# Personalization that works

The personalization that gets replies, in order: buying signals, offer-relevant company and role details, then personal details, with AI field instructions.

Personalization earns a reply when it points at a problem the prospect has and you can solve. Not all personalization is equal. In order of what works best: a buying signal, then a company or role detail tied to your offer, then a personal detail from their profile, which you can usually skip.

## The personalization hierarchy for cold outreach

1. **A buying signal**: something happening at the company right now that creates the problem you solve. A message that zeroes in on solving that specific, real problem is the best message you can send.
2. **A company or role detail tied to your offer**: who they serve, what they sell, the market they're in, or a priority common to their role, connected to the value you give.
3. **A personal detail from their profile**: their school, a hobby, a post unrelated to your offer. This often does more harm than good. It's fine to leave it out.

Use the highest tier you have true data for. When you have nothing true at a tier, drop to the one below. Never invent a detail to fill the gap.

## Personalization tier 1: a buying signal tied to a real problem

A buying signal is a recent, visible event that makes your offer relevant now. Common ones:

- Hiring for a role your offer supports, such as a first SDR or a third support hire.
- A launch, a new market or a new location.
- A change in how the company sells, such as adding a sales team or a free trial.
- The person posting about the problem you solve.

Name the signal, then name the problem it usually creates, then offer to help with that problem. Here `{signal_line}` is a column from your lead file, such as "is hiring its first SDRs":

```text
Hey {first_name}, saw that {company} {signal_line}.

The first month is usually spent building lists by hand. Mind if I send over [the give: the list we'd start with for a team like yours]?
```

The signal must be true and recent. A stale or wrong signal reads worse than no personalization at all.

## Find buying signals with Victoria Pulse

[Victoria Pulse](https://www.versionseven.ai/pulse) is VersionSeven's separate signal product. It tracks the companies in your market and flags buying signals, such as a jump in hiring, new positioning or a change in how a company sells, with a source for every fact and the dates each signal is active.

To use a signal in a Victoria AI campaign:

1. Add it to your lead file as its own column, written to fit a sentence, such as `signal_line` with "is hiring its first SDRs".
2. When you upload the file, keep the column as a custom field. See [Import a CSV](https://docs.versionseven.ai/help/import-a-csv).
3. Use it in a message as `{signal_line}`. The readiness check flags leads that have no value for it.

Reach out while the signal is active. A hiring push from six months ago is no longer a reason to write.

Without Pulse, an AI Personalization field can look for signals on the company website and in the person's recent LinkedIn posts. See "AI Personalization instructions for a buying signal" below.

## Personalization tier 2: company or role details tied to your offer

When there's no signal, use a detail about the company or the role that leads straight to the value you give: the customers they serve, a service they sell, the market they're in, or a priority common to people in their role.

```text
Weak: Hey Sam, love what Northside is doing in the HVAC space.
Better: Hey Sam, saw that Northside does commercial installs as well as residential. I help HVAC companies fill the slow months with commercial maintenance contracts.
```

The test: does the detail explain why you're writing to this person? If it could be deleted without changing the message, it's decoration, not personalization.

## Personalization tier 3: personal profile details, usually skip them

A random personal detail, such as where they studied, a hobby, a sports team or a post about something unrelated to your offer, is the weakest kind of personalization. In many cases it does more harm than good:

- It reads like a template trick, because everyone has seen "Saw you went to \[school]!".
- It can feel like surveillance.
- It says nothing about why you're writing, so the prospect still has to work that out.

It's fine to leave personal details out entirely. The one exception is a personal post about the problem you solve, which is really a tier 1 or tier 2 detail.

## AI Personalization instructions: say what, how long, how it's used and what to avoid

An AI Personalization field is only as good as its instructions. Every instruction should say:

- **What to find**: one concrete thing, from the tier you're aiming for.
- **How long**: "one sentence", "under 15 words".
- **How it's used**: "written to follow 'Saw that' in a message", so it fits the copy around it.
- **What to avoid**: "no generic praise such as 'innovative company'", "don't repeat the person's name", "only state what the sources show".

Replace the bracketed parts of the examples below with your own offer before you save them.

## AI Personalization instructions for a buying signal

A tier 1 field, such as `buying_signal`, with **Data Sources** set to **Website** and **LinkedIn Profile**:

```text
Look for one specific, recent event at the company that suggests they need [what you solve]: hiring for [roles], a launch, a new market, new funding, or a post by the person about [the problem]. Write one sentence under 20 words that names the event and the problem it usually creates for [their role]. Only state what the sources show. No generic praise. Don't use the person's name.
```

Set the **Fallback value** to a tier 2 sentence that's true for every lead in the list, such as "I work with a lot of \[their role]s at \[kind of company] on \[the problem]." Leads with no signal then get a relevant line instead of a vague one.

## AI Personalization instructions for a company or role detail

A tier 2 field from the company website, such as `offer_hook`, with **Data Sources** set to **Website**:

```text
From the company website, find one detail that connects to [your offer]: the customers they serve, a service they sell, or the market they're in. Write one sentence under 20 words that links that detail to [the outcome you deliver]. No generic praise. Don't use the person's name.
```

A tier 2 field from the person's role, with **Data Sources** set to **LinkedIn Profile**:

```text
From the person's title, headline and recent posts, write one sentence naming a likely priority for someone in their role that relates to [the problem you solve]. Phrase it as an observation, not an assumption about their problems.
```

We don't recommend an AI field for tier 3. If you write one anyway, tell it to use a personal detail only when it relates to \[the problem you solve], and to write nothing personal otherwise.

## Always set a personalization fallback

Set a **Fallback value** on every field: a phrase or sentence that reads well in the same spot and is true for every lead, ideally from the tier below. Without one, a lead the research can't support gets a generic line or is held and retried later. See [Personalization fields](https://docs.versionseven.ai/help/personalization-fields).

## Use one or two personalization fields per sequence

Put one or two AI fields in the highest-impact touch: the first LinkedIn message after connecting, or the first email. Never two in one message, and never in a connection note. Each field costs credits per lead and is one more chance of a held lead.

## Test personalization before activating

On **Sequence**, open the step, select **Preview**, pick a few real leads and select **Generate**. Read the results as the prospect would. If several come back **Generic** or held, tighten the instruction or improve the fallback. If a signal field comes back with something that isn't really a signal, narrow what it looks for. Generating a preview spends credits, so test on a handful of leads.