On Our End
Before any of this I ran a lawn mowing business, and I thought I understood it.
I had opinions about which clients were worth it and which weren’t. They weren’t lazy opinions. They came from doing the work myself, for years, and paying attention while I did it.
Some clients would ask for extra work, and I’d give them a bit of a hard time about it, because I was sure we weren’t doing well on their properties. Others I’d bend over backwards for. I never wrote any of that down. I just knew.
Then I learned to job cost properly. Actually measured it, job by job, instead of going on feel.
In some cases I was totally wrong. Some of the clients I’d been giving a hard time were the ones making us money, and some of the ones I was bending over backwards for weren’t.
That’s the part that stuck with me. It wasn’t just a spreadsheet being off. I’d been treating people differently — being short with some and generous with others — based on a feeling that turned out to be wrong, and none of them had any way of knowing why.
It changed what I thought optimizing the business even meant. The intuition I’d built up over years of being physically present for the work was simply not a reliable instrument, and I had no way of knowing that until I put a real one next to it. Everything I’ve done since, including going into marketing and building a career on measurement, comes out of that.
I mention it first because everything below is other people’s blind spots, and it would be dishonest to tell those stories without saying that I found the pattern in myself first.
A form broke
A client of mine had a contact form on his website. Ten million dollar business. Good operation, good people.
The form broke.
Nobody noticed for a month. It was a month of the summer, which in this industry is the month — it is the part of the year the whole business is built to survive on. A new person had just started at the front desk. She had no way of knowing that the number of emails arriving was wrong, because she had never seen what right looked like. And on my side, nothing looked wrong at all.
That is the part I want to talk about. Not the broken form. The fact that nothing looked wrong on our end.
Two more, same shape
I once spent a few thousand dollars sending traffic to the wrong client’s account. Everything about that campaign was healthy. The spend paced correctly. The clicks came in. The conversions tracked. Every number I looked at was a real number, accurately measured, and all of it was being delivered to somebody who wasn’t paying for it.
And once, during a snowstorm, we buried an office administrator in phone calls for a week.
That one is the one I think about. The calls went up — genuinely up, a real spike, exactly the thing you want to see. It looked like it was working extremely well for our end, so we turned it up further. What we could not see was a woman on the other end of it, drowning, who didn’t realise she was allowed to tell us to stop. Poor lady. Nobody did anything wrong. The measurement was correct. It just had nothing to say about the only thing that actually mattered that week.
Nothing was wrong. Everything was blind.
It took me a long time to see that these were the same story.
The zero form submissions were accurately zero. The campaign metrics were accurate. The snowstorm call volume was accurate. There was no bug in any dashboard I was looking at, no bad data, no mistake in the arithmetic. Every instrument I had was telling me the truth.
They were all measuring the wrong layer.
A quiet month and a broken form look identical if the only thing you can see is a count. A campaign delivering perfectly to the wrong account looks exactly like a campaign delivering perfectly. A spike that is burying somebody looks exactly like a spike that is making somebody money. In every case the number was fine and the reality underneath it was not, and there was nothing in my view of the world that could tell the difference.
That’s the actual failure. Not broken technology. A gap between what the provider can see and what is happening at the client — and a set of tools that report confidently from the wrong side of it.
Ten years of trying to see inside a phone call
The calls are where I spent most of that decade, so that’s the part I can show properly.
For years I could see two things: how long a call lasted, and where it came from. That’s it. I watched those numbers extremely closely. I got good at them. And they told me almost nothing, because a four-minute call can be a booked job or a wrong number or somebody complaining, and length and source cannot tell you which.
So we started recording the calls. Now the answer existed. It just existed inside thousands of audio files that nobody had listened to. Having the information and knowing it are not the same thing, which is a lesson I have now learned about five separate times.
So we asked clients to tag their own calls on the keypad. Press one if it was a good lead. This is the one I’d put in front of anybody building a reporting system today, because of how it failed. They hated it. They were busy, it was one more thing, and they did it randomly.
Randomly. Which means the tags came in looking perfect. Well-formed, complete, ready to chart — and meaningless. That is worse than having no data, because no data announces itself. You know you’re blind. Bad data that looks good lets you build an entire quarter’s decisions on top of noise and feel well-informed the whole time. It is the exact same failure as the broken form wearing a better suit.
So I paid somebody overseas to do the tagging by hand. Thousands of dollars, month after month, a person listening to calls about lawn care and marking them up one at a time.
And it worked. He was good at it, and he was a good guy — I have a clear memory of being on a video call and seeing my own screens glowing on his laptop in the background while he had beers with a coworker, and honestly, fair enough. The problem was never him. The problem is that I don’t trust any human being to hand-tag a million local lawn care calls. Not because people are careless. Because that is not a thing a person should be asked to do a million times, and anyone who does it that many times will stop being able to do it well, and you will never be able to tell which of the million were the ones where they stopped.
Every rung on that ladder was a workaround for the same constraint: we could not read everything, so we needed a way to guess the rest from a sample.
Then that constraint went away. We read every call now. All of them, properly — what was said, what was asked, what happened, what it was worth. There’s no sampling layer left to be wrong, no assumption to calibrate, nobody deciding at 2am that this one is probably a four.
I want to be clear that the hand-tagging wasn’t wasted. It’s what taught us what a good call looks like in the first place. You can’t automate finding a pattern until you know what the pattern is. The expensive manual version is how you earn the knowledge that the automatic version runs on. You just shouldn’t still be doing it ten years later.
The other half: stop having a separate end
The forms got the same treatment, just later.
We now run the servers the sites live on. The forms are ours, in-house, wired directly into spam filtering, conversion tracking, notification systems, the integrations that carry a lead onward — and error testing, which is the piece that exists specifically because of a broken form on a ten million dollar business in the middle of a summer.
That’s the same move as reading every call, and it took me a while to notice. The answer to nothing looked wrong on our end was not to look harder. It was to stop having an end that is separate from theirs. Every seam between a provider and a client is a place a failure can sit quietly, and the reliable fix is to remove the seam, not to stare at it more intently.
What it’s actually for
The strange part is what all that call data turned out to be good for once we had it properly.
It started as damage control — a way to stop being blind. But once you can genuinely read every conversation a business has with its market, you’re holding the most honest description of that market anybody has. What people ask. What they’re scared of. What they say right before they buy and right before they don’t.
So it does three jobs now: it educates prospects, it sells leads, and it supports clients. Same data, three directions, and all of it exists because a form broke and I couldn’t tell.
I’d like to say I’d catch all of this faster now, and mostly I would, because most of it is no longer possible to miss. But I’d be lying if I said I think I’ve run out of blind spots.
I was wrong about my own lawn business while standing in the middle of it, doing the work with my own hands, paying close attention. I only found out because I measured something I was sure I already knew. Nothing looked broken then either. It never does.
That’s the whole thing, really. The failures that cost the most aren’t the ones that show up as errors. They’re the ones where every number is correct, every report is green, everybody is doing their job — and the thing that matters is happening somewhere none of it can see.