Before we sold The Prospector, we ran it on our own agency outreach for 90 days. We figured if we couldn't prove it worked on our own business, we had no business selling it.
Some things went exactly like we expected. Some things didn't.
What we expected
We expected it to find more leads than we could manually. That happened immediately. Littlefinger scanned 247 companies in the first week. On a good week doing it ourselves, we'd find maybe 15-20.
We expected the emails to be decent. They were. Tyrion's first drafts were better than what most of us write on a Monday morning, which is when we used to do outreach.
We expected time savings. We got 7+ hours back per week.
What we didn't expect
Jon Snow got stricter on his own. By week 4, his rejection rate went from 91% to 94%. Same number of leads coming out, but higher quality going in. We hadn't changed any settings — he was learning from which leads we actually responded to.
The Hound barely rejected anything after week 3. In week 1, he sent about 15% of emails back to Tyrion for rewrites. By week 3, the rewrite rate was under 5%. The quality loop was working. Tyrion was learning what "good enough for The Hound" meant.
We found leads we never would have found manually. Not just more leads — different leads. Companies in adjacent industries we hadn't thought to target. Littlefinger's search criteria were broader than our own mental filters, and some of those "weird" leads turned out to be the best ones.
The honest part
It wasn't perfect from day one. The first week's emails were generic. By week 3 they sounded like us. The ICP scoring was loose at first — Jon Snow needed a few weeks of feedback to dial in.
Setup took about 2 hours the first time (we've since gotten it down to 90 minutes for new users). Configuring your ICP thoughtfully matters. Garbage in, garbage out — that's still true even with AI.
Why we decided to sell it
By week 8, The Prospector was consistently outperforming our manual process by every metric. More leads, better quality, zero hours of work. At that point, keeping it to ourselves felt wasteful.
So we packaged it up, wrote the setup guide, and put it on Gumroad. The version you get is the same version that's been running our outreach since January.