Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Email Marketing › Email Campaigns

Personalising Email Content

How to personalise email so it helps the reader: merge fields with fallbacks, blocks by segment and stage, recommendations, the data needed and testing it.
Guideintermediate10 min readUpdated

Personalisation changes the message, not the audience

The segmentation guide draws the line in one sentence: segmentation decides who receives a message, and personalisation changes the message for the person who receives it. The order matters. A personalised email sent to the wrong audience is still the wrong email, so the audience decision comes first and personalisation works inside it.

Within a send, personalisation takes three forms, which this guide takes in turn. Merge fields insert a value from the person's record, such as a first name, a company, a last order date or a loyalty balance, into a fixed template. Content blocks swap whole sections of the email according to a rule: a different product block for customers and non-customers, a different tip for each of several stated interests. Recommendations choose products or articles for each person from what they or similar customers have bought or viewed.

Each form needs data, and each can go wrong in front of the reader. The question to ask of every personalised element is the one the segmentation guide asks of every data field: what would be different for this reader because of it, and is that difference useful to them? A field that changes nothing the reader values adds risk without adding anything else.

Model-composed personalisation, where a model writes or selects content for each recipient at send time, raises further questions of control and review. The AI for marketing automation guide covers it under personalisation at send time; this guide is about rule-based personalisation that a person on the team can read and check.

Merge fields and their fallbacks

A merge field is the simplest personalisation and the one most often seen failing. Every merge field has three ways to go wrong, and each needs a decision before the send.

FailureWhat the reader seesThe fix
The field is emptyA greeting with a gap where the name should be, or a sentence that stops making senseA fallback value written for the sentence, such as a neutral greeting, set on every field
The field holds bad dataA name in capitals, a placeholder someone typed at sign-up, a company called "test"Clean or exclude obviously wrong values before they are used; fall back when in doubt
The field is staleA loyalty balance from last month, a plan the customer has left, a city they moved fromUse only fields refreshed before each send, or leave out values that change faster than the data does
The field label showsThe raw merge tag in the text, usually in the plain-text part or the preheaderTest sends with real and empty records; read the plain-text version

The email campaign QA checklist has a line for this: check personalisation fields have fallbacks for empty records. The fallback should be written for the sentence, not chosen once for the whole account. A greeting falls back to a neutral greeting; a sentence that names the reader's last purchase falls back to a sentence that makes sense without it, or the sentence is left out. Test sends should include at least one record with every field empty and one with the longest plausible values, because a long company name breaks a layout as surely as an empty one breaks a sentence.

Two habits keep merge fields honest. Use a field only where it reads naturally; a first name repeated in the subject line, the opening and the button reads as automation, not attention. And never use a field the reader did not give or would not expect the business to hold in an email they might forward, show a colleague or read on a shared screen.

Content blocks by segment and stage

Content blocks do more than merge fields because they change what the email says rather than how it addresses the reader. The usual rules for choosing a block come from the attributes the segmentation guide recommends keeping: lifecycle stage, engagement tier, stated preference and a few behaviour flags.

  • By lifecycle stage. A subscriber who has never bought sees a block that gives reasons and proof; a first-time customer sees reassurance and a path to a second order; a repeat customer sees recognition, early access or care advice. The same newsletter carries all three, and each reader sees one.
  • By stated preference. Someone who chose a topic at sign-up or in the preference centre sees that topic's block first. The segmentation guide's warning applies: a preference that changes nothing teaches people that choices are decorative.
  • By sign-up source. The welcome series tutorial personalises by source in one step: someone who downloaded a guide, someone who opted in at checkout and someone who signed up for an offer each wanted something different, and the series answers each, with a default branch for sources that have no special handling.

Build blocks as variations inside one email rather than as separate emails, so that the shared parts stay in one place and the fixed structure the newsletter design and copy guide recommends survives every variation, and always write a default block for readers who match no rule. Keep the number of variations small enough that someone can read every version before the send; a block that nobody on the team has seen in its final form is a block that will one day go out wrong.

Recommendations from purchases and behaviour

Recommendations choose products or content for each reader. The most dependable source is what the person has bought. A purchase says what they own, what they might need next and what they will not need again soon, which is the ground of cross-selling: the cross-selling entry describes offering a customer a related product alongside the one they are buying. In email that becomes the accessory, the refill or the care guide for the last thing bought.

Recency changes what a recommendation should be. RFM segmentation scores customers on how recently, how often and how much they have bought, and the recency score alone is a useful guide: a customer who bought last week wants help using it, not the same product again; a customer whose last order is long past their usual interval may want a reason to return. The QA checklist's line on suppressing customers who already bought what the email promotes is the simplest recommendation rule of all.

Behaviour on the website or in the product, such as pages viewed or features used, is a second source, and the one that needs most care. The segmentation guide notes that site and product behaviour can be used only where consent allows it, and that consent and privacy rules vary, so tracking should not be assumed to be permitted; the privacy and consent guide explains how consent decides what can be observed at all. Even where it is allowed, behaviour used too quickly or too precisely feels wrong. The abandoned cart tutorial makes the point for browse emails: an instant message about a page someone glanced at feels like surveillance. In a newsletter, a block that reflects a category the reader keeps returning to is helpful; a block that names the exact product they looked at last night is unsettling.

The data it needs and the line it should not cross

Personalisation runs on first-party data: what the business holds because of a direct relationship, such as a subscription, an account or an order. The segmentation guide's rule for collecting it applies here too: every field should map to a message that would be different because of it, and a field that changes nothing is a maintenance cost and, for personal data, a liability.

The law points the same way. The guide to minimising data before prompting describes the GDPR's data minimisation principle: personal data should be adequate, relevant and limited to what is necessary for the purpose, and processed for specified purposes rather than further processed in ways incompatible with them. For personalisation that means using the fields a message depends on, not every field the platform happens to hold, and not repurposing data collected for one reason for a kind of personalisation the person would not expect.

Personalisation also works inside the permission a person gave, not around it. The list building and consent guide summarises how the GDPR describes consent and how the ICO's direct marketing guidance, listed in sources, explains when consent is needed for electronic marketing in the UK. A personalised email is still a marketing email, sent on the same basis and with the same opt-out as any other. Rules differ by market and nothing here is legal advice; a new kind of data in personalisation is a question for whoever handles data protection before it is a question for the template.

Beyond the rules, a working line as general practice: personalise on what the reader told you or did with you openly, such as what they bought, what they chose and where they signed up, and leave out anything inferred or sensitive that they did not tell you. The test is whether the reader, seeing the email, would understand why it says what it says. If the answer would surprise them, the personalisation has crossed the line, whatever the data allows.

Testing whether personalisation helps

Personalisation costs effort to build and carries risk every time it sends, so it should earn its place with evidence. The fair test is a personalised version against a generic one, sent at the same time to two groups split at random, judged on clicks on the main call to action and, where volume allows, on conversions. The email metrics and testing guide covers how to run an A/B test on email with enough volume to trust, and why opens should not decide it.

Three points make the comparison honest. Test the whole personalised version against the whole generic version first, rather than one block at a time, so the question is whether personalisation helps at all. Read the result by segment as well as in total, because a block that helps customers may do nothing for prospects. And for personalisation that runs continuously, such as a recommendation block in every newsletter or a personalised step in a flow, keep a small holdout group that always receives the generic version, so the comparison stays visible as the data and the audience change.

A personalised version that does no better than the generic one is still worth knowing about. It is simpler to run the generic version, and the effort can go into content that changes the result. The campaigns hub sets out the measures each campaign type is read on, and a personalised send is read on the same ones.

A garden tool shop's personalised newsletter (illustrative)

Illustrative example — not a real result

Tinespur Garden Tools tests two personalised blocks

Tinespur Garden Tools, an invented online shop selling hand tools for gardeners, sends a monthly newsletter to 12,400 subscribers. It adds two personalised blocks: a care tip for the last tool each customer bought, with a fallback tip on winter storage for anyone with no purchase, and a seasonal product block chosen by customer stage, showing starter kits to people who have never bought, accessories to one-time customers and early access to a new range for repeat customers. The team splits the list at random for the spring issue, sends the personalised version to one half and the usual generic version to the other, and judges the test on clicks and orders over the following week.

VersionRecipientsClicks on main linkOrders
Personalised6,20041074
Generic6,20033061

Read by stage, the difference sits almost entirely with one-time and repeat customers; never-bought subscribers clicked at much the same rate in both halves. The shop keeps both blocks, rewrites the never-bought block, and holds out 500 subscribers on the generic version for the next three issues to see whether the difference lasts.

Common mistakes

  • Personalising before segmenting. The right name on the wrong offer is still the wrong email.
  • Merge fields without fallbacks. An empty record produces a gap, an odd sentence or a raw tag in front of the reader.
  • Using every field the platform holds. More fields mean more stale data and more personal data in use, with nothing gained for the reader.
  • Recommending what the reader just bought. The simplest suppression rule is the one most often missed.
  • Acting on behaviour too quickly or too precisely. A block about the exact page viewed last night reads as being watched.
  • Using behaviour without checking consent. Whether tracking is allowed is a question to answer before the block is built.
  • Never testing against a generic version. Personalisation is assumed to help, and its cost and risk are never weighed against what it adds.
  • Too many variations to read. A block nobody has seen in its final form is the one that goes out wrong.

Frequently asked questions

Does putting the subscriber's first name in the subject line help?

Sometimes, and there is no general figure that says how much or for whom. A name is the cheapest personalisation and the one readers have seen most often, so it rarely says anything the reader did not already know. Content that changes because of what the person bought, chose or said they wanted does more work. If the name is used, give it a fallback for empty and badly typed records, and test it against a version without it on clicks rather than opens.

What is the difference between personalisation and segmentation?

Segmentation decides who receives a message; personalisation changes the message for each person who receives it. The segmentation guide puts it in those terms and adds that segmentation comes first, because a personalised message sent to the wrong audience is still the wrong message. In practice the two share data: the lifecycle stage that defines a segment can also choose which block a person sees inside a broader send.

Is it legal to personalise emails with purchase or browsing data?

That depends on the market, on how the data was collected and on what people were told, and nothing here is legal advice. The site describes the GDPR's data minimisation principle, that personal data should be adequate, relevant and limited to what the purpose needs, and the consent rules the list building guide summarises from the GDPR and the ICO's direct marketing guidance. Use only data a message depends on, use behaviour only where consent allows it, and ask whoever handles data protection before adding a new kind of data.

Terms used on this page

Frameworks

Also related

Sources

  1. Regulation (EU) 2016/679 (GDPR) — EUR-Lex
  2. Direct marketing guidance — Information Commissioner's Office (UK)

Related