Most companies look at yesterday's board only when a customer complains. By then the cause is buried: was the tech late because the 10 a.m. job ran long, because the drive took twice the estimate, or because the first job started 25 minutes late and nothing recovered?
A plan-vs-actual review answers that every morning, for every tech, in about 15 minutes. It lines up what the board said would happen with what did, finds where each late arrival started, and gives you one fix to make today. Over a few weeks it also tells you whether your duration estimates and drive assumptions are honest.
The three numbers per stop
For every job on yesterday's board, you need three planned values and three actual ones:
| Planned | Actual | |
|---|---|---|
| Drive to the job | The drive the schedule assumed | Departure to arrival |
| Arrival time | The planned start | The "arrived" tap |
| Time on site | The job's planned duration | Arrived to done |
The difference between each pair is the variance. Positive means later or longer than planned.
Why late arrivals are a sum
The arrival variance at any stop is the sum of everything that went wrong, or right, before it:
arrival variance at stop k = start variance + Σ drive variance (stops 1..k) + Σ duration variance (stops 1..k−1)
Start variance is how late the first drive began. Every overrun before stop k carries forward, unless a gap or a short job absorbs it. That's why a single long job at 9 a.m. can make the 3 p.m. customer late, and why looking only at the late stop tells you almost nothing about the cause.
A worked example
One tech's day, illustrative, numbers invented. All values in minutes. Positive means late or long.
| Stop | Drive variance | Duration variance (this job) | Arrival variance |
|---|---|---|---|
| 1 | +5 (start ran late: +15) | +35 | +20 |
| 2 | +2 | −10 | +57 |
| 3 | +18 | 0 | +65 |
| 4 | −3 | +62 |
Read the table from the top, not the bottom. Stop 4 was the complaint, but the day was lost at stop 1: a late start and a job that ran 35 minutes over. The 18-minute drive overrun to stop 3 made it worse. The short job at stop 2 helped a little.
The fix list for this tech writes itself: why did the first drive start 15 minutes late, and is the planned duration for stop 1's job type too short?
The 15-minute routine
Run this before the morning rush, on yesterday's board.
Minutes 0 to 3: find the late stops
List every arrival that missed its window. If you use a threshold, such as more than 10 minutes past the planned start, keep it the same every day so the trend means something. How to measure on-time arrival covers how to pick it.
Minutes 3 to 10: trace each one back
For each late stop, walk back through the tech's day and sort the cause into one bucket:
- Late start. The first drive left late. Usually a shop, loading or morning-meeting problem.
- Duration. An earlier job ran long. Usually an estimate problem for that job type.
- Drive. A leg took much longer than planned. Usually time of day, or a planned drive measured from the wrong place.
- Insertion. An emergency or add-on was put in front of the stop.
- No recovery. The variances were small, but there was no gap anywhere to absorb them.
One cause per late stop, the biggest contributor. Don't debate ties.
Minutes 10 to 13: count the buckets
Across all techs, how many late stops fell in each bucket? One day is noise. A week shows the pattern.
Minutes 13 to 15: pick one fix
Choose the single change that would have prevented the most late stops, and make it today. Examples:
- Duration bucket keeps winning for one job type: raise that type's planned length to its median from your own records. Job duration estimates covers how.
- Drive bucket keeps winning in the afternoon: plan afternoon drives with afternoon traffic, and stop clustering cross-town jobs after 3 p.m. Plan a dispatch day around drive time has the method.
- Late starts keep winning: look at the morning, not the route.
What to track weekly
Once a week, turn the daily notes into a few numbers:
- On-time arrival rate, by tech and overall.
- Late stops by cause, as counts. Watch which bucket shrinks after a fix.
- Median duration variance by job type. If a job type is consistently long, its estimate is wrong.
- Median drive variance by time of day. If afternoon drives always run long, your planned drives are too optimistic for that hour.
These feed schedule adherence, but the cause counts are more useful than the headline rate. A rate tells you how often you missed. Causes tell you why.
Common traps
- Blaming the tech. A tech who is late at stop 4 is often a tech who inherited stop 1's overrun. Trace before you judge.
- Using planned drive measured from the shop. If the schedule assumed every drive starts at the office, every drive variance will look huge and mean nothing. Planned drive has to be measured from the previous job.
- Fixing everything at once. Change one thing per day and watch its bucket. If you change five, you won't know which worked.
- Reviewing only bad days. Good days show you which job types and areas run to plan. That's where your estimates are already right.
Where CrewLink fits
Today, CrewLink records the actuals this review needs: the tech app captures on my way, arrived and done, time on site is recorded when a job is marked done and future bookings use it, and the Reports page shows on-time rate with CSV exports. A plan-vs-actual ledger that lines up each day's plan with what happened (drive times, arrival times and time on site) is in build, along with a real-road travel model calibrated with live traffic samples. Until it ships, you can run this routine from the exports.
What to take away
- Compare three numbers per stop: drive, arrival and time on site, planned versus actual.
- Arrival variance is a sum of the start, every earlier drive and every earlier overrun. Trace late stops back to where the day slipped.
- Sort each late stop into one cause: late start, duration, drive, insertion or no recovery.
- Make one fix a day and watch its bucket shrink over a week.
- Planned drives must be measured from the previous job, or the drive variance is meaningless.
Common questions
What is plan vs actual in field service?
It is a comparison of the schedule as planned with what actually happened: planned versus real drive times, arrival times and time on site for each job. The gaps, called variances, show where the day drifted and why.
How do I calculate schedule variance for a technician?
For each stop, subtract the planned arrival from the actual arrival. To find the cause, add up the start delay, the drive variances for every leg up to that stop, and the duration variances for every earlier job. The biggest term is usually the cause.
How often should we review plan vs actual?
Daily, briefly, on the previous day's board, so causes are still fresh. Then roll the daily cause counts into a weekly summary to see patterns and check whether your fixes worked.
Does CrewLink have a plan-vs-actual report?
A plan-vs-actual ledger is in build. Today CrewLink records arrived and done taps and time on site, and its Reports page includes on-time rate with CSV exports you can use for this review.