There's a version of this post where I tell you that AI-powered personalized outreach is a game-changer and here are the results to prove it. That version would be more comfortable to write. But the actual numbers are what they are, and I'd rather share them honestly and think through what they mean than pretend we're further along than we are.
So here's the real story of the SiteLens outreach campaign, from database to inbox.
What We Built
SiteLens is our AI website audit tool. It scans a business's website across five categories — SEO health, AI readiness, trust signals, performance, and content quality — and produces a scored report with prioritized recommendations. The pitch is simple: most small businesses have no idea how their website actually performs across the signals that matter today, and they're losing customers because of it.
The outreach idea was straightforward: scrape business data for small businesses in markets we care about (Washington, Oregon, Arizona to start), scan their websites, rank the ones with the most room for improvement, and send them a personalized email that says: we already ran your audit, here's what we found, here's how to fix it.
The database ended up at 69,344 businesses total. Of those, 3,988 had valid email addresses. Of those, 8,240 scored below 65 (our threshold for a "hot lead" — a site that's visibly broken across multiple dimensions). The actual overlap between "has email" and "scores low" is the pool we're drawing from.
The Bounce Rate Problem
9.2% is high. Industry benchmarks for cold email hover around 2-5% for a clean list. Ours is almost double the top of that range.
The reason isn't hard to diagnose: we scraped email addresses from business listings, websites, and directories. This data is inherently messier than a purchased opt-in list. Small businesses close. Owners change their email. The person listed on the Google Business Profile hasn't worked there in three years. Some of the addresses were formatted correctly but never existed.
We had 32 unique hard bounces out of 349 unique recipients. That's emails that returned a permanent failure — mailbox doesn't exist, domain doesn't accept mail, that kind of thing. Hard bounces are the dangerous ones: too many of them and your sending domain starts getting flagged as a spam source.
The fix we're implementing is pre-validation. Before we send, we need to verify whether an address actually resolves. There are tools that do this via SMTP handshake without actually sending a message. We should have built that into the pipeline from the start.
The 4 Replies
Four human beings read our email and wrote back. That's a 1.1% reply rate on sends, or about 1.15% if you adjust for the bounces. It's low. But it's not zero, and I think that matters more than the raw number suggests.
The replies we got were genuine. Not unsubscribe requests, not angry responses. Actual people who read the email, recognized something they cared about, and responded. One of them had specific questions about their SEO score. Another wanted to know what it would cost to fix the issues we'd flagged. These are the conversations we were trying to start.
The conversion funnel from those four replies is still live. What we haven't done yet is close one. That part is on us — the follow-up sequences, the offer clarity, the closing mechanic. The outreach is generating leads; the sales process is the next thing to build.
What the Email Actually Said
The core message was: we already scanned your website, here's the score, here's the single biggest issue, here's what it's costing you. We kept it short — under 200 words. We referenced the specific business by name and URL. We included the actual numerical score from the audit.
The personalization wasn't AI-generated fluff. It was data-driven specificity: "Your site scored 52/100. The biggest drag on your score is mobile performance — your Largest Contentful Paint is 6.8 seconds on mobile, which is above Google's threshold for a 'poor' rating and probably hurts your local search ranking."
That's the kind of thing that makes someone stop and read, because it's clearly not a generic template. We actually checked.
The emails that convert don't convert because of clever copywriting. They convert because the recipient recognizes something true about their situation.
What We'd Do Differently
A few things, in order of impact:
- Validate emails before you send. The bounce rate is the most solvable problem on this list and also the most damaging to deliverability. Fix it first.
- Warm the domain longer. We were sending 50 emails a day from a relatively new domain. Even with good authentication (SPF, DKIM, DMARC all configured), ISPs need to see a track record before they fully trust you. We should have ramped up more gradually.
- Segment by industry earlier. A restaurant with a 55-score has completely different pain points than a law firm with a 55-score. The email that works for one probably doesn't land for the other. We're treating them identically right now.
- Build the follow-up sequence before the first send. We launched the initial outreach without a defined follow-up path. The four people who replied didn't get a structured next step — they got me responding manually. That doesn't scale.
Why We're Continuing Anyway
Because the math is still viable. We have 926 addresses left in the queue. If the reply rate holds at roughly 1%, that's 9-10 more conversations. If we improve the bounce rate and the follow-up sequence, we can reasonably improve both numbers. And we have 1,300 total eligible addresses before we even expand to new cities or states.
The product is real. The audits are genuinely useful — we've scanned over 26,000 sites now, and the scoring correlates with actual business outcomes in ways that are measurable. We just need to get the business development side to match the technical side.
The other thing worth saying: cold email at this scale is not a magic bullet. It's a volume game with a long tail. The businesses that will eventually become customers probably need to see the brand name a few times before they're ready to respond. The outreach is part of building that awareness, not a standalone conversion mechanism.
We're building toward a world where the outreach, follow-up, audit, and onboarding are all automated — where a business gets an email, books a call, and ends up on the platform without any manual work on our end. We're not there yet. But we know what the next three steps are, and that's more than we knew six months ago.