When we personalize cold email at volume, we change one sentence per person and keep every other sentence the same. That one sentence holds the research, and the rest of the email is a fixed template. We settled on this after seeing what happens when every part of every email is different.
Why do we change only one sentence?
Our view is that a fully custom body for every lead is worse for deliverability than a fixed body with one changing line. So we store only the researched sentence, and sometimes the first subject line, for each lead.
The body and every follow-up are templates. Variety comes from spintax, which swaps in different wordings of the same sentence, and not from writing a new email per lead.
What does the changing sentence look like?
Here is a made-up example for an imaginary software company that sells invoice matching to accounts payable managers. Every name in it is invented. The table shows which sentences stay fixed and which one changes for each person.
| Sentence in the email | Changes per person? |
|---|---|
| I think {company} buys from a few hundred suppliers (ex: steel, fasteners, packaging), so your team probably matches a lot of invoices by hand. | Yes, this is the only sentence that changes |
| Our software reads each invoice, checks it against the purchase order, and only sends your team the ones that don't match. | No, it is the same for everyone |
| That keeps your Fridays free for real work, and other distributors already run it that way. | No, it is the same for everyone |
| Would it help if I sent a two-minute walkthrough? | No, it is the same for everyone |
The first sentence connects straight into the product sentence. It names a situation that is probably true for them, and the next sentence says what the product does about it. We cover why that connection matters below.
I think {company} buys from a few hundred suppliers (ex: steel, fasteners, packaging), so your team probably matches a lot of invoices by hand.
Our software reads each invoice, checks it against the purchase order, and only sends your team the ones that don't match.
That keeps your Fridays free for real work, and other distributors already run it that way.
Would it help if I sent a two-minute walkthrough?Which sentences are allowed in the email?
Every sentence has to do one of two jobs. It either piques interest by showing we know something about the reader, or it provides value by telling them what the product does. Any sentence that does neither is filler, and we cut it.
The same rule keeps the personalized sentence honest. If we swap in a random other prospect's name and company and the email still makes the same sense, then it is a template with their name in it. We do not count that as personalization.
When does a personalized opening line help?
It helps when our data holds a specific event or fact, such as a named program, a funding raise, or a facility with a stage. A line built on something like that shows we read about their actual situation. We avoid claiming more than the fact supports, so we phrase it as a guess, as in "I think" in the example above.
We leave the opener out when the data only says what type of company they are. One of our programs tested an opener that described a firm's business model back to it, and the reaction was that it made zero sense. It read as if we had never looked at anything real, so we now write that segment with a plain segment opener and the company name only.
We treat describing a company back to itself as worse than no personalization at all.
| What the data holds | What we write |
|---|---|
| A specific event, such as a named program, a raise, a new facility or a launch | One personal sentence, phrased as a guess |
| Only the type of company | A plain segment opener with the company name |
Why do the sentences have to connect?
A correct fact is not enough on its own. In one test, an email listed the news about a company, then described the product, then the benefit, then the proof, and every sentence was accurate. Our verdict was that there was no connectivity between the sentences, so it read like a data dump.
The fix is to let each sentence pick up where the last one stopped. The signal and the product share a sentence, joined by why it matters to the reader. Words like "which", "keeps" and "so" carry the reader from one idea to the next, and the ask follows from the last thought.
We use a simple test on every draft. If two neighboring sentences can be swapped without losing anything, the email fails. The fact is only the anchor, and the sentence after it has to say what we think that fact means for this person.
How do we check the copy before it goes out?
Our copy checker runs some of the same checks we run on every draft. The personal sentence is only as good as the facts behind it, and our list of public data sources is where many of those facts come from.
Common questions
How much of a cold email gets personalized?
We personalize one sentence per person and keep every other sentence the same. Our view is that a fully custom body for every lead is worse for deliverability than a fixed body with one changing line.
What counts as real personalization in a cold email?
We use a swap test on every draft. If another prospect's name and company can go into the email and it still makes the same sense, we count it as a template.
When is it better to skip the personalized first line?
We skip it when our data only says what type of company the reader works for. Describing a company back to itself reads worse than no personalization, so we use a plain segment opener with the company name.