This post in 30 seconds.
- What it is: quality control in field service is an independent check that the work met your standard, done by someone other than the person who did it, on evidence captured at the job. A "Completed" status is not that.
- Why now: a callback costs around $2,500 on average according to ACCA, and a 5% callback rate can cost a business over $100,000 a year. Most of those jobs were "closed" before anyone at head office saw them.
- Who this is for: service managers, heads of field operations and quality leads at equipment makers, dealers, installers, maintenance providers and utilities who sign off work from a desk, not from the site.
The job was closed on Tuesday. The work order says "replaced contactor, tested, customer happy." On Friday the customer rings to say the unit tripped again, the panel was left off, and the technician did not mention anything about a burnt wire.
Nobody at the office saw that job. Not the part, not the panel, not the wire. The first person to check the quality of the work was the customer, and they checked it by having it fail.
If you run a service team, a field operation or a quality function, that is the shape of most quality problems: the work is real, the closeout is a form, and the review happens after the complaint. This post is about running quality control the other way round. Verify the work from the desk before the customer verifies it for you, decide how many jobs to check, and make each finding change something next week.
This is written for the person who signs off field work without standing next to it: service managers, field operations leads and quality managers at manufacturers, dealers, installers, maintenance providers and utilities. It covers what to check, who captures the evidence, how to sample without losing the plot, and how findings feed training and callback prevention. If you would rather build the review loop than read about it, start a free Venta Capture account and put one job type through it this week.
What quality control means when you cannot be on site
Strip it down and field service quality control has three parts. A standard for what "done right" looks like on this job type. Evidence from the job that can be judged against that standard. And a person who is not the technician judging it.
Most operations have the first part written down somewhere. Very few have the second and third for anything other than complaints. What they have instead is a closeout checklist the technician ticks and a status the office trusts.
A checklist the technician grades on their own work is a record of self-assessment, not quality control. It tells you the technician believed the job was complete. It does not tell you the right contactor went in, the guard went back on, or the customer was shown the repair working. Those are the things that come back as callbacks.
And callbacks are where the cost sits. ACCA puts the typical average cost of a callback at around $2,500, with a quick drive-by fix at about $400, and notes that a 5% callback rate can cost a business over $100,000 a year. That is labour, travel, the billable slot you lost, and the customer who now watches everything you do.
A quality assurance inspection done after the fact, by driving to a sample of sites, is the traditional answer. It works, it does not scale, and it lands weeks after the job. The remote version needs the technician to capture the evidence at the job and a reviewer to judge it from the desk the same day or the next.
One boundary before going further. This post is about the review process: who checks, what they check, how many they check, what happens with the result. What a completion capture should contain, and how to hold contractors to it, is covered in proof of work for field service. Read that one for the evidence, this one for the management of it.
Two jobs: capture on site, review at the desk
The mistake in most "photo documentation" advice is treating capture and review as one job done by one person. They are two jobs, done by two different people, at two different times.
Capture is designed by the office and performed by whoever is at the site. Review is performed later, by a qualified person, from a queue. Split them and both get easier.
Capture first. The technician (or the subcontractor, or in some workflows the customer) is the eyes on site, but they are not the judge of what a reviewer needs. "Send us some photos when you're done" gets you a wide shot of a closed cabinet and a thumbs up. So the organisation writes the capture: which points in the job, which angles, which questions. The person on site follows it.
For quality control the capture points that matter are usually:
- On arrival: the asset identifier and the condition as found, before anyone touches it.
- Before the critical step: the old part, the isolation, the reading.
- After the critical step: the new part with its number legible, the guard back, the test running.
- At handover: the area as left and the customer's confirmation, spoken or as a declaration.
That is a guided video capture: the flow decides what is recorded, not the person holding the phone. If your current process depends on the technician remembering what to photograph, the guided photo capture post explains why the memory approach fails on the fifth job of the day.
Now review. The reviewer is your best technician, a supervisor or a quality lead, and their time is the expensive part. So the review must not require them to be on a call, on site, or online at the moment the work finishes. The capture lands in a queue as a case, and the reviewer works the queue when they sit down for it. Send now, capture later, review when ready.
This is the asynchronous point that live remote-assistance tools miss. A live session gives you an expert on the camera for one job. A queue gives you an expert's judgement on sixty jobs in an afternoon, with no calendars aligned.
Venta Capture, a product of VentaVid, is built for this split. The office builds the flow, the person on site opens a secure, personal capture link in the browser (no app, no account), and the submission arrives as a structured case: the video and photos, the answers to each question, a transcript of what was said, the receipt timestamp and the session record. The field service page walks through the flow; the mechanics of review come next.
Why the split matters in numbers: Aquant's 2025 benchmark found that a failed first visit adds two additional visits on average and extends resolution by 14 days. A reviewer who catches the missing test on Tuesday schedules a return on Wednesday. A customer who catches it on Friday starts that 14-day clock.
What to check: the four-point review
Reviewers need a rubric, and the rubric needs to be short. Four checks cover most of what turns into a callback, a dispute or an injury.
The inspection criteria behind each row will vary by trade. An HVAC contractor's "safety steps" row is refrigerant handling and a start-up with readings; a lift engineer's is the interlock test; a fleet workshop's is the torque on the wheel nuts. Write the criteria for your job types, but keep the four rows.
If a case takes a reviewer more than two minutes, the flow is wrong, not the reviewer. Every check should be answered by one or two items in the capture, in order, labelled. A reviewer who has to scrub through a four-minute video looking for the part number will stop reviewing within a week. The fix is upstream: put a "part label, close up" step in the flow with an example photo, and make it required.
The minimum evidence set per job type is therefore a QC decision as much as a documentation one. Fewer, better-placed items beat a long gallery. The photo evidence standards post covers what makes an image usable to a second person; for QC, add "answers one of the four checks" to that list.
Sampling or 100% review: how many jobs to check
The instinct is to review everything. It feels safer, it is what the software vendors sell, and it is usually wrong for two reasons.
The first is human. Juran's inspector-accuracy studies, as summarised by Harish Jose, put 100% human inspection at roughly 87% effective. Reviewers tire, attention drops, and "everything gets reviewed" quietly becomes "everything gets clicked through." The second is cost: a reviewer's afternoon spent on sixty routine filter changes by a fifteen-year veteran is an afternoon not spent on the three jobs that needed it.
So review by risk, not by volume. The matrix below is a starting point; adjust the rates to your callback data.
A 100 percent inspection is for the jobs where a miss is expensive; a sample inspection is for everything else, and the triggers are what make the sample smart. Random alone misses the jobs that go wrong. Triggers alone miss the slow drift in someone who has never had a complaint. You need both.
Two honest notes. Sampling means some poor jobs get through unreviewed, and you should say so out loud when you set the rate. What you gain is a measurable callback rate per technician and per job type, which tells you where to move the rate next quarter. And inspection coverage should be visible: technicians should know that new starters get 100% and veterans get a random sample, because that is fair and they will agree with it.
The arithmetic, with illustrative inputs you should replace with your own. Say 400 closed jobs a month. Fifteen percent sampled plus the triggered pulls is roughly 70 reviews. At two minutes each that is about two and a half hours of a supervisor's month. If one of those reviews catches a missed test before the customer does, ACCA's average callback figure says you have paid for the review time many times over. You will not catch every one. You will catch the ones a random ride-along never would.
If you want to see what that queue looks like with your own job types in it, open a free account and build the flow for one of them. The first standard flow takes about ten minutes.
Four outcomes per case: pass, question, retake, escalate
A review that ends in a note nobody reads is not a review. Each case needs one of four outcomes, and each outcome needs a next step that happens inside the case.
- Pass: the four checks are answered. Set the status, and the evidence stays attached to the job for the day a dispute or a warranty claim needs it.
- Question: one check is unclear. Ask in the case, so the question and the technician's answer live with the evidence rather than in a text thread.
- Retake: one item is missing. Ask for that item only. In Venta Capture a retake is one click, the person on site is asked why they are recording again, and the retake becomes a new case linked to the original.
- Escalate: the evidence raises a real doubt, or the job is one where a second opinion is policy. Escalate to a manager with a reason, and they approve or reject with their name on it.
That last outcome is where the audit trail earns its keep. Every case carries who claimed it, who reviewed it, what they asked, what came back and when. When a customer disputes a job nine months later, you are not reconstructing it from memory.
A word on the signals. Submissions can be scored green, orange or red against a rule set: how many recording attempts were made, whether the file arrived through a fallback route, whether the device has been seen on another submission. Treat every one as a reason to look, never as a verdict. Several recording attempts are normal. A shared van phone will show up on many submissions. And the receipt timestamp proves when your system received the submission, not when the work was done. The evidence integrity layer tells the reviewer what it can and cannot prove; the reviewer still decides.
Turning findings into training and fewer callbacks
Findings are only worth collecting if they change something. The loop that works is weekly, short and boring.
Every week, twenty minutes:
- Count findings by check. Which of the four rows failed most? If "condition left" is 60% of your fails, that is a habit problem, not a skills problem.
- Count by technician and by job type. One person, many fails: coaching. Many people, one job type: the flow or the standard is wrong.
- Pick one training topic. One. Use the actual case in the toolbox talk: a view link opens it without a login, so the supervisor can put the missing test on the screen in front of the crew.
- Change one flow step. If reviewers keep asking for the same missing angle, add it as a required step with an example photo and stop asking.
Callbacks get their own rule. A failed check found by the reviewer before the customer notices is a scheduled return visit, planned, priced and explained. The same fault found by the customer is a callback with a complaint attached. Tag every real callback against the original case and ask which of the four checks would have caught it. After a quarter you will know which check to tighten and which technician needs a week with a mentor.
The gap you are closing is large. Aquant's benchmark puts the first-time fix rate at 86% for top companies against 53% for the bottom group. A meaningful slice of that gap is work that was done, closed and then redone. Review is how you find out which slice.
One observation from watching teams set this up. The first draft of a review rubric always has ten or more checks, because everyone with an opinion adds one. Those rubrics die within a month: reviews take eight minutes, the queue backs up, and the supervisor stops opening it. The rubrics that survive shrink to about four checks with clear evidence behind each. Start with four, and resist the fifth until the callback data asks for it.
Two adjacent points. Contractors go through the same flow and the same review as employees, on their own phone, with no app to install and no account to create; that is the only way the sample stays honest across a mixed workforce. And the same guided capture that checks the work afterwards can look at the fault beforehand: the remote diagnostics and pre-visit assessment posts cover that side, and reducing truck rolls covers what it saves. QC is the back end of the same idea: better information before action, and better information after it.
Rolling it out without a revolt
Technicians do not object to being checked. They object to being checked badly: long forms, vague criteria, monitoring that only exists because of the one colleague who cuts corners. Rollout is mostly about avoiding those three.
- One job type, one reviewer, one month. Pick the job type with the most callbacks. Build the capture flow for it. Name the reviewer and give them the two-minute rubric.
- Publish the four checks. Technicians should be able to recite what is reviewed. No surprises, no secret criteria.
- Keep the capture under three minutes. Five or six steps, each with an example photo and a one-line instruction. If the capture is longer than the closeout it replaces, it will be skipped.
- Show the findings back. The weekly loop is a team meeting item, not a management report. The best technician's capture is the example.
- Review the work, not the person. The question is "did this job meet the standard", not "do we trust Dave." The sampling matrix makes that visible: everyone gets the same rate for the same risk.
On tooling, the setup is small. Venta Capture has a free plan with no credit card, a first standard flow is live in about ten minutes, and a technician can start a capture themselves from the app-free link, or the office can send the link with the job. Flows that branch on answers, route by outcome or push into your ticketing system are set up around your process on a call. The product page covers the inbox, routing, statuses and escalation the reviewer works in.
The evidence QC reviews comes from before and after documentation and technician video updates; keeping it consistent across technicians is the subject of field service documentation.
Frequently asked questions
What is quality control in field service?
It is an independent check that completed work met a defined standard, judged on evidence from the job by someone other than the technician who did it. A completion status or a self-ticked checklist records that the work was declared done; quality control confirms it was done right.
How many completed jobs should we review?
Review by risk, not by a flat percentage. New technicians, new contractors, safety-critical and warranty work get 100%; routine work by proven technicians gets a 10% to 20% random sample; anything with a retake, a complaint or an odd signal gets pulled regardless. Adjust the rates against your callback data each quarter.
Can we run QC on subcontractors who will not install our app?
Yes. A guided capture link opens in the phone's browser with no app and no account, so a subcontractor follows the same steps and lands in the same review queue as an employee. The evidence, the reference number and the receipt time are the same either way.
Do photos and video prove the work was done on that day?
They prove less than people assume, and you should be precise about it. A server-side receipt time proves when your system received the submission, not when the work happened; device and location data help narrow the window. Recording in the session, with a sealed record and integrity signals, makes the evidence far harder to reuse or edit, but a reviewer still judges the content.
Does remote review replace site audits?
It replaces most of them and makes the rest better. Remote review handles the volume: every sampled job, every triggered pull, the same week. Site visits are then reserved for escalated cases and for the periodic check that the capture itself reflects reality.
Will technicians accept being checked?
Most will, if the capture is short, the criteria are published and the findings come back as coaching rather than as a report. Publish the sampling rates so everyone can see that new starters are checked more than veterans for a reason. Use the best technician's capture as the example, not the worst.
What does Venta Capture cost?
Prices are not published. There is a free plan with no credit card required, enough to build one capture flow, send it to your own phone, and run a first month of reviews on one job type.
Check the next twenty jobs
You do not need to review every job. You need to review the right twenty this week, with four checks, from a queue, and do something with what you find.
Pick the job type that produced the most callbacks last quarter. Write the four checks for it. Build the capture flow with one item per check, send it with the next twenty jobs, and have one reviewer work the queue on Thursday afternoon. By the second Thursday you will know which check fails most and which technician needs a conversation, and you will have found it before the customer did.
Venta Capture is free to start, with no credit card, and a first standard flow is live in about ten minutes. Stuck on the flow? Book a free setup call and we build it with you.
The VentaVid team builds Venta Capture for service, claims and inspection teams.
Start for free and build your first review flow
If you would rather see the review queue with your own job types in it first, book a demo and bring last quarter's callback list.