How Long a Small Business Should Keep Customer Data
A nine person design studio I spoke to last year had a minor incident: one compromised email account, caught within a day, no money lost.

A nine person design studio I spoke to last year had a minor incident: one compromised email account, caught within a day, no money lost. The awkward part came during the clean-up, when they found a spreadsheet in a shared folder called clients-export-final.csv. It held 14,000 rows going back to 2017, including names, phone numbers and home addresses of people who had bought a single poster print and never come back.
Nobody had decided to keep that file. Someone had exported it once to fix a mailing list, and it had simply stayed. The incident itself was small. The thing that would have made it reportable and genuinely damaging was six years of data that nobody in the company needed, wanted or remembered.
Deletion is the cheapest security control you have
Every other control is a bet that your defences hold. Deletion is the only one that removes the thing being defended. You cannot leak, ransom, mis-send or subpoena a record you no longer hold, and unlike a password manager rollout or an endpoint tool, it costs nothing per month.
The law agrees in principle and then declines to give you a number. Under UK and EU data protection rules, the storage limitation principle says personal data should be kept no longer than necessary for the purpose it was collected for. Necessary is left to you to define. California's rules now require you to disclose your retention periods or the criteria you use, which again means you have to pick.
And pulling the other way are rules that require you to keep things. In the UK, HMRC expects business records to be kept for six years from the end of the accounting period. In the United States the general IRS window is three years, longer in several situations. So an invoice and a marketing list look identical in a folder and have completely different obligations attached.
That combination is why most small businesses end up keeping everything. It is the only option that never requires a decision.
A default schedule you can argue with
Here is what I suggest as a starting point for a business under about fifty people. Treat the numbers as defaults to be adjusted, not as legal advice, and check the specifics for your jurisdiction and sector.
| What you hold | Keep for | Why that long |
|---|---|---|
| Invoices, accounts, receipts | 6 years plus current year | Tax record requirements |
| Payroll records | 6 years | Tax and employment claims |
| CVs from unsuccessful applicants | 6 months | Discrimination claim window |
| Personnel files after someone leaves | 6 years | Contract claim limitation |
| Customer support tickets | 24 months | Useful history, then dead weight |
| Marketing list with no engagement | 24 months | Consent goes stale, deliverability suffers |
| Website and server logs | 90 days raw, 14 months aggregated | Enough for investigation |
| CCTV footage | 30 days | Standard guidance for premises |
| Call recordings | 6 months | Quality and dispute window |
| Card numbers | Never | Let the payment processor hold them |
The row that saves most businesses the most risk is the marketing one. A list of 40,000 addresses where 31,000 have not opened anything in three years is not an asset. It is a liability with a worse open rate attached, and cutting it usually improves your sender reputation at the same time.
Where the standard advice falls down
The usual recommendation is to write a data retention policy. I have read a number of these at small companies and most of them have the same defect: they describe deletion that has never once happened. The document sits in a shared drive, the data sits in the CRM, and the two have no relationship.
That is worse than having no policy. In an incident, a written policy you did not follow is evidence that you understood the obligation and did not meet it. A regulator asking how long you keep support tickets is not impressed by a paragraph. They want the date of the last deletion run and how many records it removed.
The second place this goes wrong is anonymisation. Teams reach for it because it sounds like a way to keep the data and the compliance. In practice, stripping names from a table that still contains postcode, date of birth and purchase history produces pseudonymised data, which is still personal data and still in scope. Genuine anonymisation means you cannot get back to a person even by combining it with something else, and that bar is higher than most exports clear. If you want the analysis, aggregate to counts. If you want the records, keep them properly and delete them on schedule.
And then there are backups, which quietly undo everything. I have seen a company delete a customer's record from the live database, correctly and promptly, while holding seven years of nightly snapshots containing that same record. Set backup retention to something defensible, typically 35 to 90 days for operational backups, and document that a restore followed by re-deletion is your process for a request that arrives mid-window.
How to run this without a compliance team
Build the inventory first, in a spreadsheet. One row per place data lives. Columns for the system, what sort of data, who owns it, the retention period, how deletion happens and the date it last ran. An hour of work for most small businesses, and the hour is mostly spent remembering tools you stopped using.
Hunt for the shadow copies. This is where the real exposure sits. Email attachments, one-off exports in cloud storage, a spreadsheet on a laptop that left with a contractor, the old CRM whose subscription lapsed but whose data is still there. The studio's 14,000 row file was in this category, and so is nearly every one I have found.
Use the retention settings you already pay for. Most business tools have them, and almost nobody switches them on. Workspace and Microsoft 365 both support retention rules on mail and files, chat tools can expire messages after a set period, and payment processors will tell you what they hold so you do not have to.
Put a thirty minute deletion run in the calendar quarterly. The first one will take two hours. Every one after that is short. Afterwards write one line in a log: date, what was deleted, how many records, who did it. That log is the only thing that makes the policy real.
Then update your privacy notice to match. Publishing the periods is increasingly expected anyway, and it has a useful side effect. Once the numbers are public, somebody will notice when the practice drifts.
When keeping it longer is correct
If there is a live dispute, a regulatory enquiry or a credible threat of a claim, stop the scheduled deletion for the affected records and write down the decision and the date. Deleting on schedule during a dispute looks much worse than keeping data too long.
Suppression lists are the other honest exception. If someone asks never to be contacted again, you need to keep something to honour that, and a one-way hash of the email address with no other fields is the usual approach. Keep that indefinitely and say so in the notice.
Finally, retention belongs in your joiners and leavers process rather than living on its own. When an account is closed, the exports that person made are part of the job, which is covered in our guide to offboarding an employee securely. And before adding another tool that will hold customer records for years, run it through the checklist for vetting a new online service. More of this sits in our business security section.
FAQ
How long should a small business keep customer data?
There is no single legal figure. Financial records usually need six years, support tickets and stale marketing contacts are fine at twenty four months, and card details should not be held at all.
Does deleting a record from the database count if it is still in backups?
Not fully. Keep operational backups to a defensible window of roughly 35 to 90 days, and have a written process for handling a deletion request that lands inside that window.
Can we anonymise data instead of deleting it?
Only if a person genuinely cannot be re-identified, including by combining your data with anything else. Removing names from a table that keeps postcode and purchase history does not reach that bar.
What proves we actually follow our retention policy?
A short log of deletion runs with dates and record counts. A policy document on its own shows you knew the obligation, which is the wrong half of the evidence to have.
More in Business Security
Business Security
A One Page Incident Response Plan for a Small Team
A twelve person agency I worked with lost ninety minutes on a Friday evening because nobody knew who could change the hosting...
Business Security
How to Secure a WordPress Site for a Small Business
Your small business website runs on WordPress, someone set it up a few years ago, and nobody has logged in since except to...
Business Security
How to Write a BYOD Policy for a Small Business Team
In most small businesses, half the work email is read on personal phones and nobody has written down a single rule about it.