How we built an estimator that prices like the owner

Hey,

This one's different. No tip this week. Instead, here's something we've been building since spring, how we built it, and what it taught us about the time between the on-site estimate and the signed proposal.

The problem wasn't the work

In June, on a call with a tree service owner we work with, he named the thing holding his business back. It had nothing to do with crews or equipment. Proposals were the choke point.

"If we increase our lead volumes," he said, "what it does is it drags out the quoting process."

He'd been holding back on taking more leads, because a quote that sits for two weeks has less of a shot than one that goes out this week. Every good multi-tree job meant about an hour on site, then two or three more at a desk getting the price and the wording right. What he wanted was a quote anyone on his team could put out that priced and read like his.

His ask was simple: "How do we both do it quickly and efficiently, but also have it accurate from a number standpoint?"

Step one: don't guess, read his history

We started going way back through his own jobs. We pulled four years of quotes out of his CRM, thousands of line items, his whole price catalog, and his shorthand for writing up a job.

What came out was a clear system. He prices on time and resources, never per tree: how many guys, how many hours, what equipment, whether a crane comes out, how much debris leaves the property. When we asked what moves the price most, he ranked it: access first, then the size of the tree, then cleanup, then timeframe, then condition.

Step two: prove it before he sees it

We set a bar before he saw anything: the estimator had to rebuild 95% of his past jobs within $50, or it didn't leave our hands.

We took 218 jobs he'd already sold and ran them back through. 95% came out to the dollar. 97.7% landed within $50. The few that missed were jobs we'd left out on purpose: commercial programs, storm work, discounted travel jobs.

We sent him a working calculator the next day. On our next call he confirmed the math works.

Step three: find out the math was the easy part

Then he told us what he actually wanted, and it changed the whole project.

He didn't want another form to fill out. He wears a Plaud, a small recorder that captures his on-site estimates and transcribes them, and he feeds those transcripts to AI. He wanted to say "this part of the job is this, describe it this way," and have the proposal built from that.

That's when we figured out the math was maybe a third of the problem. The rest was two things no calculator touches.

The walk is not the proposal

We took one real job. He'd walked the property with the homeowner and looked at about twenty trees. The proposal he sent had four lines.

He left a dozen trees off on purpose, ones he'd called low priority or someone else's problem. The rest got grouped by type of work, each line marked included or only recommended, plus one small item at no charge. Plant health care went on a separate quote.

That editing is the most valuable judgment in the whole process, and it's exactly what generic estimating tools get wrong. We wrote his habits down as rules: what makes the cut, how lines get grouped, what's included versus recommended. Could you write down your own version of those rules? That question ended up driving the whole project.

The writing is what eats the time

The second thing was the words. Each proposal opens with a letter to the customer, covering what his team saw on the on-site estimate and what they recommend tree by tree, and why, before a single price shows up.

Our first pass pulled the wrong text, his internal work notes. The real voice was sitting in a different field: 215 jobs of his own letters to customers. That became the model. Nothing gets invented. It's built to write the way the owner already writes, and the owner reviews it before anything goes out.

He told us the writing was the part that takes the most time.

Then we ran it for real

We ran his recorded notes from that twenty-tree estimate through it, start to finish. It picked the same four lines and marked the same ones included and recommended. It left off the same trees, put plant health care on its own quote, and kept the same no-charge line.

The one miss: it guessed five crew hours on the biggest line where he'd figured 4.75. That told us something too. The owner should always make the final call on hours. That's his eye, and it should stay his.

What we're building now

All of that was built around one owner. Now we've rebuilt it so any tree service company can use it to price jobs and turn on-site estimate notes into proposals that read like the owner wrote them. It runs on four things you already have:

The customer gets a clean proposal with no hours or rates showing. The crew gets the breakdown.

Both pieces work today, the pricing and turning on-site estimate notes into a finished proposal. We're putting them together into one tool now.

A few things worth knowing before Wednesday. You don't need years of clean data to start, because the rate card is what a crew and your equipment cost you, and you already know that. It works alongside whatever CRM you use, since the proposal comes out ready to send or paste in. Your numbers and your proposals stay yours. And it means your pricing stops living only in your head, which matters whether you hand the company down, sell it, or do something else entirely.

It isn't open yet. Joining the list costs nothing and doesn't sign you up for anything. If you want in when it is, reply "estimator."

Come see it Wednesday

I'm doing a live walkthrough next Wednesday, October 28, at noon Central. A short look at how it works, then your questions. It's raw. No slides, no editing.

Got a question you want answered? Reply with it ahead of time and I'll make sure we get to it.

Join the live walkthrough here (Wednesday, October 28, noon Central)

Talk soon,

Jacob