Which Data Points Are Worth Personalizing On

Personalisation works when the data point you use implies the prospect has a problem you can solve. Most personalisation fails because the data is merely true. Their headcount, their funding round, their office location are all accurate and none of them tell the reader why you are writing. The question is not how much you know about them, it is which single fact makes the next sentence inevitable.

The Test a Data Point Has to Pass

One question separates useful personalisation from decoration:

Does this fact, on its own, imply a problem?

"You raised a Series B four months ago" is true and implies nothing. "You have posted three SDR roles in six weeks and none are filled" implies a hiring bottleneck, a pipeline target someone is missing, and a manager under pressure. Same effort to source, completely different value, because the second one earns the sentence that follows it.

Apply the test ruthlessly and most of what enrichment tools return gets discarded. Industry, headcount, revenue band, tech stack, and location are segmentation data. They are excellent for deciding who to contact and nearly useless as the opening line of a message, because they describe a state rather than a tension.

That distinction is the whole article. Segmentation data tells you who belongs on the list. Personalisation data tells you what to say. Teams routinely use the first for the second, and the result is a message that proves you ran a query.

The Four Tiers

The inverse relationship is the whole difficulty. Tier 4 is free and worthless for copy. Tier 1 is valuable and requires actual monitoring infrastructure. Most teams personalise at tier 4 because it is what their enrichment returns, and then conclude personalisation does not work.

The Best Data Points, Specifically

Ranked by how reliably they earn a reply:

• A role they are trying and failing to fill. A job posting that has been live for two months is a problem someone owns right now, with a budget already approved. Hard to beat.

• A new leader in the relevant function. Someone three months into a role is actively looking for things to change and has unusual licence to buy. The window is real and it closes.

• A tool they just adopted, or just removed. Both imply a project in motion and a team mid-decision.

• A public commitment with a date. A stated target, a launch, an expansion. Anything with a deadline creates pressure you can speak to.

• Their own words about the problem. A post, a talk, a job description that describes the difficulty in their language. Quoting someone accurately is the highest-credibility opening available, and the hardest to fake.

What consistently underperforms despite being popular: funding rounds (over-used to the point of being a tell), company anniversaries, generic congratulations, anything about the weather or their city, and any observation drawn from their personal social media. The last category actively backfires, which is the subject of relevance beats personalisation.

Sourcing Without It Becoming a Research Job

The honest constraint: tier 1 data does not come out of a standard enrichment field. It comes from watching for changes.

Three approaches, in increasing order of effort:

• Monitor rather than enrich. Enrichment answers what is true about a company now. Monitoring answers what changed. Job postings, leadership changes, and tooling changes are all observable on a schedule, and the change is the signal.

• Derive the trigger from data you already have. Two snapshots of headcount by function, taken a quarter apart, produce a growth signal without any new source. Most teams hold the raw material and never diff it.

• Reserve manual research for the accounts that justify it. Reading someone's actual words takes minutes and cannot be automated well. Spend it on the accounts you most want, not evenly across the list.

The practical architecture that follows: segment broadly using firmographics, personalise narrowly using triggers, and accept that a meaningful share of your list will have no trigger available. For those, a well-written role-plus-context message is the correct fallback, and pretending to a trigger you do not have is worse than not having one.

One Fact Per Email

A recurring failure is stacking personalisation. Three researched details in one email do not read as three times more attentive; they read as a dossier, and they crowd out the point.

Use one fact, and let it do one job: justify why you are writing to this person now. Everything after it should be about the problem, not about them. Two details is occasionally defensible when the second is genuinely the consequence of the first. Three is always a symptom of someone proving they did research rather than using it.

This connects directly to length. In a 50 to 125 word email there is room for one relevance line, and that is the constraint doing useful work: how long a cold email should be and where the ask goes.

Quality of the Underlying Data

A personalisation strategy sits on top of a list, and a stale field produces a confident error.

The dangerous ones are the fields that look fine when wrong. Job titles change without the record changing. A funding round described as recent may be two years old. A "current" tool may have been replaced. Referencing any of these incorrectly is worse than sending a generic email, because it demonstrates inattention specifically.

So the rule is simple: the more specific the claim, the more recently it must have been verified. A firmographic can be a few months old without much risk. A trigger event older than a quarter is not a trigger, it is history, and using it dates you precisely. The maintenance side of this is covered in why list hygiene shows up in the numbers.

Where the Human Belongs

Sourcing and drafting scale. Judging whether a data point actually implies a problem does not, because it requires knowing the domain.

A model given a funding round will write a confident sentence about growth challenges. It has no way to assess whether this company has that challenge, and it will not tell you it is guessing. The judgement about whether the inference holds is the part that has to stay human, and it is cheap to do at the template and sample level rather than per email. The specific checks are in what to check before an AI-written email goes out.

FAQ

What is the best personalisation data point for cold email? A role they have been trying to fill for months. It is a live, owned problem with budget already attached, and it implies the difficulty rather than merely describing the company.

Is mentioning a funding round good personalisation? It is popular and weak. Everyone uses it, so it signals a list rather than attention, and knowing someone raised money does not tell them why you are writing.

How many personalised details should one email contain? One. Two only when the second follows from the first. Three reads as a dossier and crowds out your actual point.

Can personalisation be fully automated? Sourcing and drafting can. Deciding whether a fact genuinely implies a problem cannot, because that requires domain knowledge a model does not have and will not admit to lacking.

What should I do for prospects with no trigger event available? Use role plus context and write well. A clear, relevant, generic-but-honest email outperforms a fabricated specific every time.

How fresh does personalisation data need to be? In proportion to how specific the claim is. Firmographics tolerate a few months. A trigger event older than a quarter is no longer a trigger and referencing it as current works against you.

Want personalisation built on signals that imply a problem rather than fields that fill a template? That is what Lidgen's list building is for. Book a demo.