Consent by design: email capture, lead forms and onboarding under UK GDPR and PECR
TL;DR
Difficulty: Beginner
In the UK, PECR decides whether you may send marketing email, and the UK GDPR decides what valid consent is.
- Non customers need consent: an unticked box, naming you and the channel, never bundled with a download.
- Customers and trial users can be emailed under the soft opt in, if you offered a simple opt out right where you took the address.
- Since 5 February 2026 charities have their own soft opt in, and PECR fines reach £17.5 million or 4% of turnover.
- Keep a record of every consent with the exact wording, and check a suppression list before every send.
Most sign up forms collect addresses that cannot lawfully receive a single marketing email. In this article I show what the UK GDPR and PECR actually require, what the Data (Use and Access) Act 2025 changed, and how I rebuild popups, checkouts, gated guides and onboarding so the list you build is one you can use.
Two laws decide whether you may email someone, not one
Most people think a sign up form follows one rule, called GDPR. In the UK there are two.
The UK GDPR (the General Data Protection Regulation as it applies in the UK) decides what consent is and what makes it valid. The Privacy and Electronic Communications Regulations, PECR for short, decide when you need that consent before you send marketing by email, text or private message. Regulation 22 of PECR is the one that matters here.1 Both are enforced by the Information Commission, and both were changed by the Data (Use and Access) Act 2025, with the parts that matter here in force since 5 February 2026.12 The same Act turned the Information Commissioner's Office (ICO) into the Information Commission on 30 September 2026, and its website and guidance still use the name ICO, so I do too.3
So every form answers two questions. Under PECR: may I send this person marketing email at all? Under the UK GDPR: if I rely on their consent, is it real, and can I prove it?
I am an engineer who builds these forms, not a solicitor. What follows is what the law and the ICO say, and how I turn it into forms that work. Beyond the basics, ask a solicitor.
Worked example: one popup, two questions
Kettle and Leaf is an invented online tea shop in Bristol, run by Amara, at kettleandleaf.example. Nobody real, and every number in this article is made up so I can show each step.
Her site has a popup: "Join the club for 10% off your first order", one email field and a button. Underneath, in small grey text: "By signing up you agree to our terms and privacy policy."
The PECR question: a weekly email about new teas is marketing, and most of these visitors have never bought anything, so she needs their consent. The UK GDPR question: is pressing that button consent to marketing email? The text never mentions email or marketing and ties the sign up to her terms. So the popup collects addresses she cannot lawfully email: the work of collecting, none of the value.
Valid consent has to pass five checks, and all of them happen on the form
The UK GDPR defines consent as a "freely given, specific, informed and unambiguous indication" of someone's wishes, given "by a statement or by a clear affirmative action".4 Since February 2026 that definition sits in Article 4(1)(11), because the 2025 Act renumbered Article 4. The only change to the definition itself is a cross reference to a new rule about scientific research, which does not touch marketing.4
Article 7 adds the conditions around it.5 You must be able to show the person consented. A request inside a longer document must stand apart, in plain language. People can withdraw at any time, must be told so, and withdrawing must be as easy as agreeing. And consent is not freely given if something the person wanted was made to depend on a consent it did not need.
The ICO adds practical tests. Silence, boxes ticked in advance and inactivity are not consent. Consent sits apart from your terms, with separate options for separate purposes, and names your organisation and any third party relying on it.6 For email marketing in particular, the ICO says the request should name you, say it covers email marketing, and be asked separately for each channel, because a general consent to "direct marketing" is not specific enough.7
I turn the law into five checks, each a question the form must answer yes to:
- Freely given: could the person say no and still get what they came for?
- Specific: does it say this business, this channel, this kind of message?
- Informed: do they know who you are and that they can withdraw?
- Unambiguous: did they do something deliberate, like ticking an empty box?
- Provable: could you show who agreed, when, how, and to which wording?
A box that arrives already ticked fails the fourth question on its own. That one is so common it has its own article on dark patterns.
Worked example: rebuilding Amara's popup
Freely given: the popup exists for the email list and the code is the reason to join, so nothing unrelated is held hostage, provided the form says plainly that the club is a marketing email. Specific: "agree to our terms" says nothing about email. Fail. Informed: no business name in the consent line, nothing about withdrawing. Fail. Unambiguous: the press is deliberate, but it agrees to terms, not to marketing. Fail. Provable: the shop stores the address and a date, nothing else. Fail.
The rebuilt popup reads: "Get 10% off your first order, plus a weekly email from Kettle and Leaf about new teas. Unsubscribe any time with one click." One email field, one button, "Send me the code and the weekly email". The button now names exactly what happens, so pressing it is a clear affirmative action for one specific thing, and the shop records the wording and the time with the address.
The soft opt in: people who bought, or nearly bought, can be emailed without a tick
PECR has one long standing exception to the consent rule for businesses, known as the soft opt in. Regulation 22(3) lets you email someone without their consent if all of this is true: you got their contact details yourself in the course of a sale or negotiations for a sale, you are marketing your own similar products or services, and you gave them a simple, free way to refuse when you collected the details and in every message since.1
The ICO's guidance fills in the details, and they matter.7 The details must come directly from the person, not from a list, and not even from another company in your group. Negotiations means the person actively showed interest in buying, and the ICO says this includes signing up to a free trial or asking for a quote. Browsing your site is not enough. Similar means what people would reasonably expect from what they bought, and you cannot use it to promote someone else's products.
The opt out at collection is where most shops trip. The ICO says one hidden in your privacy policy is not simple, and one first offered later, say in the order confirmation, is too late.7 It belongs right where you take the address. Every later message then needs an unsubscribe that does not make people log in.
I treat it exactly as in the article on data warehousing for marketing automation: the same three conditions, carried as a permission column per person and checked inside every query that builds a mailing list.
The soft opt in only answers the PECR question. Under the UK GDPR you still need a lawful basis for using the data, and where the soft opt in applies, the ICO says that basis is likely to be legitimate interests, ideally with a short legitimate interests assessment written down.8
Worked example: the checkout and the free trial
At Kettle and Leaf's checkout, under the email field, there is an unticked box in normal sized text: "Please don't send me emails about new teas and offers. You can also unsubscribe from any email." Tom, an invented customer at tom@example.com, buys a tin of Assam and leaves the box empty.
Can Amara email Tom next week about her new Darjeeling? The details came from Tom, during a sale. Darjeeling is her own tea, a similar product. He was offered a simple refusal when she took his address, and every email carries an unsubscribe. Yes. Can she email him about the deli next door's cheese hampers for a commission? No. It is not her product.7
Now a software example. Rotawise is an invented staff rota tool in Manchester. Priya, at priya@example.com, starts a 14 day free trial. On the sign up screen there is the same kind of unticked opt out box for product news. The ICO counts signing up to a free trial as negotiations for a sale, so Rotawise can email Priya about its own features and plans under the soft opt in, as long as every email carries an unsubscribe.7 If Priya had only read the pricing page and left, there would be nothing to rely on.
One point for software sold to businesses: the ICO says you can email corporate bodies, such as limited companies, without consent, but sole traders and some partnerships count as individuals, and you should keep a "do not email" list.9 The UK GDPR still protects the named person behind a work address.
What the 2025 Act changed, and what it left alone
The Data (Use and Access) Act 2025 changed three things that touch email capture. All three apply from 5 February 2026.
First, charities gained a soft opt in of their own. New regulation 22(3A) lets a charity email people without consent if the sole purpose is to further its charitable purposes, and it got their details when they showed interest in those purposes or offered or gave support. The same simple, free refusal applies at collection and in every message.1 The ICO's guidance adds firm limits. It only covers details collected on or after 5 February 2026. It does not cover purchases that are incidental or just for convenience, the ICO's example being a drink bought at a charity café. And it cannot be used to promote anyone else, not partner charities, not commercial sponsors.710
Second, fines. PECR is now enforced through the penalty rules of the Data Protection Act 2018, and regulation 22 is on the list of provisions that carry the higher maximum: £17.5 million or 4% of worldwide annual turnover, whichever is higher.2 Email marketing breaches used to be a far smaller risk on paper. They are not any more.
Third, Article 6 of the UK GDPR now names direct marketing as an example of processing that may be necessary for legitimate interests.11 That is easy to misread as "no consent needed for marketing email". Article 6 is about the lawful basis for using the data. PECR still decides whether you may send the email, and for people who are not customers that still means consent.7
The commercial soft opt in in regulation 22(3) is unchanged, and so is the definition of consent for marketing.14 The ICO marks several consent pages as under review because of the Act, so I recheck them before relying on a detail.6
Worked example: a rescue in Brighton
Downs Hedgehog Rescue is an invented registered charity in Brighton, planning a spring appeal. Sam, at sam@example.com, signed up as a volunteer through the rescue's own website in March 2026, and the form had an unticked box to opt out of emails about the rescue's work. Sam's details came directly, after 5 February 2026, from someone offering support. The rescue may email Sam about the appeal under the charitable purpose soft opt in.
Jo, at jo@example.com, gave an email address for a receipt when buying a coffee at the rescue's café. That is a purchase made for convenience, so it does not count. The rescue needs Jo's consent.
Alex donated in December 2025, before the new rule began. The rescue cannot rely on the new soft opt in for Alex's existing details. It needs consent, or a fresh interaction after 5 February 2026 that collects the details again with the opt out offered.7
And an email promoting the garden centre that sponsors the rescue? Not under the soft opt in, for anyone.
The free guide must not cost a marketing consent
The commonest lead generation pattern I see is the gated guide: enter your email to download it, and by doing so you join the newsletter. It feels like a fair trade. The law reads it differently.
Article 7(4) asks whether something the person wanted was made conditional on a consent it did not need.5 The ICO is direct about it: consent is not valid where it was a precondition of a service but the processing is not necessary for that service.6 Sending someone a PDF does not need their consent to a weekly newsletter. So the newsletter consent collected this way is not freely given, and the soft opt in does not rescue it either, because downloading a guide is not negotiating to buy.7
Two clean ways to build it. Email the guide, which needs the address, with a separate unticked box for the newsletter. Or, as I often prefer, let people download it with no email at all and invite them to the newsletter on the last page. Fewer addresses, every one from a person who asked.
Worked example: Amara's brewing guide
Kettle and Leaf offers "The Bristol guide to brewing loose leaf tea" as a PDF. The old form said: "Enter your email to get the guide. You'll also join our newsletter." In a month, 1,200 people downloaded it.
None of those 1,200 newsletter consents is freely given, because the guide was conditional on them, and downloading a guide does not make anyone a customer. So Amara can lawfully send marketing to none of them. She can send each person the guide, and that is all.
The rebuilt form: one email field, a button that says "Email me the guide", and beside it an unticked box: "Also send me the weekly Kettle and Leaf email about new teas. Unsubscribe any time." Same month, same 1,200 downloads, and 456 people tick the box. Those 456 are a list she can use.
A consented list is smaller, and worth more
The usual objection is the drop in sign ups. It is real. What matters is the usable list, and what it earns.
The share of people who say yes is a rate like any other:
Here is the opt in rate, is the number of people who ticked the box, and is the number of people who saw the form and submitted it. It only means something if you count and the same way every month, which is the point of Foundations: ratios and rates.
To put a value on a subscriber, I use expected value: the chance they order in a month, times the margin an order brings.
Here is the expected margin in pounds per subscriber per month, is the probability that a subscriber places an order in a month after receiving the emails, and is the margin in pounds on an average order. It is the simplest case of the expected value formula, with one outcome that pays and one that pays nothing, explained from scratch in Foundations: expected value.
Worked example: what the 456 are worth
From Amara's rebuilt form:
Here is , the people who ticked the box, and is , everyone who downloaded the guide. So 38% of them also asked for the email. On the old form the collected number was 1,200, but the usable number was zero, so the comparison that matters is 456 against 0.
Now the value. In Amara's (invented) records, 4% of subscribers place an order in a month, and her average order brings £14 of margin after the tea, the packaging and the postage.
Here is and is in pounds, so is 56p per subscriber per month. Across the 456, it is , about £255 a month from one month's guide downloads, from people who asked to hear from her.
Along the bottom are the two forms, each with two bars: addresses collected, and addresses that can lawfully receive marketing email. Up the side is the number of addresses. The takeaway: the bundled form looks like the bigger list and is in fact an empty one, 1,200 addresses to store and protect and never use.
In my experience a list built this way also gets fewer spam complaints, because nobody on it is surprised. If you want to compare two wordings of the box properly, that is an A/B test with enough people in each arm, and the article on sample size, lift and test design shows how many you need before the difference means anything.
Asking for consent by email is itself marketing
When a list turns out not to be clean, the instinct is to email everyone and ask them to confirm. Do not, if you were not allowed to email them in the first place.
In 2017 the ICO fined Flybe £70,000 for sending more than 3.3 million emails titled "Are your details correct?" to people who had opted out of marketing, asking them to update their preferences, with a prize draw for those who did. Honda Motor Europe was fined £13,000 for 289,790 emails asking customers to confirm their marketing choices. Honda argued the emails were customer service. The ICO disagreed: an email asking whether someone wants marketing is itself marketing.12 The ICO's current guidance makes the same point from the other side: a service message becomes direct marketing as soon as you add marketing material to it.9
So the clean route to permission runs through moments the person started: the next purchase, the next login, with a plain question at checkout or in the account.
Worked example: cleaning the old popup list
Amara's old popup collected 3,000 addresses over two years under the wording from the first worked example. Of those, 1,100 have since bought something.
The 1,900 who never bought: no valid consent, no soft opt in. She does not email them, not even to ask, and deletes the addresses on her normal retention schedule.
The 1,100 who bought: her checkout never offered an opt out, so the soft opt in was not set up, and a later opt out is too late.7 No "please confirm" campaign either. From now on her checkout and account page carry the opt out box and the new consent wording, and when any of the 1,100 buys again, their status is recorded properly from that moment. Slower than a mass email, and it does not end in a fine.
Proof is a record, not a feeling
Article 7(1) puts the burden on you to demonstrate consent.5 The ICO says the record should show who consented (a name or an identifier), when (a timestamp), what they were told (the consent wording, versioned and dated), how (the form and the submitted data), and whether and when they withdrew. Keep the record for as long as you rely on the consent.13 For the soft opt in, the ICO suggests a simple flag or preference field per person that shows which basis applies, plus a suppression list you check before every send.7
In the shops I build, this is one small table, written by the form and never edited by hand. A row looks like this:
{
"email": "ben@example.com",
"basis": "consent",
"channel": "email",
"purpose": "weekly_new_teas",
"wording_version": "popup_v3_2026_03_02",
"source": "guide_form",
"page": "https://kettleandleaf.example/guides/brewing",
"given_at": "2026-03-14T09:12:44Z",
"withdrawn_at": null
}
The wording version points to a stored copy of the exact text the person saw. Without it, the record says someone agreed, but not to what.
The ICO also suggests refreshing consent every two years if in doubt.13 I flag older consents with no engagement, and those people get a plain question next time they are in their account.
Worked example: the complaint that ends in a minute
Ben, the owner of the row above, writes to Amara: "I never signed up to your emails." She finds his row: consent, from the guide form, 14 March 2026 at 9.12am, wording version popup_v3_2026_03_02, which she opens: "Also send me the weekly Kettle and Leaf email about new teas. Unsubscribe any time." She replies with the date, the page and the wording, apologises, and unsubscribes him on the spot. The record is not for winning arguments with customers. It shows the ICO, if it ever asks, that the consent was real.
Leaving must be as easy as joining
Article 7(3) says withdrawing consent must be as easy as giving it.5 The ICO's guidance puts it the same way and adds that withdrawal should ideally use the same method and one step, and must not leave the person worse off.13 For the soft opt in, every message needs a free opt out that does not ask people to log in.7
Separately, Article 21 gives everyone the right to object to direct marketing at any time, and once they object the marketing must stop.14 The ICO calls this an absolute right: there are no grounds to refuse.8 It also covers people you emailed under the soft opt in, so an unsubscribe from a customer ends the soft opt in for good, until that customer actively consents later.
In practice that means three things. The unsubscribe link works in one click. It writes to the same consent table, so withdrawn_at gets filled in. And every mailing list is built by a query that excludes anyone on the suppression list, so an unsubscribe in one tool cannot be undone by an import into another.
Worked example: one unsubscribe, three tools
Hollie, at hollie@example.com, clicks unsubscribe on a Tuesday evening. The email tool marks her unsubscribed, and a webhook writes the time into her consent row and adds her to the suppression list. On Wednesday Amara uploads recent customers to her ad platform, and on Friday she imports last year's competition entries into the email tool. Both lists are built by queries that leave out the suppression list, so Hollie is in neither. Without that check, Friday's import would have quietly subscribed her again. Sending audiences to other platforms brings its own questions about the companies you share data with, covered in how data processors can get you fined.
Onboarding is the best moment to ask, and to ask well
Someone who has just bought, or just started a trial, is paying attention and wants something from you. That is the moment to ask what they want to hear about, rather than assume.
I build it as one short preference step. Two or three topics, each with a plain description and an honest frequency, each unticked. One channel per box, because the ICO wants consent asked separately for email and for text.7 And a settings page that changes it later in the same words. Getting the copy on that step right is worth an afternoon, because "about twice a month" is a promise people can hold you to.
Worked example: Rotawise's first login
When Priya first logs into her Rotawise trial, one screen asks what she would like to hear about, with two unticked boxes: "Tips for building rotas faster, about twice a month" and "New features, roughly monthly". A third line, "Text me when a shift swap needs my approval", is a service message, and is labelled as part of the product, not marketing.
Priya already falls under the soft opt in for product emails. Rotawise asks anyway, because her answer says which emails she will open. She ticks the tips only. From then on she gets the tips and nothing else, and her choice is written to the consent table with the wording version of the screen she saw.
The short version: build the form so the right answer is the easy one
Here is the whole decision, as I run it before any email address goes onto a marketing list.
Read it from the top. Each diamond is a question whose answer stops you or moves you on. The order matters: the suppression list comes first, because an objection beats every other basis. The takeaway is that there are only two doors into a marketing list, consent and the soft opt in, and a business that is not a charity has no third.
None of this asks for less ambition, only for fewer addresses you cannot use. A clear box, a recorded wording, an opt out at the checkout and a suppression list checked on every send take a few days to build, and then quietly do the right thing every day. Trust built that way tends to pay you back later, which is the subject of GDPR as a strategic advantage.