Data Protection: Minimising, Mapping and Deleting
Map the personal details a workflow holds, flag fields with no purpose or keep time, check which saved records are overdue and plan consent and deletion habits.
In this lesson you will
- Write a data map for a workflow
- Explain purpose, minimisation and storage limitation
- Find records past their keep time
- Plan consent and deletion habits
About 25 minutes
“Why do you still have my phone number?”
A customer who asked for a quote in the summer emails Ravi to ask why the shop still holds her phone number, and what the AI step did with it. Ravi cannot answer. The automation copies each enquiry into a sheet, a log and a note for the assistant, and nobody ever decided how long any of that should stay. This lesson is a practical outline of how to answer such questions before they are asked. It is a guide to good habits and not legal advice, because the rules differ by country and by kind of business.
Three ideas that carry most of the weight
Data protection rules in many places, including the UK and EU, are built on principles that are easy to state. At the time of writing, three are especially useful for automations. Purpose: collect personal details for a reason you can name, and use them for that reason. Minimisation: keep to what is adequate, relevant and limited to what you need. Storage limitation: do not keep personal details longer than you need them. The UK regulator’s guidance says the law sets no fixed time limits. You decide how long you need each kind of detail, and you should be able to explain why. Keeping data ‘just in case’ is not a reason.
Make a data map
A data map is a plain list of the personal details your workflow touches. For each field it records why you have it, how long you keep it and whether it goes to an outside service such as an AI step. Writing this list is often the most revealing hour in a project. The program below holds a six-field map for the shop with made-up keep times, and flags fields without a purpose or a keep time and fields sent to the AI step. Then it applies a made-up 90-day keep time to five saved enquiries and lists which are past their date.
from datetime import date, timedelta
# the data map: (field, purpose, days to keep, sent to the AI step)
fields = [
("name", "address the reply", 90, False),
("email", "send the reply", 90, False),
("phone", "", 90, False),
("order_text", "write the quote", 365, True),
("address", "deliver the order", 365, False),
("birthday", "", None, False),
]
no_purpose = [f[0] for f in fields if not f[1]]
no_retention = [f[0] for f in fields if f[2] is None]
to_ai = [f[0] for f in fields if f[3]]
print("fields:", len(fields))
print("without a purpose:", no_purpose)
print("without a keep time:", no_retention)
print("sent to the AI step:", to_ai)
# delete after 90 days: which saved enquiries are past their date?
today = date(2026, 11, 1)
received = [date(2026, 7, 20), date(2026, 8, 5), date(2026, 9, 30), date(2026, 10, 15), date(2026, 8, 1)]
for r in received:
due = r + timedelta(days=90)
print(f"received {r:%d %b}: delete on {due:%d %b}", "(overdue)" if due < today else "")
print("overdue:", sum(r + timedelta(days=90) < today for r in received), "of", len(received))
fields: 6
without a purpose: ['phone', 'birthday']
without a keep time: ['birthday']
sent to the AI step: ['order_text']
received 20 Jul: delete on 18 Oct (overdue)
received 05 Aug: delete on 03 Nov
received 30 Sep: delete on 29 Dec
received 15 Oct: delete on 13 Jan
received 01 Aug: delete on 30 Oct (overdue)
overdue: 2 of 5
The map has 6 fields. Two have no stated purpose: phone, birthday. One has no keep time: birthday. Fields with no purpose are the first candidates to stop collecting, since collecting what you cannot explain is a risk with no benefit. The only field sent to the AI step is order_text, which the lesson 9 redaction step should have cleaned first. The deletion check used the date of the first of November 2026. 2 of the 5 saved enquiries were past their 90 days, for example the enquiry received on 20 Jul, which should have been deleted on 18 Oct. That check is exactly the kind of job an automation does well: a scheduled workflow can list overdue records every week for a person to confirm and delete.
Habits for automations that touch personal data
- Collect only what the workflow needs. If a field has no use, remove it from the form.
- Send an AI step the minimum, and redact the rest, as lesson 9 showed.
- Write down who can see the data, including any outside service, and read that service’s privacy information.
- Set a keep time for every field, and automate the reminder to delete.
- Make it easy for a person to ask what you hold, to correct it or to ask you to delete it, and know who answers.
- Keep logs free of private details, and keep them no longer than you need.
Consent is its own topic. If you rely on a person’s agreement, for example for marketing email, record when and how they agreed and make it as easy to withdraw as it was to give. Some data, such as payment card numbers and government identifiers, should not pass through casual automations at all.
Keeping everything because storage is cheap. Old data does not stay harmless. It goes out of date, it can leak and it must be explained. Decide a keep time, and automate the clearing.
The fields, keep times and dates here are made up. Data protection and privacy rules vary by country, sector and situation, and change over time. This lesson is an outline of good habits and is not legal advice. If you handle personal data seriously, read your regulator’s own guidance and take advice.
Practice for this lesson
Check what you learned
Five quick questions. Pick an answer and you will see straight away whether it is right and why.
What is a data map?
- A picture of the shop
- A log of errors
- A list of the personal details a workflow touches, with reason and keep time
- A list of passwords
It records why you hold each field and for how long.
What does minimisation mean?
- Keeping to what is adequate, relevant and necessary
- Collecting nothing at all
- Shrinking files
- Using shorter forms only
Too little data can fail the purpose, too much is excessive.
What does storage limitation mean?
- Limiting hard drive size
- Storing data twice
- Not keeping personal details longer than needed
- Never deleting
You decide how long you need each kind of detail and explain why.
Why flag fields with no purpose?
- Collecting what you cannot explain is a risk with no benefit
- They are always illegal
- They take no space
- They improve accuracy
They are the first candidates to stop collecting.
What job suits an automation here?
- Choosing the keep time alone
- Listing records past their keep time for a person to confirm
- Deleting everything silently
- Ignoring consent
A weekly overdue list is easy to automate; a person confirms.
Summary
- Collect for a stated purpose and only what you need.
- Set a keep time for every field and explain it.
- A data map shows what an outside service sees.
- This is an outline, not legal advice.
Questions
How long should I keep personal data?
There is no single answer. In the UK the guidance leaves it to you to decide how long you need each kind of data and to be able to explain why. Check your own rules.
Do these rules apply to a small shop?
Rules about personal data often apply to businesses of any size, but details vary by place. Check your regulator's guidance for small businesses.
Is this legal advice?
No. It is an outline of good habits with made-up examples.
