All articles
Data study8 min read

We analyzed 40 local cold email campaigns. One lever changed everything.

Not volume. Not the subject line. An opening sentence that cites a verifiable fact from the prospect's site. Here are the numbers, and the prompt we use.

In 2026, a B2B cold email gets a reply 8.5% of the time on average. Our first twelve campaigns ran at 3.1%. Same list, same product, same sender. We sent more to compensate, and burned our domain instead.

So we stopped guessing. We instrumented 40 local outbound campaigns (~18,400 emails, French B2B niches, first half of 2026) and changed one thing at a time. One lever changed everything, and it is neither volume nor the subject line.

The gist in 30 seconds
  • The #1 lever is not volume or the subject line: it is a verifiable opener (a fact the prospect confirms in 10 seconds on their own site).
  • Replies by personalization level: 2.1% (generic) → 4.3% (merge fields) → 9.7% (verifiable opener).
  • Cost: 2-4 minutes per lead, because only the first sentence changes. The rest is a template.
  • The exact prompt and the 3-touch sequence are below, copyable.

The real problem: invisible personalization

"Hope you are doing well." "I love what {company} does." These openers are not personalized, they are disguised as personalized. The prospect has read them a thousand times. Their brain files them as spam before the second line.

Merge fields ({first_name}, {city}) barely help: they only prove you own a spreadsheet. What changes everything is a fact the prospect recognizes as true about them, instantly.

An opening sentence should not flatter the prospect. It should prove you opened their site.
The rule we keep from these 40 campaigns

The method: three levels of personalization

We sorted every email into one of three levels, then measured the reply rate of each across the 18,400 sends.

1
Level 1
Level 0: Generic

No reference to the prospect. Good copy, but interchangeable. Replies: 2.1%.

2
Level 2
Level 1: Merge fields

First name, city, trade auto-inserted. Looks like personalization. Replies: 4.3%.

3
Level 3
Level 2: Verifiable opener

A sentence citing a real site defect (form, mobile, CTA), checkable in 10 seconds. Replies: 9.7%.

2.1%
Replies, generic opener
4.3%
Replies, merge fields
9.7%
Replies, verifiable opener

Level 2 does not just double level 1. It is 4.6x the generic opener and doubles merge fields. It also converts better into meetings: 2.8% of level-2 emails led to a booked slot, versus 0.9% at level 1.

Where the effect is strongest: the niche matters

Not all niches react the same. Where a site defect directly costs the prospect money (a lost visitor is a lost quote), the verifiable opener explodes.

Web agencies
Niche ·A broken site is their own storefront
Replies (level 2) ·11.2%
HVAC / plumbing
Niche ·Broken form = missed calls
Replies (level 2) ·8.9%
Dental practices
Niche ·Failing online booking
Replies (level 2) ·7.4%
Restaurants
Niche ·Secondary site vs Google / social
Replies (level 2) ·5.1%

Lesson: do not hunt for the perfect opener, hunt for the niche where your observation hurts. The same email goes from 5 to 11% just by changing the target.

The prompt we use for the opening line

Here is the exact prompt. It takes a prospect's public data and returns a single sentence. Test it on three of your prospects before sending anything.

Prompt: verifiable opening line
You are an SDR. Here is a prospect's public data:
- Name: {{company_name}}
- Site: {{url}}
- Audit findings: {{site_score}}/100, {{defect_list}}
- Google rating: {{rating}} ({{review_count}} reviews)

Write ONE cold-email opening line (18 words max) that:
1. cites a VERIFIABLE fact from the site checkable in 10 seconds (no vague praise),
2. uses zero superlatives ("great", "amazing", "leader"),
3. makes the reader feel a human actually opened their site.

Return only the sentence, no quotes.

The "18 words max" constraint and the superlative ban are what stop the AI from sliding back into generic flattery. Without those two rules, it writes "I love your modern website", and you are back to 2%.

The full sequence, three touches

Only one sentence changes from prospect to prospect. The rest is fixed. That is what makes the method survivable at 200 leads a week.

Cold email sequence: 3 touches
Touch 1 (Day 0): Verifiable observation
Subject: {{first_name}}, your contact form
Hi {{first_name}},
[Line generated by the prompt above.]
We fix exactly this for {{trade}} in {{city}}. Worth 15 minutes?

Touch 2 (Day 3): Peer proof
Hi {{first_name}},
A {{trade}} near you had the same issue. After the fix: {{result_number}}.
Want me to show you how in 15 minutes?

Touch 3 (Day 7): Clean break
Hi {{first_name}},
I'll stop here. If it ever comes up, here's a one-page audit of your site: {{audit_link}}.
All the best.
When this method does not work
Three cases. One: the prospect has no site to audit (fully physical business). Two: enterprise sales cycles, where the opener matters less than a peer intro. Three: a commoditized niche where your observation costs the prospect nothing. In those cases, the verifiable opener is not your lever, change the angle.

Where AutoLeads fits

None of the above requires AutoLeads. You can audit by hand with Lighthouse, write the sentences yourself, send with the tool you already know. AutoLeads just runs the 10-criteria audit and the sentence generation on every lead automatically, then pushes it into Lemlist with the first sentence already written. We complete your sending stack, we do not replace it.

See for yourself
Don't take our word for it

Open the demo campaigns and judge the lead quality yourself, no card required. Worst case, you lose ten minutes.

If you remember one thing: stop optimizing the subject line, and start opening your prospects' sites. The first sentence is where the reply is won.

Frequently asked questions

What is a verifiable opener, exactly?

A sentence citing a fact the prospect can confirm by opening their own site in 10 seconds: a broken contact form, a low mobile score, a CTA below the fold. Not vague praise like "I love your work", which everyone receives and nobody believes.

Does it scale if every email is different?

Yes, because only the first sentence changes. The rest of the sequence is a template. We generate the opener with AI from the audit data, which brings personalization time down to 2-4 minutes per lead instead of ten.

Is cold email legal in France?

In B2B, email prospecting is allowed without prior consent if the subject relates to the person's professional role, with a clear opt-out. See the CNIL guidelines on commercial prospecting. In B2C, consent is required.

Do you need AutoLeads to apply this method?

No. You can audit sites by hand (Lighthouse, a look at the form and mobile view), write the sentences yourself, and send with Lemlist. AutoLeads automates the audit and sentence generation so you do not spend all day on it, but the method works without us.

What reply rate should you target at first?

On well-targeted local outbound, 6-10% replies is realistic once the opener is verifiable and the niche is narrow. Below 3%, the problem is almost always targeting or the opener, rarely the subject line.

Book a demo