Stability testing and shelf life explained, where that 24 months comes from

Every supplement pack carries a best-before or expiry date. Most new brand owners assume it is a convention, that 24 months is simply what everyone writes.

It is not supposed to be a convention. It is supposed to be a conclusion from data, and the difference matters when a buyer, a regulator or an importing country asks you to show the data.

So this article covers what stability testing actually is, the two kinds, what each is good for, which one your situation needs, and the specific things it will not tell you.

This is orientation, not legal or regulatory advice. Shelf-life substantiation requirements, labelling rules and import registration requirements change and differ by market. Verify the current position with a competent regulatory consultant.

What stability testing is

You put the finished product, in its final pack, in a controlled chamber, and you test it at intervals against the specification you claim.

Three words there are doing all the work.

Finished product, not the extract, not the blend. The product.

In its final pack, because the pack is part of the result. The same powder in a PET jar with a desiccant and in a paper pouch have different shelf lives.

Against the specification you claim, which means if your label says 5% withanolides, the test is whether it is still 5% at month 18, not whether the capsule still looks like a capsule.

What is actually measured

And this is where most brands are surprised, because they expect a single pass-or-fail.

Assay of the actives you declare, withanolides, curcuminoids, boswellic acids including AKBA, bacosides, whatever your label names. This is the one that decides the number.

Physical appearance, colour, odour, caking, capsule shell integrity, tablet hardness and friability, phase separation in liquids.

Moisture content, which drives most of the degradation in herbal powders.

Disintegration or dissolution for tablets and capsules, a tablet that hardens over time can stop disintegrating within specification even though its assay is fine.

Microbiology, total plate count, yeast and mould, and specified pathogens. This often fails before the assay does in high-moisture or syrup products.

pH for liquids, and peroxide or acid value for oil-based products, which is how rancidity shows up.

And leachables or pack compatibility if the pack is new, particularly for liquids in plastic.

Real-time stability

The honest method. You store the product at the conditions it will actually see and test it at intervals until you reach the claim.

Typical pull points: 0, 3, 6, 9, 12, 18, 24 months.

The obvious problem: to substantiate 24 months you wait 24 months.

Which is fine for a product you already sell, and impossible for a launch.

Accelerated stability

The workaround. Store at elevated temperature and humidity, test over a few months, and extrapolate.

The underlying idea is that chemical degradation speeds up with temperature in a reasonably predictable way, so a short exposure to harsh conditions approximates a long exposure to normal ones.

Hence the rule of thumb that circulates in this industry: six months accelerated supports a provisional 24-month shelf life.

Treat that rule of thumb with real caution, for three reasons.

First, extrapolation assumes the degradation mechanism is the same at both temperatures, and for botanicals it often is not. Heat can trigger reactions that simply do not occur at ambient conditions, so you get failure that overstates real-world risk; and conversely, slow processes like microbial growth or moisture migration through a pack seal can be understated by a short hot test.

Second, physical changes do not extrapolate at all. Caking, colour change in a natural pigment, and oil separation follow their own logic.

Third, whether a regulator or an importing country accepts accelerated data for your product class is a question with a specific answer, and the answer differs by market. Several registration regimes require real-time data, or accept accelerated data only provisionally while real-time runs in parallel.

Which you need

And in practice most brands need both, sequentially.

For a domestic launch: accelerated data to support a provisional shelf life, with real-time running in parallel on the launch batch. If real-time contradicts the extrapolation at month 12 or 18, you shorten the claim on subsequent artwork.

For export registration: ask the buyer or the importing authority what they require before you commission anything, because a study run to the wrong protocol is a study you pay for twice. Real-time data is commonly required and commonly the long pole in a registration timeline.

For a reformulation of a product you already sell: you need new data. A changed excipient, a changed capsule shell, a changed supplier for an extract, or a changed pack can all change stability, and the old study does not transfer.

For a pack change alone: you still need data, because the pack is part of the result.

What it costs you, in the things that are not money

We do not publish figures, because testing costs vary by lab, scope and number of pull points, and a number in an article dates badly.

But there are three non-monetary costs worth planning for.

Time. Real-time data cannot be bought faster than it accrues. If your export registration needs 12-month real-time data, that is 12 months, and no supplier relationship changes it.

Batch commitment. A study consumes product from a specific batch, held in reserve. That batch is now partly unavailable for sale, and if you change anything you start again.

And decision risk. A study can tell you your 24-month claim is not supportable, which means new artwork, a shorter claim, a reformulation, or a better pack. Brands sometimes avoid testing because they suspect this, which is the worst available reason not to test.

The shelf life you print is not the shelf life you get

And this is the part most worth internalising.

A stability study is run at controlled conditions. Your product will not live there.

It will sit in a warehouse without climate control through a north Indian summer. It will travel in a container across the equator. It will stand on a shelf in a shop with a glass front in direct sun. It will be opened by a customer who leaves the cap loose in a Mumbai monsoon.

Some of those are addressed by storage conditions on the label, which is why those lines exist and why they should be specific rather than decorative.

And some are addressed by the pack. An induction seal, a desiccant, an opaque bottle and a moisture-barrier film are not cosmetic choices, they are the mechanism by which the number you printed survives contact with reality.

We have written separately about packaging choices that generate returns, and most of those returns are stability failures that the pack could have prevented.

Questions to ask your manufacturer

Specific enough that a vague answer is visible:

“Do you have stability data for this exact formulation, in this exact pack?” Not a similar one.

“Real-time, accelerated, or both? For how long?”

“Which parameters were tested at each pull point, and can I see the data rather than a summary letter?”

“What specification was the assay measured against?” If the study checked appearance and microbiology but never assayed the active, it does not support a potency claim.

“Are retention samples held, and for how long after expiry?”

“If I change the pack or an excipient, do you re-run it, and at whose cost?”

And: “Who holds the data if I move manufacturer?” The answer to that belongs in your agreement at the start.

What stability data cannot tell you

Being fair about the limits, because stability data gets over-read.

It does not tell you the product works. A stable product is one that is still what the label says. Whether what the label says does anything useful is a completely separate question, and no stability study addresses it.

It does not cover a different batch. Stability runs on specific batches; routine release testing is what tells you each batch left the plant to specification.

It does not cover misuse or abuse. A bottle left in a car in May is outside every study ever run.

And it does not substitute for per-batch testing. A good stability file and a poor release process still ships a bad batch.

The short version

1. The expiry date is meant to be a conclusion from data, not a convention
2. Stability testing means the finished product, in its final pack, against your claimed specification
3. The pack is part of the result, same powder, different pack, different shelf life
4. If the label claims 5% withanolides, the test is whether it still is
5. Measured: assay, appearance, moisture, disintegration, microbiology, pH, peroxide value
6. Microbiology often fails before the assay in high-moisture products
7. Rancidity shows up as peroxide or acid value in oil-based products
8. A tablet can harden out of disintegration specification while its assay is fine
9. Real-time: store at actual conditions, pull at 0/3/6/9/12/18/24 months
10. Honest, and impossible for a launch, because you have to wait
11. Accelerated: harsh conditions, short duration, extrapolate
12. The six-months-supports-24 rule of thumb deserves real caution
13. Extrapolation assumes the same degradation mechanism at both temperatures
14. For botanicals it often is not, heat triggers reactions ambient does not
15. And slow processes can be understated by a short hot test
16. Physical changes like caking and colour shift do not extrapolate at all
17. Acceptance of accelerated data differs by market and product class
18. Most brands need both: accelerated to launch, real-time in parallel
19. If real-time contradicts the extrapolation, shorten the claim on new artwork
20. For export, ask what protocol is required before commissioning anything
21. Any reformulation, supplier change or pack change needs new data
22. Three non-money costs: time, batch commitment, decision risk
23. Avoiding testing because you suspect the answer is the worst reason not to test
24. The printed shelf life assumes controlled conditions your product will never see
25. Warehouse heat, container transit, sunlit shelves, a loose cap in the monsoon
26. Storage lines on the label exist for this and should be specific
27. Seal, desiccant, opaque bottle and barrier film are the mechanism, not decoration
28. Ask whether the data is for this exact formulation in this exact pack
29. Ask to see the data, not a summary letter
30. A study that never assayed the active does not support a potency claim
31. Ask who holds the data if you move manufacturer, settle it in the agreement
32. Stability does not tell you the product works, only that it is still itself
33. And it never substitutes for per-batch release testing

Related guides: how to choose a contract manufacturer, supplement costing and margins, supplement label compliance in India. Related reading: real-time vs accelerated stability for registration, packaging choices that cause returns, reading a certificate of analysis, and expiry, batch and storage, the boring but useful bits.