Six situations where you should not buy this

Every vendor page on this site argues for buying the product. This one argues against it in the cases where it should not be bought, partly because it is true and partly because a pilot that was never going to work costs us more than the deal was worth.

The cost nobody puts in the business case

A storage saving is easy to model. The other side of the ledger is not, and it is usually larger in year one:

• Migration compute, and the calendar time to run and validate it.
• Dual-running while you satisfy yourself nothing was lost.
• Every script, dashboard and runbook that touches the data.
• People learning something new for an infrastructure layer that was previously invisible.
• A new dependency in a path you will rely on for a decade.

None of that appears in a ratio. All of it appears in your quarter.

The six cases

1. Storage cost is your only requirement. If you never need to read the archive again and nobody will ask questions about its integrity, a general-purpose compressor at maximum settings gets close enough. On some database corpora we land near xz on size alone. Paying for verification and queryability you will not use is a bad trade, and we would rather say so before the pilot than during the renewal.
2. Your data is mostly high-entropy. Hashes, UUIDs, encrypted blobs, already-compressed media. There is no structure to model, so we floor out at roughly what a general compressor achieves. A tool claiming a large win on random data is broken.
3. The volume is small. A 27% saving on 400GB is not a project. Percentage improvements need scale before they become money, and below a few hundred terabytes the engineering time usually costs more than the storage does.
4. It is photographs or video. WebP and x265 beat us clearly and it is not close. We win on 16-bit medical, depth and astronomical imaging, where lossless is required. For ordinary media, use the media codecs.
5. You are mid-migration on something else. A format change during a warehouse move or a cloud migration is a compounding risk with a shared blast radius. The saving will still be there in six months.
6. Nobody has asked you a question you cannot answer. This is the real test, and it is the last one for a reason. Our advantages are queryable-in-place, verifiable, erasable. If no auditor asks whether records were altered, no data subject asks to be deleted, and nobody needs to search the archive without restoring it, then those properties are features you are buying and will not collect.

The buying signal is a question you cannot currently answer — not a storage line that looks large.

When it is worth it

The inverse of the sixth case, essentially. The organisations that get real value have a specific problem they have been working around, usually one of three: an auditor asking for proof records were not altered; a legal team asking to delete one person from something that cannot be edited; or an analytics team quietly maintaining a second copy of cold data because the archive cannot be searched.

In each case the storage saving is the smaller half. It also means the pilot has a clear success criterion — the question gets answered or it does not — instead of a percentage that everyone stares at, unsure what it was supposed to prove.

Why publish this

Because the alternative is worse for us. A customer who buys on a storage figure, migrates, and discovers the saving did not clear the switching cost is a bad reference, a renewal problem, and a story they tell other people in their industry. A prospect who reads this and decides not to proceed costs us nothing and occasionally comes back two years later with an auditor’s question and a budget.

It also keeps us honest in the other direction. Writing down where we should not be bought makes it much harder to quietly widen the claim when a quarter is looking thin.