The best posts on buying a small business are the ones like this: what actually moved the numbers after close.
Worth noticing if you're buying: all four of his examples leave a trail you can find before you close, and each one tells you something.
Leads that got one touch: the call records and the CRM. That's money left on the table, and you can count it.
Measurement remakes: production records and material costs. Rework shows you what's really happening to your margins.
Finished jobs waiting on install: the gap between production-complete and install dates. A handoff nobody owns.
Installs running long: estimated vs actual hours, if anyone logs them. If they do, that's proprietary data you can price and schedule from. If they don't, you're guessing, and starting to log it is value you add on day one.
He found them by running the business for 30 months.
A buyer can find them in diligence and walk in on day one with a plan to fix them with the proper ops audit.
I acquired a window treatment manufacturing and installation business with an SBA loan in April 2024 doing about $3M in revenue.
We did $3.9M in 2025 and are on pace for around $4.8M this year.
One thing I’ve learned is that growing a business from $3M to $5M is a very different exercise than I expected.
When I bought the company, I figured the biggest opportunities were pretty obvious. Improve marketing, grow the sales team, expand geographically, increase production capacity.
Most of which we have done and have resulted in growth.
However, a surprising amount of our growth has come from fixing really basic stuff.
We manufacture our own shutters, sell several other window treatment products, and handle our own installations. We’re effectively running a manufacturing business, a retail sales business, and a service business under one roof.
This creates some complex workflows with many opportunities for things to go wrong like…
-30% of our leads never get more than 1 touch and never get booked.
-A measurement error turns into a full-job shutter remake.
-Production finishes a job that sits in the warehouse because nobody scheduled the installation.
-An installer takes 5 hours on a job we estimated at 2 and we have to reschedule the day’s remaining jobs.
Individually, none of these things are a huge deal, but multiply them across thousands of jobs and they add up to a 5, probably 6, figures of dollars 💸.
So we’ve been putting a ton of effort into relatively boring things. Better follow-up processes, production scorecards, tracking install hours, reducing remakes, and getting clearer about who owns what.
The other thing I’m realizing is how much of the business still depends on me or a couple of key people knowing what needs to happen.
That’s probably the biggest challenge I’m working through right now.
I want to get to $10M+. I want to make more acquisitions and I want to spend more of my time working on those opportunities instead of day-to-day operations.
But I’m becoming increasingly convinced that buying another business before I’ve figured out how to make this one less dependent on me would be a mistake.
I could probably push this business to $6M or $7M by continuing to do more of what we’re already doing, but if that growth requires me to be even more involved than I am today, I’m just making myself a bigger bottleneck. Eventually, I’ll be the very thing preventing us from growing beyond that point.
That’s the part of scaling a small business that’s been harder than I expected. My original thesis was to use this business as a platform for bolt-on acquisitions, and that hasn’t changed. Before I start adding spokes, though, I need to make sure the hub is strong enough to support them!
It’s taking longer than I’d like (and 60% growth in 30 months ain’t bad!), but I’m confident we’ll get there in the next 6–12 months.
1
5
122
Call the shop at 6:40 on a Tuesday night.
That's check #1 of 11.
The rest are the office problems a QoE never looks at, how to put a dollar number on each, and where it goes in the deal.
18
4
1,038
most companies are using AI to do the same work, faster.
that's the smallest thing it does.
the bigger change is work that never got done at all, because nobody could justify the hours.
no firm pays an analyst to read every deed, permit and zoning letter in a county every morning. the math never worked, so it went unread.
In sept someone requested a zoning letter on a 204-unit apartment property in WA state. ten days later it sold for $102M.
it was a marketed deal. the letter still told anyone watching that a buyer was underwriting it, ten days before the close.
A $102M trade is a sales event for a dozen businesses that aren't the buyer. the lender financing it. the property manager the new owner hires. the contractors bidding the renovation. a seller who may now have 45 days to find a replacement property.
every one of them would rather hear about it ten days before the close than read about it after.
we've built systems like this for CRE desks, for data center builders, and for a biotech team. different records, same pattern:
the information was already public
it mattered to one team, at one moment
nobody had the hours to find it in time
so when we look at a business, the first question is: what did you stop doing because you couldn't staff it?
try it on your own team. write down what you'd do if hours were free. read every filing in your market. call every customer who went quiet. check every change a competitor makes.
that list is worth more than any efficiency audit.
efficiency is the floor, not the point.
2
46
Yesterday I said I'd scored all 338 licensed HVAC contractors in King County.
First thing that jumped out is that the work piles up at the top.
Seattle publishes its permits as open data. In the last 12 months, 2,242 went to shops on my list.
One shop pulled 952 of them.
The top three pulled about two-thirds. Only 38 of the 338 pulled any at all.
Payroll says the same thing. 64 of these shops match a 2020 PPP loan. 7 reported 100+ jobs. 30 reported fewer than 10.
The rest of the list: 172 run from a house, 114 got licensed in the last five years, and 175 have one person on the license.
338 licenses. A handful of real dogs.
Hunting an HVAC deal around Seattle?
DM me MAP and I'll send you the top 5.
1
71
I run a few small businesses with no employees. Agents handle the office work. I watch it, fix it when it breaks, and make the calls.
Now I'm taking that same setup into the trades.
I started with HVAC in King County (WA): 338 licensed contractors, scored from public records.
- Licenses
- Principals
- Insurance
- Parcels
- Permits
- PPP loans
Small enough that I can check the output by hand before I trust it. Every Monday at 7am it rescores the list and tells me what changed.
County and trade are just inputs. Once King County HVAC holds up, I add plumbing and electrical in the same county, then other metros. One list I trust before I start the next.
The other half is going into shops. The person on the phones, the dispatcher, whoever chases estimates: each one gets their own AI help. The owner keeps the team. The team gets faster.
Automation is not the answer.
If you own a trades company, you're buying one, or you roll them up, follow along.
This week I'll be dropping more sauce on what you can already know about an HVAC shop before you send an LOI. Six public records, no NDA, and one of them tells you whether the owner holds the building.
2
1
90
Muse is going to nuke a lot of consumer apps
Great product from @Meta
2
86
The easiest AI service we’re selling to home service businesses
A 45-minute AI Operations Audit.
We map your biggest bottlenecks (usually lead response, scheduling, or quoting) and hand you a clear list of ready-made tools that fix them this month.
No custom code.
No 6-month project.
Just diagnosis + prescription.
Average result so far: 8–15 hours saved per week + fewer missed jobs.
If you want one of these, comment “AUDIT” or DM me “audit” and I’ll send the details.
Multi use baseball facility
No better place to work while the kiddo gets some work in baseball & soccer
If you don’t have your own sports facility for your kid then idk what you’re thinking
Huge ROI
2
67
I think you truly need to eat shit, and fail hard to appreciate the moment when you reach that next level
Fail forward friends
1
57
Yeah so this is real
Also might be kind of invasive meta can read our minds now but whatever we’ll talk about that later
Super cool service we’ll definitely have to try out!
We’re sharing the next major milestone in our non-invasive brain-to-text decoder research: Brain2Qwerty v2.
Building on v1, which was published today in @Nature, Brain2Qwerty v2 is the highest-performing end-to-end pipeline capable of real-time sentence decoding from raw brain signals. It advances beyond character-level performance to decoding words and semantics, enabling accuracy for overall communication.
We believe this research has the potential to make a real difference for the millions of people who suffer from brain lesions or disorders that prevent them from communicating.
🧵👇
1
88
Anthropic coming for your startup
Just nerfed a ton of VC funded companies
Who’s next?
Introducing Claude Tag, a new way for teams to work with Claude.
In Slack, Claude joins as a team member with access to the channels and tools you choose. Tag Claude in and delegate tasks to it while you focus on other work.
3
95
Ten dollar VPS
n8n and Claude wired together
you just replaced most of what an automation agency bills four-five figures for
The whole setup takes about ten minutes.
replace your automation agency if they can't strategically help you
building the automation is no longer enough
2
2
78
The automation was six boxes connected by lines.
They invoiced you f$5k
That gap has completely collapsed, and a lot of people have not noticed yet.
AI automation agencies are done
RIP
1
44
Think twice about the bouncy house biz
Crazy
11 injured, 3-year-old girl dead after bounce house accident:
bit.ly/4x8k8Un
1
70
Greg Bennett retweeted
When you’re starting something new, whether it’s a business or career change
You cannot tell anyone a thing about it unless they have already taken the chance you’re about to take
Someone who is pretty close to me, is about to leave their cushiony job to get into a business
They told way too many people already and they’re already getting negative feedback before the business even started
What you have to realize is nobody wants to see you do better than them or even see you take a chance at becoming better than them
You must have delusional confidence in yourself in order to see any sort of major success in life
Stay delusional kings
Hermes is one of the most impressive agents ever shipped.
It’s also the wrong tool for almost everyone downloading it.
If you’re trying to automate a VA (inbox, scheduling, follow-ups, content), you do not want an agent that rewrites its own playbook as it runs. You want one that does the same thing tomorrow that it did today.
The whole selling point of Hermes is that it learns and “improves” itself. That’s exactly the property you don’t want when the work is boring and repeatable. The day it teaches itself something slightly off, you’ve turned a VA into a problem you can’t trace, because the thing edited its own brain and didn’t tell you.
Most people don’t need a self-evolving research agent. They need a predictable one. One strong model, does exactly what you scoped, nothing underneath it to drift. That’s the whole philosophy behind what I run (nano claw).
Want a self-evolving agent and you have the technical ability to catch it when it drifts? Hermes, all day.
Want to replace a VA and stop thinking about it?
You want boring. Boring wins.
2
1
53










