Field notes

Systems · November 11, 2026 · 7 min read

Before a CRM for solar installers, find where the cash falls out

A long-running installer with no CRM didn't need software first. It needed three leaks closed that the owners could already name.

The owners of a residential solar and electrical company told us exactly where their money was going missing in the first meeting. Service calls rang reps' personal phones. Finished service work sat for weeks before anyone sent an invoice. Upsells lived in one owner's head.

They had no CRM. Most people would start there: pick a platform, migrate the contacts, train the team, and hope the leaks close on their own.

We didn't. A CRM for solar installers is worth having, but it's a container. If you pour broken handoffs into it, you get broken handoffs with a login. So we started with the leaks, and built the system around them.

Why we didn't buy an all-in-one platform

The company runs 10 to 15 installs a month. We priced the all-in-one solar platforms against that volume and against building around the tools they already had. The flat platform fees only compete at two to three times their install count.

The reps also already sold in a proposal tool they liked. The owners' theory of the business was that reps stay because selling there is easy. Ripping that out to fit a platform would have broken the one thing that worked.

So the plan was to keep the proposal tool, add a real CRM beside it, and connect both to the ops board the install team already used.

Leak 1: service calls landing on personal phones

Service calls came in almost every day. They rang whichever rep a homeowner had a number for. No record, no routing, no way to know if anyone called back.

We gave the company one main line with a call tree behind it. A recorded human greeting comes first, then a short menu. Press 1 for a new quote and a person's phone rings, at any hour. Billing questions and messages go to voicemail with a call-back task and a text alert, so nothing sits.

Press 2 for service and an AI agent answers. This was the part the owners pushed back on. They didn't want AI talking to their customers, and that was a fair position. So we wrote the agent a short job description:

  • Take the caller's name, address, the issue and the system details.
  • Create the service job and alert the team.
  • Don't quote prices, talk savings or discuss incentives.
  • Hand anything safety-related to a person.
  • Disclose the recording, since some of their service area requires every party's consent.

The narrow scope is what got it approved. We tested it on live calls before launch, and testing caught a gap where a declined call counted as "answered" and dropped the caller. We fixed it and confirmed with two more live calls. At their volume, we estimate the agent costs $15 to $40 a month.

Every service job it creates also lands on the ops board automatically, linked both ways so the two systems never disagree.

Leak 2: finished work that never gets billed

Uninvoiced service work is the quietest leak in home services. The tech finishes, the customer is happy, and the job just stops moving.

We designed the service pipeline around the stage most setups skip: Invoiced. A job moves from new request to triaged, scheduled, in progress and completed, and then it has to reach Invoiced before Paid. A job that sits at completed without an invoice triggers an alert to the person who owns billing.

The last step, sending the invoice automatically from the accounting system, is next on the list.

Leak 3: upsells that live in one person's head

A service visit is the best sales appointment a solar company gets. The tech is already on the roof, already trusted, and often looking at a system that would benefit from a battery or an electrical upgrade.

At this company, those opportunities depended on one owner remembering them. Now the AI agent asks about upsell signals during intake, and an upsell flag on any service job opens a sales opportunity on its own, assigned to a rep.

The leaks nobody had named

Looking at how a job actually moves turned up leaks the owners hadn't listed, because they didn't look like leaks.

An adder priced at a fifth of market. We benchmarked every adder on their pricing sheet against three comparable installers. One roof-type adder came in at about one-fifth of what the others charged. Three other items we first flagged as underpriced matched the closest comparable exactly, so we withdrew those. A benchmark is only useful if you report what didn't hold up too.

A typo that cost 10 days. The lender behind many of their financed deals rejects a design package over one or two wrong characters in an equipment model number, and each rejection restarts a 10-day review. We rebuilt the equipment catalog in the proposal tool with exact model numbers, so the mistake can't be made at the point of sale.

A tax credit that had already ended. The website and the proposal template both still advertised the homeowner solar credit that expired Dec 31, 2025. That's a compliance problem dressed up as marketing. We removed it from the site and flagged the template for a rewrite.

Pricing reps had to do math to quote. Travel was priced per mile and small systems per panel. We replaced both with flat bands: 5 travel rings by distance from the nearest of two crews, and 5 size bands for small systems. Reps type an address into a map and see the ring and the adder.

What the CRM is for, once the leaks are closed

None of this required a big software decision up front. It required knowing where the money fell out, then putting each handoff somewhere it couldn't get lost.

That's what a CRM for solar installers should do. It holds the handoffs: the call that becomes a service job, the completed job that has to become an invoice, the service visit that becomes a sales opportunity, the signed contract that becomes an install. In this case, a signed contract in the proposal tool now creates the job on the ops board with the signed PDF attached, and a nightly check catches anything that slipped through. Nobody retypes it.

Built in that order, the CRM shows up already doing a job. Bought first, it's one more place the work can stall.

What we can show so far

Ten weeks in, we can show what shipped. We can also show one early measurement: about a week after the new website launched, the company was named in 50% of the AI answers to 19 local homeowner questions, up from 36% on the old site. Google Maps rank hadn't moved yet.

What we can't show yet is dollars. The monthly report covering leads, close rate, contracts, revenue funded and cost per sale needs a full quarter of data first. We'll post the numbers when they're real. The full engagement, system by system, is in the case study.

The owners knew where the cash was going on day one. The work was making sure it stops.

Want these systems built for you?

Tell us how a job moves through your company today, from the first knock to the review. We'll tell you which of the five systems to build first.