Signal based personalization in cold email means writing each message from a live fact about the prospect (their role, company stage, a hiring push, a new plant) instead of dropping a name into a template. The signal explains why you are writing and why now, and the rest of the email connects it to one problem you solve. Inside Artemis, the system behind the AGP growth machine, messages are written from those signals, tested in variants, and tuned toward positive replies.
Nerdy Joe is owned by Agency Growth Partners (AGP), so this is our own system and we are not neutral about it. Below is how the writing stage works, before and after examples you can borrow, and criteria you can use to judge any provider or your own process.
The step before this one is placement. Personalization only pays off once mail reliably lands in the inbox, covered in mailbox warmup and inbox placement.
Why templates stop working at agency volume
A template is written once for an imagined average prospect. At low volume, a good template can carry a campaign. At agency volume, across many industries and roles, the average prospect does not exist, and the email reads like it.
Templates also fail on timing. "I help companies like yours grow" was equally true last year and will be equally true next year. Nothing in the email says why the prospect should care this week.
AGP's line on this is short: "Not templates." Artemis writes from live signals, so each message has a specific reason to exist.
What a signal is, and which ones Artemis uses
A signal is a fact about the prospect or their company that suggests a need. Artemis writes from live signals such as these:
| Signal | What it tells you | Example angle for an agency |
|---|---|---|
| Role | What the person owns and is measured on | A new head of marketing who owns pipeline targets |
| Company stage | Growth, funding or maturity | A company scaling past founder-led sales |
| Hiring activity | Where the company is investing | Several open sales roles, so pipeline is a priority |
| New plant opening | Physical expansion into a new market | A manufacturer that needs buyers near a new site |
| Hiring push | A sudden, concentrated ramp | A burst of hiring that signals a deadline or launch |
These signals come from the data stage. Every contact arrives with company and intent context attached, which is why the list work in defining your ideal client profile for outbound matters so much. A vague profile produces vague signals.
What makes a signal usable
Not every fact is worth writing about. A signal is usable when it passes three checks:
- It is true and current. Old news reads as lazy research.
- It connects to your offer. A new plant matters to an agency that generates regional buyer pipeline. It does not matter to one that redesigns logos.
- It changes the ask. The signal should shape what you propose, not just the first line.
How the writing stage runs inside Artemis
Here is the sequence, step by step, as it runs on a client engine.
- Signals attach to each contact. Role, company context and intent context arrive with the record from the data stage.
- AI drafts from the signals. Artemis drafts messages built around the signal, not around a fixed paragraph.
- Drafts get rewritten and checked. AI rewrites can shorten a message, make it friendlier or fix grammar, and Artemis scores the email content.
- Variants are set per step. Each step in the sequence has A/B variants, so the engine learns which version earns better responses.
- Optional human review. An Autopilot mode can hold emails for human review before they send.
- The campaign self-optimizes. Over time, campaigns shift toward the messages that earn positive replies.
- Messages adapt by channel. The same signals shape LinkedIn messages, adapted for that format.
A dedicated growth engineer owns this stage for each client. They set the offer and angles, read the results, and change what is being tested.
Before and after: three illustrative examples
The examples below are written for this post. They are illustrative only, use invented generic companies, and are not real campaigns or results. The sender is a fictional marketing agency.
Example 1: Template vs. role and stage signal
Use this pattern when the prospect is new in a role that owns growth.
Before (template):
Subject: Quick question
Hi [First Name],
I hope this finds you well. We're a full-service marketing agency that helps companies like [Company] grow their revenue with proven strategies.
Would you be open to a 30-minute call next week to learn more?
Best,
[Your Name]
After (signal based):
Subject: Your first 90 days at Example Software Co.
Hi [First Name],
Congrats on taking over marketing at Example Software Co. New marketing leads at companies past their first funding round usually get asked for pipeline numbers fast, often before the team is fully built.
We run outbound for B2B software companies at that stage, so pipeline starts moving while you hire.
Worth a short call to see if it fits your plan for the quarter?
[Your Name]
What changed: the email names the role change and the stage, connects both to one pressure (pipeline numbers), and makes a narrow offer.
Example 2: Template vs. new plant signal
Use this pattern for manufacturers or distributors expanding to a new site.
Before (template):
Subject: Partnership opportunity
Hi [First Name],
We work with manufacturers to improve their marketing and generate more leads. Our team has years of experience in your industry.
Let me know if you'd like to chat.
Thanks,
[Your Name]
After (signal based):
Subject: Buyers near the new Example Industrial plant
Hi [First Name],
Saw that Example Industrial Co. is opening a new plant. A new site usually means new capacity before there is a full order book in that region.
We build outbound to plant managers and buyers within a set radius, so the sales team has conversations lined up as the plant comes online.
Open to a quick look at who we would target?
[Your Name]
What changed: the signal sets the timing (capacity before orders) and the offer (buyers near the new site), and the ask is low effort.
Example 3: Template vs. hiring push signal
Use this pattern when a company is suddenly hiring in a function your offer supports.
Before (template):
Subject: Grow your sales
Hi [First Name],
Are you looking to grow your sales? We help businesses like yours get more customers.
Can we set up a call?
[Your Name]
After (signal based):
Subject: The sales roles Example Logistics Co. is hiring for
Hi [First Name],
Noticed Example Logistics Co. has several sales roles open at once. When a team grows that fast, new reps often wait weeks for a full calendar.
We fill calendars for new reps with qualified first meetings, so ramp time goes into selling instead of prospecting.
Would it help to see how that would work for your new hires?
[Your Name]
What changed: a hiring push points to a specific near-term pain (new reps without meetings), and the offer addresses only that.
A/B variants per step, and what to test
Every step in an Artemis sequence can carry A/B variants. The goal is not to test everything. It is to learn one useful thing at a time.
| Test this | Example variant A | Example variant B |
|---|---|---|
| Which signal leads | Role change | Hiring push |
| The call to action | Short call | "Want to see the target list?" |
| Length | Five lines | Three lines |
| Subject line | Names the signal | Names the outcome |
| Follow-up angle | New proof point | Simpler ask |
Change one variable per test, give each variant real volume, and judge it on positive replies. A clever variant that gets more replies but fewer interested ones is a worse email.
Self-optimization toward positive replies
Artemis campaigns self-optimize toward the messages that earn positive replies. Over time, more volume goes to the versions that start real conversations.
Optimizing for positive replies, not opens or total replies, matters for two reasons. Open tracking is noisy, and a message that provokes "remove me" replies is not winning anything. Positive replies are the closest early signal to meetings.
For context on what that looks like at scale, AGP reports a live campaign on the Artemis dashboard with a 0.98% reply rate and 8.39% of replies interested over the last 30 days. That is one campaign AGP reports, not a benchmark or a promise. For how reply and interest rates turn into meetings, the pillar on how AGP's growth machine works walks through the projection math.
AI drafting, rewriting and content scoring
AI does three jobs in this stage inside Artemis:
- Drafting. It writes first versions from the signals attached to each contact.
- Rewriting. It can shorten a message, make it friendlier or fix grammar.
- Scoring. It scores email content before it goes out.
What AI does not do here is decide the offer. The offer, the target and the angle come from the strategy work in the Advise stage. AGP's services page says it directly: most agency pipelines fail at the offer, not the list. No amount of rewriting fixes an offer nobody wants.
Autopilot and human review
An Autopilot mode in Artemis can hold emails for human review before they send. That is useful in three situations:
- A new campaign. Review the first batches until you trust the drafts.
- A sensitive market. Regulated industries or high-value accounts deserve a human read.
- A new signal type. When you add a signal the engine has not written from before, check how it is being used.
Once a campaign is proven, review can be lighter. The choice stays with the people running the engine.
How to judge any personalization approach
Use this checklist on a provider, a tool or your own team:
- [ ] Each email is built around a specific, current signal, not a merge field
- [ ] The signal connects to the offer and shapes the ask
- [ ] Signals come attached to the contact record, not researched by hand one at a time
- [ ] Each sequence step has variants, and tests change one thing at a time
- [ ] Success is measured on positive replies
- [ ] AI drafts can be rewritten, checked and scored before sending
- [ ] Emails can be held for human review when needed
- [ ] The same signals shape LinkedIn messages, not just email
If you are comparing outside teams, our guide to what a cold email agency does covers the rest of the service to check.
Who this is not a fit for
If you sell one narrow service to a tiny, well-known list of accounts, hand-written emails from the founder may beat any system. Personalization at scale solves a volume problem you may not have.
It is also not a fit if you cannot say which change at a prospect's company makes your offer urgent. Signals only work when they map to a real need. Work that out first.
What comes next in the machine
Once messages are written from signals, the next stage keeps email and LinkedIn in sync so the same signals reach prospects on both channels. That is covered in email and LinkedIn outreach for agencies. After that, replies get sorted and worked, covered in AI reply handling and follow-up.
Want this built for your agency?
If you want signal based messaging and the rest of the full AGP machine built and tuned by a dedicated growth engineer, book a call with David. You will get a plain answer on whether your offer is ready for it.
Sources
Frequently asked questions
What is signal based personalization in cold email?
It means writing each email from a live, verifiable fact about the prospect, such as their role, company stage, a hiring push or a new location, instead of filling a name and company into a fixed template. The signal is the reason the email exists and the reason it is being sent now.
What counts as a buying signal for B2B outreach?
Useful signals are changes that create a need: a new role, a funding round or growth stage, active hiring in a function you support, a new plant or office, or a new market. A signal is only useful if it connects to a problem your offer solves.
Is first-line personalization the same as signal based personalization?
Not quite. A first line that mentions a podcast or a post can be personal without being relevant. Signal based personalization ties the whole email, including the offer and the ask, to a change at the prospect's company.
Can AI write personalized cold emails well?
AI is good at drafting from structured signals, rewriting for length and tone, and checking grammar. It still needs a clear offer, accurate signals and a human who can review messages, which is why Artemis has an Autopilot mode that can hold emails for review before they send.
How many A/B variants should each cold email step have?
Enough to learn something, and few enough that each variant gets real volume. Most teams test a small number of variants per step and change one thing at a time, such as the opening signal or the call to action.
How do you measure whether personalization is working?
Measure positive replies, not opens or total replies. A message that earns more interested replies from the same audience is better, even if it gets fewer replies overall.