All notes Field notes

Why isn't automated personalization converting?

In short
  • Personalization helps a cold email get noticed. In a field experiment, adding the recipient's name to the subject line raised opens by about 20% and leads by about 31%.
  • Getting noticed is not a reason to reply. Gartner found that 73% of B2B buyers actively avoid suppliers who send irrelevant outreach.
  • What is worth personalizing is the reason for contact: what is changing in how the company sells, and what problem that creates. Written down, that reason becomes a hypothesis the market can confirm or reject.
  • AI can do the research and the first draft. Choosing the reason still needs someone who understands how the client sells.

In October we started an outbound test in the US with software companies that sell through partners. The first email to each of them doesn't open with the person's latest post, their alma mater or a compliment on their podcast. It opens with a change in how the company sells. One of them, anonymized and shortened, starts like this: "[Company] recently opened a channel account manager role to recruit resellers and MSSPs in one US region." The rest of the email follows from that sentence. This note explains why we write first emails this way and what the research says about personalization.

What does personalization actually change?

Personalization does have a measurable effect. Sahni, Wheeler and Chintagunta (Marketing Science, 2018) ran field experiments in which companies sent marketing emails with and without the recipient's name in the subject line. Adding the name raised the open rate from 9.05% to 10.80%, about 20%, and the share of emails that produced a lead from 0.39% to 0.51%, about 31%.

Effect of the recipient's name in the subject line In field experiments, adding the recipient's name to the subject line raised the open rate from 9.05% to 10.80%, about 20%, and the share of emails that produced a lead from 0.39% to 0.51%, about 31%. Source: Sahni, Wheeler and Chintagunta, Marketing Science, 2018. Marketing emails to consumers, not cold B2B outbound. Open rate About 20% higher 9.05% 10.80% Without name With name Emails that produced a lead About 31% higher 0.39% 0.51% Without name With name Each panel has its own scale, starting from zero.
Open rate and share of emails that produced a lead, without and with the recipient's name in the subject line. Source: Sahni, Wheeler and Chintagunta, Marketing Science, 2018. Marketing emails from companies to consumers, not cold B2B outbound.

Two details matter here. First, the name carries no information about the offer. The authors' explanation is that a message that looks addressed to you is more likely to get your attention. Second, these were marketing emails from companies to consumers, not cold outbound to B2B decision-makers. The result is a fair argument for personalization, but it is an argument about attention.

Why doesn't a personalized first line give anyone a reason to talk?

A decision-maker who opens your email still has to decide whether replying is worth their time. That decision depends on whether the email is about something that matters to their business now. Gartner's 2025 survey of 632 B2B buyers points in the same direction: 73% said they actively avoid suppliers who send irrelevant outreach. Inboxes are starting to filter as well: some AI email clients now label cold pitches and move them out of the main inbox (Superhuman, 2025).

73%of B2B buyers actively avoid suppliers who send irrelevant outreach. Source: Gartner, survey of 632 B2B buyers, 2024, published 2025.

A line about someone's latest post is relevant to the person, but not necessarily to the reason you are writing. When every outbound tool can produce that line, it also stops telling the reader much about the sender. After ApolloNEXT this year I wrote:

After ApolloNEXT

Your competitors can use the same platforms, follow the same playbooks and generate messages based on the same public information. I don't think adding another layer of "personalization" will be enough to create a lasting advantage.

What should you personalize instead?

I would personalize the reason for contact, and take it from how the company sells: who it sells to, through whom, what is changing in that model, and what problem the change is likely to create. That gives an order to the first message: a change in the company's sales model, the problem that follows from it, and a specific way you could help.

Three levels of personalization Level 1, the person: name, latest post, alma mater; for the reader it gets attention. Level 2, a signal: a new role, a partner program, a new market; it shows why now. Level 3, a reason from how the company sells: who it sells to, through whom, what is changing and the problem that follows; it gives a reason to reply. Conceptual diagram, no data. 1 The person Name, latest post, alma mater For the reader: gets attention 2 A signal A new role, a partner program, a new market For the reader: shows why now 3 A reason from how the company sells Who it sells to, through whom, what is changing For the reader: a problem worth replying about
Three levels of personalization and what each gives the reader. Conceptual diagram, no data.

Here is that order in an email from our US test, anonymized: no company, person or program name, and no numbers that would identify the company.

"[Company] recently opened a channel account manager role to recruit resellers and MSSPs in one US region.

For a young channel, the first partners shape the whole program. The best fits are usually partners that already have a practice in your category and are hearing related questions from their clients.

For [Company], that could mean a verified list of those partners in the region, followed by an email and LinkedIn campaign. Your channel team would get the interested partners with context, while terms and onboarding stay with you."

The first sentence is the signal. The second is the problem we think it creates for this company. The third is what we could do about it, and where our part ends. None of it depends on the person's profile.

The table below shows the difference on examples. They are fictional.

Personalization lineReason for contact
"Loved your post about the conference in Austin.""You opened a role to recruit MSPs in the Midwest. The first partners usually come from firms that already sell to your buyers but don't work with you yet."
"Congrats on five years at the company.""Since you moved from direct sales to a partner model, new revenue depends on partners who start selling in their first quarter."
"I see we both studied in Boston.""Your new mid-market plan reaches a different buyer than your enterprise deals. Someone has to run those first conversations."

How do you know the reason is good before you send?

You don't, and that is the point of writing it down. A reason for contact is a hypothesis about a company you haven't spoken to yet. "Software companies hiring a channel manager need more potential partners than their team can find on its own" is a sentence the market can confirm or reject. "Channel is a good market" is not.

Before sending, decide what will count as confirmation, what as rejection and what as an unclear result, and record every reaction, including refusals and questions. Our US test works this way. It is still running, and we don't have results yet, so this note makes no claims about reply rates. When we have them, we will write about what held up. More about how we run such tests: offer and segment testing.

Where does AI help, and where does it not?

AI is useful for most of the work around the reason: finding signals such as job posts and partner announcements, summarizing what a company sells and to whom, and drafting a first version of the email. That is how we use it.

Two decisions stay with a person. One is whether a signal is a reason to write at all: a new channel role can mean a company wants more partners, or that it is replacing someone. The other is whether a reply is a real opportunity for the client, which depends on knowing how that client sells. I wrote more about who that person should be in my notes after ApolloNEXT.

Three questions before you send the first email

  • If you remove the personalized line, is there still a reason for this company to reply?
  • Does that reason come from how the company sells, or from facts about the person?
  • Can you write the reason as one sentence the market could reject?

If the answer to the first question is no, a better first line will not fix the email.

Sources and notes

Sahni, N. S., Wheeler, S. C., and Chintagunta, P. (2018). Personalization in Email Marketing: The Role of Noninformative Advertising Content. Marketing Science, 37(2). Field experiments with three companies; open and lead rates as reported in the study. The emails were marketing messages to consumers, not cold B2B outbound. gsb.stanford.edu

Gartner, press release of June 25, 2025: survey of 632 B2B buyers conducted in August and September 2024. gartner.com

Superhuman, product announcement (2025): AI labels such as cold pitch keep those emails out of the main inbox. Vendor source. superhuman.com

The email excerpt comes from MSGA's US outbound test (October 2026), anonymized with the author's consent. The table examples are fictional. Statistics from outreach tool vendors were left out because they do not report sample or method.

Let's take a look at the architecture of your sales.

Two ways in, depending on where you are.