Skip to content
SheetLink Forms

Privacy · 2026-09-03 · 8 min read · By Arden Talbot, founder of SheetLink

Deletion requests when your CRM is a sheet

Someone asks you to delete everything you hold about them. In a spreadsheet workflow this is a ten-minute job, provided you know where the copies are.

A ledger-styled illustration of a single row being ruled out of a register while a short note is added in the margin.

The email that starts it

Please delete all the data you hold about me. It arrives without warning, usually with no reference number, sometimes from an address that does not match anything you can find.

The panic is unnecessary. For a business whose leads live in a spreadsheet, this is a short task. The reason it feels alarming is that most people have never done it once, so there is no routine to fall back on.

Building the routine takes about an hour and makes every future request a ten-minute job.

It is also worth knowing that the volume is low. Most small businesses receive a handful of these in a year, often prompted by a person tidying up their own exposure rather than by any dissatisfaction with you. Treating the first one as an opportunity to build the process, rather than as an incident, is the difference between an hour spent once and an hour spent every time.

This is a working explanation for people who run websites, not legal advice. If your processing is unusual or high-risk, take proper advice.

First, verify who is asking

You should not delete records on the strength of an unverified email, and you should not demand a passport either. The standard is reasonable measures proportionate to the request.

For contact form data, replying to the address held on the record is usually sufficient and elegant: it confirms control of the address without collecting anything new. If the request arrives from a different address claiming to be the same person, ask for something that ties the two together, such as the date and subject of the original enquiry.

Be careful not to turn verification into an obstacle course. An excessive identity process is itself a compliance problem, and it reads as obstruction to the person on the other end.

Then work out what you actually hold

This is the step that takes the time, and it is the reason to do the mapping exercise before you ever receive a request.

Search the spreadsheet for the address and the name separately, because people submit under variations. Search your inbox, because notification emails are copies. Search any CRM, help desk or mailing platform. Check whether the person appears in an export that still exists.

The uncomfortable discovery for many businesses is how many places one enquiry ended up. That count is the real output of a first deletion request, and it usually leads to reducing the number of copies long before the next one arrives.

What you can keep

Erasure is not absolute. Data needed for a legal obligation, for establishing or defending legal claims, or for a contract still in force can generally be retained, and so can a minimal suppression record.

That last one matters practically. If you delete every trace of someone who asked not to be contacted, nothing stops them being re-added when they appear in a future import. Keeping a minimal record of the request itself, enough to honour it, is normally the right call.

Where you keep something, be able to say why. One sentence per category is enough: invoices retained for tax purposes, suppression entry retained to honour the objection.

What you cannot do is keep the record because it might be commercially useful later. That is precisely the interest erasure is designed to override, and citing it in a response is the fastest way to turn a routine request into a complaint.

Doing it in the sheet

The mechanics are dull, which is the goal. Filter for the person, confirm the rows are theirs, delete them, and note the action.

Two cautions. First, deleting a row in a shared sheet is immediately visible to anyone with access and recoverable from version history, so if the request concerns something sensitive, be aware that history retains it for a period. Second, if anything downstream references specific cells, deleting rows can break it, which is a good reason for downstream tools to reference the sheet rather than fixed positions.

A common refinement is to blank the identifying columns rather than delete the row, keeping the date and the source so monthly counts stay accurate. That satisfies erasure as long as the remaining data cannot identify the person.

The inbox is the hard part

Every notification email is a copy, and they are scattered across whoever was on the distribution. Searching your own mailbox is easy; being sure about a colleague's is not.

The practical approach is to search the shared mailbox thoroughly, ask individuals to search theirs, and reduce the problem structurally afterwards. A pipeline where the notification is a pointer rather than the record leaves far less to clean up, because the email says a submission arrived rather than reproducing all of it.

This is one of the quieter arguments for keeping notifications brief.

Backups, honestly

Backups will contain the data and cannot usually be edited. The accepted approach is to delete from live systems, record that the data remains in backups, and ensure it is not restored into live use.

Do not claim more than you have done. Telling someone every trace has been erased when a backup will hold it for another ninety days is a worse answer than explaining the position plainly.

If your backup retention is indefinite, that is worth fixing for its own sake, and it is the same conversation as retention on live data.

Respond within the month

The deadline is generally one month from receipt, extendable in limited circumstances for complex requests. Most contact form requests are not complex.

The response should confirm what was done, name anything retained and why, and state that the person can complain to a supervisory authority if unhappy. Three short paragraphs.

Keep a log of requests received and actions taken. It is the evidence that your process exists, and after the third entry it also becomes the training material for whoever handles the next one.

Turning it into a routine

  1. Verify the requester proportionately, ideally by replying to the address on file.
  2. Search every location on your data map, not just the spreadsheet.
  3. Delete or anonymise, keeping only what you can justify.
  4. Record what you kept and why.
  5. Reply within a month, plainly.
  6. Log the request.

Written down once, this stops being a compliance event and becomes a task.

The structural lesson

Every difficulty in a deletion request traces back to the same cause: the same enquiry existing in several places. The legal test is easy; the archaeology is not.

Which makes the strongest preparation an architectural one. When submissions land in one spreadsheet with notifications acting as alerts rather than copies, the answer to what do you hold about me is one search, and you can give it with confidence.

That confidence is worth more than the time it saves. The tone of a response to a data request is read closely by the person who sent it, and a prompt, specific reply defuses situations that a vague one escalates. Being able to say precisely what you held, what you deleted and what remains, on the day you receive the request, is the whole difference.

FAQ

How long do I have to respond?

Generally one month from receiving the request, with an extension available for genuinely complex cases. A contact form enquiry is rarely complex, so plan to answer well inside the month rather than relying on an extension.

Can I refuse a deletion request?

In limited circumstances, such as where data is needed for a legal obligation or for defending legal claims. Refusals must be explained, and blanket refusals because deletion is inconvenient are not defensible.

Do I have to delete the notification emails too?

They are copies of the same personal data, so yes in principle. This is the strongest practical reason to keep notification emails short and to treat the spreadsheet as the record.

What about backups I cannot edit?

Delete from live systems, ensure the data is not restored into use, and be honest that backups retain it for their retention period. Overstating what you have deleted is worse than explaining the limitation.

Is blanking the identifying columns enough?

If what remains genuinely cannot identify the person, directly or by combination, then it is no longer personal data. Be strict about that test: a free-text message mentioning a company and a role can still identify someone after the name column is cleared.

Can I keep a record that they asked to be deleted?

Keeping the minimum needed to honour the request is normally appropriate, otherwise the person reappears in the next import. Keep it minimal and do not use it for anything else.

What if I cannot find them?

Say so, describe where you searched, and ask for anything that would help locate the record such as the date or the address used. A clear account of a thorough search is a legitimate response.

Do I need a formal process document for this?

A page is plenty for a small business: who verifies, where to search, what may be retained, and who signs the reply. The point is that the next request does not depend on the same person being available, and that you can show a process existed rather than reconstructing one after the fact. Add a log of requests underneath it and the document maintains itself.

One place to look

When every submission lands in one spreadsheet, answering a data request is a search rather than an investigation.

Start freeSee the live demo

Data residency for form submissionsWhat happens when your form endpoint goes down