I make a hotel management software called The Vestly, and like everybody who builds something, I had the same problem everybody has: I needed people to talk to. Customers. Leads. After getting burned by paid ads, I stopped chasing them and built a Claude skill to find customers instead. This is the story of why.
The old ad looked great until I picked up the phone
The numbers on the screen were honestly pretty nice. The campaign got me 19 leads. Each one cost about 27 rupees. I spent around 515 rupees total and reached almost 30,000 people. If you just look at that, you would think, wonderful, leads for the price of a cup of tea.
But here is the thing. A number on a screen is not a customer. So I started calling.
It was like ordering a box of apples and opening it to find most of them are painted rocks. They look like apples in the photo. They are not apples.
Almost none of these people were in the area I actually serve. Some of them told me flat out that they never filled in any form. A couple said their kid was playing with the phone and tapped through it by accident. The “interest” was not interest at all. It was noise that happened to have a phone number attached.
I even turned on OTP verification to stop the accidental taps. It helped a little. But really it just confirmed what I already felt in my gut. In the Indian Meta lead market, this is just how it goes. It has been this way for years. The system is built to count form fills, so that is exactly what you get: form fills, not people who want your product.
So that 27 rupees a lead was a lie. The real cost, the cost of one actual conversation with someone who fit, was much, much higher.
So I Built a Claude Skill Instead
I did not want a pile of strangers. I wanted to find the specific hotels that actually need software like mine, and go talk to them. That is a simple idea. The hard part is doing it carefully every single time.
So I built a Claude skill to do exactly that. Instead of sitting back and letting an algorithm decide who sees my ad, the claude skill goes out and finds the right businesses, checks if they are a good fit, and gets everything ready for me to reach out.
Here is the result. The skill found and cleaned up 40 leads. I did cold outreach to them. More than 25 of them were genuinely interested in The Vestly. I gave demos. I signed a few deals.
Think about that for a second. My old ad gave me 19 mostly fake apples. The claude skill, for The Vestly, gave me 40 real ones, and 25 of those wanted to take a bite. That is why I never bothered running an ad for this product. I already knew where that road ends.
Also Read : Ad revenue optimisation on a regional language digital news property
The Tweaks That Made My Claude Skill Work
Now, the interesting part is not that I used a tool. Anybody can use a tool. The interesting part is what I taught the tool to do.
Most lead tools just hand you a long list of businesses with phone numbers. That list is mostly useless, for the same reason a phone book is useless if you are selling ice. The names are real, but the list has no idea who actually wants ice.
So I taught the Claude skill a few small habits. A careful, patient person would do these same things by hand. I just made the skill do them automatically, every time, and unlike a tired human it never gets lazy or bored.
First, it filters tightly by place and by type. Not just “hotels,” but the kind and size of property that actually needs my software, in the regions I can actually serve. This one trick alone killed the whole “out of my area” disaster that wrecked the ads.
Second, it looks for signs of pain before reaching out. A hotel that clearly has the exact problem my software fixes deserves a conversation. A random hotel does not. So the skill hunts for those signs first.
Third, it goes for the right person instead of just any inbox. Reaching the owner who can say yes beats reaching a front desk email that goes nowhere.
Finally, every lead comes with a note explaining why the skill picked it. Because of that, the message I send actually speaks to that hotel’s real situation. It is not a copy-paste blast that everybody can smell from a mile away.
What I actually learned
A cheap lead and a good lead are not the same animal, even though they look identical on a dashboard. Twenty-seven rupees means nothing if the person on the other end never wanted to hear from you. Meta rewards the ad system for counting form fills, so naturally it hands you form fills. You cannot be surprised when a machine does exactly what you paid it to do.
Now, cold outreach has a bad reputation. People think it is painful and pointless. But cold outreach to a carefully chosen list is a completely different thing from cold outreach to a random pile. The problem was never the calling. The problem was the list. So once I fixed the list with a few simple tweaks, the rest more or less took care of itself.
I am not saying ads are worthless. For getting your name out there, they have their place. But for a product like The Vestly, where every customer deserves a real demo and a real relationship, I did not even want to roll the dice on another junk-lead pile. From day one, the outreach machine was the better bet. And it keeps working. It does not burn money for every set of eyeballs. Instead, it just keeps finding better people for me to talk to.
So if you are fighting the same battle with junk leads, here is my one piece of advice. Stop chasing a low cost per lead. Chase fit instead. First build the list carefully, and then the customers will follow.
If you want to use my outreach and lead generation claude skill, ping me and I will share it with you.
Building The Vestly, a hotel Property Management System. More on what I am up to at iniyan.in.