How to Personalise Cold Emails at Scale

The personalisation paradox

Truly personal emails don't scale. Fully automated emails don't convert. The goal of personalisation at scale is to operate as close to 'personal' as possible while still sending at volume. Every shortcut you take costs some reply rate. The art is knowing which shortcuts cost the least.

The good news: most of the reply-rate benefit of personalisation comes from the first 20 words. If you can personalise the opening line convincingly, the rest of the email can be templated without significant loss. This is the core insight that makes scale possible.

Dynamic variables: beyond {{first_name}}

Dynamic variables are placeholders in your email template that get replaced with prospect-specific data when the email is sent. Most people only use first name and company name. The real leverage is in custom variables that reference the trigger, the segment, or a specific pain point.

For example: a {{signal}} variable filled with 'recently expanded into Southeast Asia' or 'just announced Series B' or 'hiring for a growth role'. A {{pain_point}} variable filled with 'managing outreach manually' or 'high SDR ramp time'. These two variables on top of name/company turn a generic template into an email that reads as personalised.

• Minimum custom variable: {{signal}} — add one line about why you're reaching out now.
• More powerful: {{pain_point}}, {{industry_problem}}, {{recent_achievement}}.
• Keep variable fill text short — one phrase, not a sentence.

Email with custom variables: Hi {{first_name}}, {{signal}} — which usually means {{pain_point}} becomes a real bottleneck. [Rest of email template...]

Spintax: natural variation at zero extra effort

Spintax lets you define multiple variations of a word, phrase, or sentence inside a single template, with one variant chosen randomly per send. This serves two purposes: it makes emails feel less robotic (different prospects see slightly different phrasing), and it helps deliverability by reducing identical content fingerprints that spam filters recognise.

A simple example: {Hi|Hello|Hey} {{first_name}}. More powerful: {I noticed|I saw|Quick observation:} you're {growing your team|scaling your sales function|adding headcount} — {that usually means|which often means|which often comes with} [the problem you solve].

• Apply spintax to greetings, transition phrases, and CTAs — not to the core value statement.
• Keep variant lists to 2–4 options per spin. More than that becomes hard to QA.
• Read every combination mentally before sending — some combinations sound unnatural.

Segment-level personalisation: write once, sound personal everywhere

Instead of personalising every email individually, write one email per segment. A segment is a group of prospects who share a specific characteristic — same industry, same job title, same company size. Your message to a 50-person SaaS startup's sales leader is different from your message to a 500-person enterprise VP Sales, even if you're selling the same product.

With 5–6 well-written segment templates, you can cover 80% of your prospect universe without any per-lead research. The email reads as personalised because the problem, the language, and the examples are all specific to that type of person — even if they weren't researched individually.

• Build segments around: industry (SaaS vs e-commerce vs agency), seniority (VP vs manager), and company size.
• Test one segment at a time — don't change copy and audience simultaneously.
• A good segment template should be unrecognisable as a template to anyone in that segment.

Using AI for personalisation research

AI tools can dramatically speed up first-line research. Feed a tool the prospect's LinkedIn URL, their recent posts, and their company's about page, and ask it to write three personalised opening lines. You still need to review and edit — AI often misses nuance or produces generic outputs — but it can cut per-lead research time from 10 minutes to 2.

The human role in AI-assisted personalisation is curation and judgment. You know which opening line will resonate; AI gives you three candidates to choose from. As models improve, the curation step will get easier. The judgment step — understanding what actually matters to this type of buyer — will always be yours.

• Always review AI-generated lines before sending. They can be subtly wrong in ways that damage trust.
• AI is best at summarising public information. It can't tell you what the prospect actually cares about.
• Use AI for volume (Tier 3 and Tier 4) accounts — handcraft Tier 1 yourself.