Throw It Away: The Case for Deleting Data on Purpose

Somewhere in your organisation right now, there's a server quietly holding a spreadsheet nobody has opened in three years. Nobody remembers what it was for. Nobody's job depends on it. And nobody, when asked, will agree to delete it.

That's the strange contradiction sitting at the heart of most companies' data strategy. We talk about data as the new oil, the new gold, the most valuable asset on the balance sheet. And yet walk the floor of most analytics teams and you'll hear the same complaint: most of what we're sitting on is rubbish. It's out of date. It's inconsistent. Nobody trusts it enough to build a decision on it. It just sits there, taking up space, because getting rid of it feels riskier than keeping it.

It isn't. Keeping data you don't use and can't trust is the risk. And it's time more businesses treated deletion as a strategy, not a failure.

The hoarding instinct has an alibi, and it's a bad one

Ask anyone why a dataset is still sitting in storage and the answer is almost always some version of "it might be useful one day." It's a reasonable-sounding sentence. It's also doing a lot of quiet work to avoid a harder question: useful for what, exactly, and to whom?

The honest answer, most of the time, is nobody knows. Nobody can point to a decision the data feeds, a report it powers, or a person who'd notice if it vanished tomorrow. It was collected because collecting it was easy, or because a system defaulted to logging it, or because someone three reorganisations ago thought it might matter. "Might be useful one day" isn't a plan. It's the absence of one.

A recent report puts a number on how far this goes: organisations are conservatively estimated to be leaving 55% of their data dark, collected, stored, and never touched again. Dig into the same survey and 3 in 10 of the c-suite call dark data a serious risk to their organisation. It's worth sitting with what that actually means: seven in ten leaders are, in effect, saying that not knowing what's happening to most of their data is fine.

Storage got cheap. Judgement didn't.

Part of why this happened is simple economics. Storage used to be expensive enough that someone had to justify keeping anything. That constraint is largely gone. A terabyte costs a fraction of what it once did, so the instinct to be selective went with it. But "cheap to store" was never the same question as "worth keeping," and treating them as interchangeable is how organisations end up with digital landfills instead of data assets.

The report puts real figures against that landfill. Organisations with 250–1,000 employees are estimated to store anywhere from 20 to 200 terabytes of data a year. Using the lower end of the dark-data estimate, that works out to somewhere between £26,400 and £264,000 a year spent storing data that never gets used, for the average organisation, and considerably more for data-heavy sectors like financial services or healthcare. That's not a hypothetical cost. It's a line item, recurring, for an asset nobody's using.

And storage is the cheap part of the risk. Unused, unmanaged data still has to be secured, still falls under GDPR, and still shows up as a liability the moment a regulator or a hacker comes looking. 94% of the c-suite in the same research say it's imperative they have access to accurate, complete, up-to-date data for decision-making. But every dataset kept "just in case" that nobody maintains is quietly working against that goal, diluting the pool of trustworthy information with data nobody's kept clean.

What "throw it away" actually means

None of this is an argument for recklessness. Some data has a genuine, ongoing reason to exist: regulatory retention requirements, a model that's actively trained on it, a report someone actually reads. The point isn't to delete indiscriminately; it's to stop keeping things by default and start keeping things by decision.

That means asking, dataset by dataset: is this trusted? Is it tied to something an actual person actually does? Does keeping it create more exposure than value? If nobody in the business can answer those questions with a straight face, that's the answer.

Deletion, done deliberately, isn't data loss. It's data hygiene. It's the discipline of making sure that when someone in your organisation reaches for "the data," what they find is something they can actually trust, not one honest dataset buried under nine dubious ones.

Where this leaves you

If your organisation has never run an exercise to find out how much of its data is genuinely dark, that's usually the first sign it's carrying more than it realises. It's not a small undertaking to map what exists, where it sits, who owns it, and whether it's still earning its keep. But it's considerably smaller than the bill for not doing it.

If "we might need it one day" is doing a lot of load-bearing work in your organisation's data strategy, it's worth asking what that "one day" is actually costing you, because going by the numbers above, the honest answer is: a lot more than anyone's bothered to check.

This is the first in a short series on why more data isn't always better data. Next: what happens when AI is trained on the data nobody trusted in the first place.