Ivona MikulčićNovember 27, 20256 min readLast update: Oct 7, 2026
Tiramisu Cold Email: How We Booked Rand Fishkin and Aaron Ross With 35 Emails (Step by Step)
Type “tiramisu cold email” into Google and you’ll get mascarpone, ladyfingers and espresso. This is not that. Tiramisu is the name we gave to a cold email technique for B2B outreach, and we used it to fill a webinar guest list with people we had no business getting a reply from. Two campaigns. 35 emails total. Guests like Rand Fishkin, Aaron Ross, Jill Rowley, Anthony Iannarino, Tim Soulo and Morgan Ingram said yes.
In this post, I’ll show you exactly how we wrote those emails, what we changed between campaign one and campaign two, and how to run the same play in 2026 without spending your entire week inside a spreadsheet.
What the Tiramisu cold email strategy is
The Tiramisu strategy is adding one hyper-personalized line to a cold email that changes completely for every single prospect, then building the rest of the message around it.
Every prospect eats their own Tiramisu. One custom sentence (or block) per person, written from real research, dropped into a custom variable so the campaign still runs as a sequence.
That’s the whole idea. The execution is where people quit.
Why we call it Tiramisu
As a cake, tiramisu is good. So is this cold email strategy.
It’s about writing genuine intro lines and adding custom sentences that change for each prospect. The name stuck internally, people in the community started using it, and now here we are writing a sales article that keeps getting confused with a dessert recipe.
How the Tiramisu tag works in your .csv
The entire work happens in your .csv file that you eventually upload to lemlist. Here’s how it looks.
You create a column called Tiramisu, you write one custom block per row, and your email template calls it like any other variable.
Imagine the following cold email structure:
Hey {{firstName}}, {{Tiramisu}} The rest of the email…
The {{firstName}} tag is replaced with Guillaume. And {{Tiramisu}} becomes:
“I’ve checked out your $1M ARR article the other day. It’s brilliant and congrats on reaching this huge milestone. I remember reading the first episode (the $250k ARR one you wrote) and really appreciate that we can follow your journey in such a detailed way.”
The result is a unique email that only Guillaume will receive.
And it works because it grabs attention and shows you’ve done real research on each person.
You may think this takes a long time. It does. Writing 20 of those by hand is a full afternoon, minimum. The good news is that the research you do while writing them is the same research you’d need for the follow-up call anyway, so nothing is wasted.
Campaign 1: 20 invitations to people we wanted to interview
The goal was to bring industry leaders to our live webinars, where we’d produce actionable sales and growth advice people could use to close more deals. Attendees could ask questions live, since every interview is recorded in front of our community.
The first campaign went out to 20 people we wanted to interview.
Here’s why it worked:
The
{{Tiramisu}}tag was replaced with a genuine compliment for that person. In most cases it was 1-3 sentences commenting on their recent work, plus one more sentence on why we admire them.{{Tiramisu2}}was changed with our topic proposal. The actual webinar subject we wanted to record with them.The pitch showed how their participation would bring them exposure to our audience.
Credibility is key. We wanted everybody to know we’re practitioners of our craft with a real track record, so the email carried proof of that. Today we’d pull those signals automatically with AI Agentic Enrichment instead of digging through tabs. For reference, the proof point we lead with now is simple: lemlist is rated 4.6/5 on G2 with 2,000+ reviews.
The next step was always clear. One ask, one action.
We also A/B tested two different subject lines on that send.
Pro tip: always A/B test
Let your prospects decide what subject line, image, call-to-action or sequence works best. Don’t waste your time debating these in internal meetings. Split-test it instead, then let the reply data settle the argument.
Recommended read: How to run A/B tests in lemlist?
Campaign 2: 15 prospects and a stronger pitch
A few months later, the second campaign began.
By then we’d already interviewed a batch of sales and marketing experts, so we had more credibility and wanted to change the pitch. We targeted 12 + 3 people for new episodes. The first 12 were part of the Tiramisu campaign.
What we changed:
We bragged a little. Having recorded several episodes already, we named the previous guests. Those names bring respect in a prospect’s eyes and answer the “who else has done this?” question before it’s asked.
We simplified the CTA even further. We didn’t want anyone thinking too hard about what to do next.
Same structure, more social proof, shorter ask. The strategy held up across both campaigns, and it’s how we ended up with a guest list most teams would call unrealistic.
How to run Tiramisu at scale in 2026
Here’s the honest tradeoff with the original method: it’s manual. One person, one afternoon, maybe 20 to 30 prospects. That was fine for a webinar guest list. It falls apart the moment you need 500 prospects a month.
That’s the part that changed since we first ran this.
Smart messaging now writes the Tiramisu line for you. It reads the prospect’s website, LinkedIn activity and company context, then generates the custom opener in the same variable slot you’d have filled manually in your .csv. You keep the structure of the strategy and drop the data-entry part of it.
AI Agentic Enrichment handles the layer underneath: finding the signal worth complimenting in the first place. Funding news, hiring patterns, recent posts, tech stack changes.
And lemAgent goes further, running the research-to-draft loop as an agent so your reps review instead of write.
One warning, because this matters. Automation applied to lazy research doesn’t fix the email. It ships the same weak compliment to 500 people instead of 20. Feed the AI a real signal and it writes a real Tiramisu. Feed it a company description scraped from a homepage and it writes the thing everyone deletes.
Pro tip: review your campaign before you hit send
lemlist lets you add last-minute changes for specific people just before your campaign goes out. You’ll find it in the “review” stage.
Use it. Even with AI-generated openers, a 30-second human pass per prospect catches the line that reads slightly off, and it’s where you can push personalization further for a specific person or a small group you really want.
Think of this review stage as the manual version of what lemAgent now does on autopilot. Same quality check, different amount of your time.
More info: personalization superpowers for cold emails.
What to remember about the Tiramisu strategy
Our community loved the webinars, and getting to pick the brains of people you admire is still one of the better perks of this job. You can see all episodes on our YouTube channel.
The things that actually make Tiramisu work:
Everything comes down to your research. The output of that research is a custom sentence for each prospect.
Anything can be a
{{Tiramisu}}tag. Your intro, your value proposition, your CTA, name it.Every prospect eats their own Tiramisu. One sentence, one person.
Volume is the constraint, not creativity. Automate the research and the drafting, keep the review.
If you want proof that personalized outbound still books meetings at scale, read how ElevenLabs built their outbound motion with lemlist. Then grab a starting point from our cold email templates and write your first batch of Tiramisu lines this week.
Over to you
Pick 20 prospects you genuinely want to talk to. Spend an afternoon writing one real sentence for each of them. Then compare the replies to whatever you sent last month.
If the afternoon is the blocker, let the platform do the research and the drafting for you. Start a 14-day free trial and send your first Tiramisu campaign.
Convert strategy into pipeline
Launch your first outbound campaign with verified leads, AI sequences, and multichannel outreach from one platform.