you can measure, not by brand name: the Brix range of the paste, the strength of the vinegar, the heat range of the pepper, the grind of the spice. Ask for the property that moved to be checked against its certificate at receiving. Where one ingredient keeps causing trouble, a more standardized form can help. For heat and color, an oleoresin (a concentrated spice extract) can hold steadier from lot to lot and over shelf life than ground spice. If the supplier itself is the problem, qualifying a second approved source is its own project, and it goes better before a shortage than during one.

Process fixes

Write the missing step into your process notes with a target and an acceptable range: the order of addition, when heat-sensitive ingredients go in, the longest a batch can hold before filling, which equipment the product runs on. Agree with the plant on an in-process check for whatever drifted, and on which reading means stop and call. Keep it collaborative. The plant knows its equipment far better than you do, and the best process notes are written with the production lead, not handed to them.

Formula fixes

A formula fix usually means building back margin. On pH, a formula that reaches its target only on a good day, with nothing spare for drift after the run, needs rebalancing. If texture failed, the thickener (a starch, or a gum or similar hydrocolloid) needs to be one chosen for the real kettle, the real acidity, and the real hold, not for a stovetop. If flavor or heat swung with an ingredient, the formula may need a steadier form of that ingredient or a correction step written into the spec.

Any formula change is a new version. It can touch the ingredient list, allergens, the nutrition panel, claims, and the process authority's review, so the people responsible for each need to see it before it runs. Then prove it at the plant on a small, controlled batch before you commit a full run, using the pilot batch protocol.

The held product

What happens to the bad run itself is a decision for you and the plant, and anything touching safety stays with the plant's food safety program and your process authority. Depending on the cause, product may be released if it meets spec and the defect is truly cosmetic, reworked where the plant's procedures and, for an acidified product, your process authority allow it, or destroyed. Rework has to be documented as a link between lots, or it quietly breaks your traceability. And do not let the cost of held inventory rush the root cause. Shipping a run you have not explained is how one bad run becomes two.

Write the Post-Mortem, Then Tighten the Spec

When you have a cause, put it on one page:

  1. What happened, with lot codes and dates.
  2. What the evidence showed, from the side-by-side, the records, and the certificates.
  3. The cause or causes, and how confident you are.
  4. What changes in the formula, an ingredient spec, the process notes, or the release checks, each with an owner and a date.
  5. How you will know it worked, and when you will check.

Write it about the system, not about people. The useful question is not "who missed it?" but "what in the spec let this happen, and what would have caught it sooner?" Then share it with the plant. A post-mortem the production lead has read and agreed with is one they can run against.

Finally, change the spec so the same miss has a much harder time slipping through:

  • Add the check that would have caught it to the finished product spec, with a target and a range, so it becomes a release decision instead of a surprise. Using the finished product spec as release criteria covers how targets, ranges, and holds fit together.
  • Narrow any range that let a wrong product pass, or start measuring the property that was missing.
  • Replace brand-name ingredients with functional, measurable ingredient specs.
  • Give every changed document a new version number, and confirm the plant has the new one and has retired the old.
  • Keep getting a sample from every batch. Quality control is mostly the plant's job, and it should be. A sample from every batch, tasted against your gold standard at the same age, is how you catch the next drift while it is still small.

If this run first reached you through customers rather than the plant, build a complaint log alongside the post-mortem. Reading customer complaints like a product developer shows how to tell one bad run from a pattern across many lots.

The Part That Does Not Fit on a Checklist

Any brand can do the steps above, and doing them carefully often gets you to an honest cause. The harder part is on the formula side: deciding how much margin a formula needs, which form of an ingredient will hold up on a particular plant's equipment, and how to rewrite targets and ranges so they protect the product without making the plant's job impossible. That calibration depends on the product, the plant, and what the run itself revealed, and it is the work I do with brands at the bench and on the spec sheet, one product at a time.

What you can do today is protect the evidence. Hold the product, keep the retains, get the records, and keep the conversation with your plant open. A run you can explain is a run you have a real chance of preventing.

Need Help With Your Formulation?

Whether you're scaling your first recipe or reformulating an existing product, let's talk about how to get it right.

Book a Free Discovery Call

Related Resources