Your Software Stack Is Only as Good as Your Worst User

Four months into running Reeves Electric, I had Jobber set up correctly, CallRail running on every marketing channel, and a tech team that was essentially using a $200/month field service platform to send text messages. Jobs closed on paper. Notes fields blank. Customer signatures missing. Status updates untouched from "dispatched" to "complete" with nothing in between.

The software wasn't the problem. I was.

You Bought the Right Stack. You Just Never Finished the Job.

When I launched in February 2022, I spent real time on the technology decisions. Jobber at the core, CallRail for attribution, QuickBooks Online for accounting. I'd watched enough shops run on spreadsheets and gut feel to know I didn't want that. So I picked tools with intention, got the accounts set up, did the vendor onboarding call, and went back to running jobs.

What I didn't do: write down what I actually expected my techs to do inside the software. Not in any detail. I showed them the app, walked through the dispatch flow once, assumed they'd figure out the rest. For the first four months they were using maybe a third of the actual feature set. The platform was sending me organized, incomplete data.

Same thing happened with my intake process. My NPS in month two was a 4. The problem wasn't the phone system. Nobody had written down what "correct" looked like on a call. No script, no checklist, no defined outcome. My CSR was improvising every conversation because I'd never told her what a good call sounded like. Once I defined it and started reviewing recordings against it, NPS hit 81 by month nine.

Same problem. Different department.

Most Shops Are Paying for a Tool Their Techs Use Like a Notepad

Here's what I've seen across my own records and what guys I've talked to describe consistently. Job gets dispatched through Jobber. Tech acknowledges it in the app, drives to the job, does the work, and then — nothing. He closes it out that night from the truck or the next morning from his driveway. Notes say "fixed it." No before photo. No after photo. No customer signature. No follow-up items flagged. Status sat at "in progress" for six hours while your dispatcher had no visibility into whether the tech was still on site or already at lunch.

This happens because nobody defined what done looks like. Not in writing. Not with specificity.

Every week at Reeves Electric, my dispatcher and I pull four call recordings — two that booked, two that didn't. We tag what went well and what went wrong. It takes about 45 minutes. You can run the exact same discipline on job records. Pull four completed work orders from last week. Open each one and ask: does this record tell the full story of what happened on that job? Could I hand this to my bookkeeper, my attorney, or a new tech covering a callback and have them understand what was done?

If the answer is no, you have a workflow problem. Possibly a training problem too. But start with workflow.

Here's what makes incomplete records expensive beyond the obvious administrative headache. If my techs aren't updating job status in real time, the downstream data falls apart. Where did this job come from? How long did it actually take versus what we quoted? What did we bill versus estimate? That data drives my marketing decisions and my labor planning. If it's garbage going in, I'm flying blind — which is the exact thing I bought the software to avoid.

You trained your techs on the buttons before you wrote down what the buttons were supposed to accomplish. That's the sequencing problem. It's not a training problem yet.

Write the Workflow Before You Train Anyone on the Software

Most shops solve this in the wrong order.

They train the tech on the software first. Vendor onboarding, maybe a YouTube tutorial, maybe a 30-minute walkthrough. The tech learns how to click the right buttons. Then goes into the field and uses those buttons in whatever sequence feels natural, because he has no idea what the business outcome of each step is supposed to be.

Reverse it. Write the workflow first. Train to the workflow. Then the software is just the tool that executes it — and the training sticks because there's context behind every button.

I've said this about intake a dozen times: if you haven't written a document that defines what a great call looks like from first ring to booked job, your CSR is freestyling. If you haven't written a document that says "here is what a completed job record looks like at dispatch, at arrival, at job close, and at invoice" — your tech is freestyling. Techs freestyling in your field service software costs you money in ways you won't see for weeks.

The CallRail example is the cleanest analogy I have. The tool is $50 to $150 a month. The AI call scoring upgrade they'll pitch you is another $150. I tried it. Wasn't worth it for us, at least not yet. What made CallRail worth every dollar was the discipline around it — the weekly review session, the tags, the pattern recognition over time. A $50 tool used with discipline beats a $300 tool used passively. That's true whether you're talking about CallRail, Jobber, or the notepad your tech currently prefers.

The One-Page Field Workflow: What It Actually Contains

At five trucks, I can't be in the field checking work every day. The written workflow is what lets me trust the output without standing over anyone's shoulder.

Here's what ours actually requires at each stage.

Dispatched: Tech confirms the appointment in the app. Customer gets an automated text with ETA. If the tech is running more than 15 minutes late, he updates the status and the customer gets another text. Two taps.

On-site: Before photo of the work area. If it's a service call with an unclear scope, a photo of the panel, the circuit in question, or whatever the customer described. Takes 45 seconds.

Diagnosis complete: Note entered in the job record describing what was found versus what was quoted. Not "checked breaker." Something like: "Customer reported tripping AFCI in master bedroom circuit. Found improper double-tap at breaker, aluminum wiring on two conductors not originally flagged in quote. See photos." If the scope changed, this is where the change order gets written before any additional work starts.

Job close: After photo. Customer signature captured in the app. Follow-up items flagged in a dedicated field — panel issues spotted, code items not in original scope, any conversation about an EV charger or panel upgrade. A blank follow-up field at close is a lost upsell conversation and a liability if the customer calls back claiming we caused a new problem.

Invoice: Sent from the app before the tech leaves the driveway. Not that night. Not the next morning. Before he leaves.

One page. Five stages. Specific outputs at each one. No tech needs to call the office to know if they did it right.

What Happens When You Skip This (And How I Found Out)

In the fall of 2022, one of my techs closed a panel repair job in the wrong status — marked it "ready to invoice" instead of triggering the actual invoice send. The work was good, the customer was happy, but the invoice sat in a queue I wasn't monitoring closely yet. The customer never got a bill.

I caught it three weeks later during my monthly P&L close. Went looking, found the record, sent the invoice late with an apology, customer paid — but the receivable had been sitting there 21 days for no reason other than a status click that never happened.

Shops that don't run a weekly revenue review would never have seen that until quarter-end reconciliation. I caught it because I had the cadence. But it shouldn't have happened in the first place, and it wouldn't have if the workflow had a defined "invoice sent" step the tech was responsible for completing before the job closed.

Priya follows written protocols for medication and procedure every single day. Not because she's incompetent — because the protocol is what makes the outcome consistent no matter how busy the floor is or how many hours into her shift she is. A good nurse following a protocol is more reliable than a great nurse working from memory.

I was doing the same thing to my techs. No protocol. Then wondering why the results varied.

What to Do Monday Morning Before You Buy Another Tool

Block 45 minutes Monday morning. Pull four completed job records from last week. Not your best ones — random ones. Grade each record against what complete should look like. Before photo, diagnosis note with found-versus-quoted language, customer signature, follow-up items flagged, invoice sent before the tech left the property. You'll find out quickly where the gaps are.

Pick one problem. Write one fix. Don't try to repair photos, notes, and the invoicing flow in the same month. It won't stick. Find the single most common incomplete field across those four records and write the workflow fix for that problem only. One paragraph per stage, specific expected output, no ambiguity.

Train to the workflow, not to the software. Call a 20-minute shop meeting. Walk through the one-page document. Show what a complete record looks like in the app. Set a 30-day window where you're reviewing this one thing specifically.

At day 30, pull four more records. If it improved, move to the next gap. If it didn't, you're probably having a different conversation — about one specific person, not the whole team.

When I built out admin support at Reeves Electric, I hired the bookkeeper before the CSR because the workflow was more contained and I could train to it cleanly. Same logic here. Fix one discrete workflow before you touch the next one. Don't rebuild your entire Jobber setup and your dispatch process and your invoicing flow in the same week.

Write the document first.

---

FAQ

My techs say the software slows them down on the job — is that a training problem or a workflow problem?

Usually both, but start with workflow. If the required steps aren't clearly defined, techs fill that ambiguity with friction — they're making micro-decisions on every job about what to enter and when. That feels slow. A clear, written sequence reduces the decision load because there's nothing to figure out. Once the workflow is written, do the training. If it still feels slow after that, look at whether you've required steps the software could automate — like the customer notification on dispatch — instead of requiring the tech to do manually.

We're on Housecall Pro and thinking about switching to Jobber. Should we fix our workflows first or migrate and start fresh?

Fix the workflows first. If you migrate without written workflows, you'll rebuild the same habits in the new interface within 60 days — and you'll be dealing with migration headaches at the same time. Get your one-page field workflow running cleanly in Housecall Pro, run it for 30 days, verify it's working, then evaluate whether the platform switch is still necessary. You may find the problem was never the platform.

How do I hold techs accountable for incomplete job records without it becoming a disciplinary thing every week?

If you've never written down what a complete record looks like and you're grading techs against an unwritten standard, that's a management problem. Write the standard, share it, train to it, give it 30 days. After that, the weekly record review becomes a coaching conversation against a known benchmark. Most techs who are consistently incomplete after 30 days either don't understand the workflow — which you can fix — or they've made a choice. That second conversation is shorter than you think it needs to be.

What does a realistic software training cadence look like for a crew that's in the field five days a week?

Twenty minutes to introduce one workflow element beats a two-hour quarterly session. Once the one-page workflow is live, the ongoing cadence I'd suggest is a five-minute check-in at the Monday morning truck huddle — one thing that went well in the records last week, one thing that needs to be tighter. Specific. Short. Review four records weekly yourself; surface the pattern monthly with the team.

Is there a point where adding an integration actually makes adoption worse?

Yes, and it happens faster than you'd expect. Every integration adds a dependency on the data underneath it. If the underlying job record is incomplete because the tech didn't fill the notes field, your Zapier-to-NiceJob review request fires on a job the customer barely remembers, at the wrong time, with no personalization. Automated on top of incomplete is still incomplete — it's just faster. Clean the core data first, then automate on top of it.

How do I know if the problem is the software or just one tech who won't use it?

Pull records across the whole team and sort by tech. If four out of five have complete records and one consistently doesn't, you know what you're dealing with. If all five have the same gaps in the same fields, the standard was never defined clearly enough for anyone to hit it. The weekly record review makes this obvious within a month. One outlier after a clear standard has been set and trained is a different conversation than a team-wide gap that never had a standard to begin with.