01 / Proof

Real numbers, labeled by how much they are worth.

This is a young firm with a small number of engagements, and pretending otherwise would be the fastest way to lose the kind of buyer this is written for. So every claim below carries the class of evidence behind it, and the weak ones say so themselves.

No client is named. Written permission to use a name is not on file for anyone, so you get the shape of the business, the systems, and the arithmetic.

Deployed: Built, live, and running in a client's business today.
Diagnostic: Measured inside a real business. The findings are real; the systems were not built, so no result is claimed.
Surge internal: Surge's own systems, running Surge. Not a client engagement.

DeployedFollow-up and response

The estimates were written. Nobody was chasing them.

a $5M+ roofing company in Arizona

What was true

The company was writing more estimates than it ever had and closing a smaller share of them. Chasing lived in people's heads, and heads get busy in storm season. A quote went out, went quiet, and became somebody else's roof.

We pulled every open estimate out of the CRM and counted them by hand against the follow-up tasks attached to each one.

What we built

A follow-up system inside the CRM the company already ran. Every estimate that goes quiet gets worked on a schedule the owner approved, in the company's own voice. Nothing new to buy, no new logins, nothing for the office to learn.

Alongside it: every new lead gets an answer in seconds instead of whenever somebody got off a roof, and the review request fires on its own when a residential job closes.

What changed

The pile has a number, an age, and an owner now. It is worked every day by software instead of remembered by a person who is on a roof.

Reviews went from 138 to 159 over the engagement and the 4.9 rating held the whole way.

Read this with it: That figure is the size of the pile we found, counted once from the company's own export. It is not revenue recovered. Recovered dollars get published when they are measured against the baseline we froze, and not before. Worth saying plainly: the owner and one partner personally sourced 71% of that month's pipeline, so any future claim about signed work has to account for them and not for us.

DiagnosticReceivables

The work was finished. The money had not shown up.

a Colorado restoration company

What was true

Restoration work gets done fast and paid slow. Insurance-funded jobs stack up in receivables, and chasing them is the task that loses to every emergency, every day, forever.

We read the whole job history through the CRM's API rather than sampling it, so the number is a census and not an estimate.

What we built

A full map of the pile: every stuck invoice, its age, its stage, and what it would take to move it. Plus the follow-up design that would work it without anyone having to remember.

What changed

The company could see the whole receivable position in one place for the first time, aged and sorted, instead of rediscovering old invoices by accident.

The engagement paused before the follow-up went live, so there is no recovery number here. I publish those when they are real.

Read this with it: Diagnostic only. Nothing was deployed here, so nothing is claimed beyond what the audit found.

DeployedReporting and visibility

Monday's numbers used to cost half a day.

a $5M+ roofing company in Arizona

What was true

The owner's weekly picture lived in five systems. Getting one honest view meant somebody copied figures into a spreadsheet on a weekend, and by the time it was done it was describing last week.

What we built

One page that pulls from all five systems and refreshes every morning before the owner is awake: leads, sold work, pipeline, and margin by branch. Nobody compiles anything.

Six city pages and the local search setup underneath them, so the phone rings from the towns the crews actually drive to.

What changed

The owner opens one page and knows how the company is doing. The half-day of copying is gone and the numbers stopped being a week old.

05 / One we turned off

A text that was technically correct and practically useless.

We built a message that told the homeowner their crew was starting that day. It ran for sixteen days and we killed it.

Crews arrive at five or six in the morning. The message landed around nine, by which time the crew had been on the roof for three hours. So it did not read as helpful, it read as a system that did not know what was happening at the house. Being accurate was not the same as being useful.

It came out, the owner was told why in writing, and the acceptance test we run before anything goes live now asks the question that would have caught it: not whether the message is true, but whether the customer is glad it arrived.

Every engagement produces at least one of these. Publishing them is the only way a prospect can tell the difference between a company that tests its work and a company that has never looked.

Surge internalOur own instruments

The systems we run on, and what they cost us to admit.

These are Surge's own instruments, not client work, and they are labeled that way so nobody mistakes one for the other. They are the reason a small firm can hold a standard a larger one usually cannot be bothered to write down.

The implementation system: Nine phases across four arcs: decide, build, prove, operate. A build is meant to go from kickoff to accepted in 21 days, and the client's own time inside that is measured in hours, not weeks.
The acceptance test: Before we invoice, we put five real drafts in front of you and ask one question about each: would you send this as written, yes or no. Three yes is a pass. Fewer and we tune for another week or tell you honestly that it should be scoped down. We publish the test because being willing to fail it in front of you is worth more than any feature.
The pattern library: Twenty-one documented patterns from real engagements. Two of them are graded as measured; the rest are observed or inferred, and they are labeled that way internally so nobody quotes a hunch as a finding.
The agent test suite: Ten tests every client-facing agent passes before it is allowed near a customer. Zero fabrications, zero data leaks between clients, zero refusals of questions it should have answered. One fabrication fails the suite no matter how the other ninety-nine went.

Tell us what is not working.

Two minutes to write it down, a reply within one business day, and the option to book a call straight away. You keep whatever I find either way.

Get in Touch