Field Notes
What a Fractional CTO Does for a Builder (It Starts With the Invoices)
The first thing I did as a fractional CTO one week in August was read a software bill.
GoHighLevel, the CRM platform I run for every client, had spent the summer on a promotion. Its AI tier was free from June through August. Fifteen of the eighteen client sub-accounts I administer were sitting on that tier, and on September 1 the promotion ended and the tier started billing. $850 a month, across the book, for as long as nobody looked.
So I pulled the usage. Ninety days of real AI activity across all eighteen accounts came to $20.74. Four accounts had used it at all. Eleven showed exactly two events each, logged 83 days earlier, the day the feature was switched on and never again.
Nobody was stealing anything. A promo was ending, the way promos do. I moved every account to usage-only billing, checked that no client-facing bot went dark, and closed the tab. Twenty minutes.
That is the job.
Reading the bill before the bill reads you. And in a construction company, there is almost never anyone whose job it is.
The seat a builder never filled
Ask what a fractional CTO does and the answer people expect is code. Some of it is. Most of it is ownership. In a building company, a CTO owns the website as a working system rather than a brochure. The CRM and what goes in it. The proposal system. The client portal. The analytics, and whether they connect to a dollar. The vendors, the bills, and the keys.
A builder anywhere from two million a year to twenty has never employed this person. Which means every item on that list belongs to nobody. The site belongs to whoever built it in 2019. The CRM belongs to whoever opened it last. The passwords belong to a sticky note. The bills belong to autopay.
Nothing on that list is broken, exactly. It is unowned, and unowned things drift.
The fixes nobody sees
Custom Crafted Homes, a design-build firm on Cape Cod, is where I built that department this year. The site we inherited had passed through several vendors over a decade, each of whom did their own job and left the parts that were nobody's. Every builder's site has a version of this list. This is what was under that one.
- A phrase that hundreds of people a month search for, ranking to a page that returned a 404.
- Redirects from the old site pointing the wrong way, so traffic for one service landed on a page about another.
- A homepage with no H1 heading at all. Google was guessing what the business was.
- A sitemap frozen at an old date, telling search engines nothing had changed in a year.
- Scheduled posts silently missing their publish time. The site's internal clock only ran when a visitor hit an uncached page, and the caching plugin made sure nobody did.
- The www and the non-www versions of the domain, treated by Google as two different businesses splitting one reputation.
- Pages no other page linked to, invisible to a crawler.
- A disavow file for links the site should never have had.
Not one item there is a redesign. No client will ever see any of it. The photos did not change and the logo did not change. What changed was what the site was standing on.
The count of phrases that site ranks for sat at 66 last December. Come July the number was 264, and page-one placements had climbed 10 to 26 across a single month inside that stretch. I won't list which phrases, because the phrases are the point of the work and they belong to the client now. What I will say is that every dollar of marketing that builder spends from here lands on a foundation that holds, and nine months ago it would have landed on the list above.
The plumbing is where a fractional CTO starts. The marketing half of the seat, the part that decides where the traffic comes from, is its own subject.
Nobody is watching your vendor either
This summer I looked at a site for a friend's business. Not a builder, but the lesson travels.
The site had been maintained by one web vendor for years. I audited the vendor's entire book: 1,616 client domains spread across five servers, a floor rather than a count. My friend's site had received one content edit in five years. Its analytics tag had been dead for three, reporting nothing to anyone, while the vendor's dashboard presumably stayed green.
Then I sampled 23 of the vendor's other clients. Thirteen of them had been moved to modern tracking. This one had not.
That is not a bad vendor. That is a vendor with sixteen hundred accounts and a queue, and an account that never made it to the front. Attention is finite, and nobody inside my friend's company was in a position to demand it. A fractional CTO's job, a great deal of the time, is to be the attention. To be the person the vendor's queue cannot skip, because somebody is going to ask.
One more lesson from that audit, and it is on me. My first draft of the findings argued the client was paying for nothing. Then I found out what the client was actually paying that year, which was nothing. The argument collapsed and got rebuilt as a plain condition report. Establish what a company actually pays before building the case about what it gets.
Keys, and how far a bad day travels
The same platform that produced the $850 bill also issues API tokens, and its default is an agency-level key that can read every sub-account at once. Nothing has to leak for that to be a problem. A single credential that opens eighteen doors is a blast-radius problem before it is ever a leak problem, and the fix is boring: one token per account, scoped to that account, rotated on a schedule.
The other habit is reading the limits page instead of the marketing page. A platform whose sales page promises one ceiling and whose limits page quietly states a lower one will fail you on a Friday afternoon with a client on the phone. Somebody has to have read the fine print before the workflow depended on it.
The two rooms a builder needs built
Once the plumbing holds, the build list is short, and it is the same for nearly every builder because 785 contractor websites said so. In that study, 95.8 percent of general contractors showed no proposal system a client could reach and 93.8 percent had no client portal. Those are the first two things a fractional CTO builds after the fixes above.
The first is the bid. Not an attachment exported from Word, but a page on the builder's domain, with a measured 3D study of the house on the client's drawings and the numbers explained beneath it. The Bid Room is what that looks like when it ships.
The second is the portal. A gated page in the builder's brand where a signed client watches their own house go up, with job-site photos and the next week's schedule. The builder above has two of them live. The homeowners open the portal instead of dialing the superintendent, and that year of not dialing is where referrals are born.
Under both of those sits the CRM, its pipeline stages defined on paper, and the measurement that connects a visit to a signature, so the owner knows which channel produced each signed job. What the site itself has to do to feed those systems is a shorter list than you would guess.
What the seat costs
A full-time CTO costs something like $250,000 a year in salary, and a builder needs about a fifth of one. Most fractional CTO services are sold as strategy. This one starts with the invoices. The fractional version is the same person for part of the week, for a modest base, then a percentage of the revenue that can be traced back to the work, with the tracing rules agreed in writing first. The full economics, and how the CMO half of the seat fits, are in the pillar of this series. Everything the seat operates, from the site up, is listed layer by layer.
Every builder already has a technology department. It is spread across a vendor, an autopay, and a sticky note, and nobody is reading the invoices.