Evidence
What a season of your own data will tell you.
A real register from a real diagnostic, with the customer anonymized. Eleven operating companies, 93 branches, $393.7M of contract value and 47,097 crew-weeks of production data, read across 24 separate systems. Every item came out of software the group already owned.
Ordered by size, not by ease.
Consulting decks lead with whatever is easiest to fix. This register does the opposite. Items are ordered by what they are worth, and each carries a confidence rating that separates something we measured from something we are still asking about.
Several items are open questions rather than findings. They are left in, labeled, because a register that only contains good news is not a register. One item exists purely to say that a capability cannot be delivered to most of the group until somebody changes a setting in the system they already run.
Renewal retention varies by 35 points between branches
Retention ranges from 65% to 100% across the branches with decided renewals, on the same contract book and broadly the same customer type. The gap between the weakest branches and the group median is the largest single number in the study, and it is visible by region as well as by branch.
measuredinvestigationWork delivered below its own recorded cost
$19.8M of work was delivered below cost on $391.3M of priced revenue. It is overwhelmingly a single-region problem, and that region is simultaneously the best run in the group on routing and on retention. The loss sits in a few dozen large tickets rather than a long tail.
measuredinvestigationA fifth of paid labor never reaches a work ticket
Crews clocked 3.24M hours. Only 2.62M reached a job ticket. Job costing therefore sees about 81% of what is paid, and the missing share is not evenly spread between branches.
measuredinvestigationThe price book has stopped being a control
Between 57% and 89% of services carry an overridden price, and minimum-price rules are effectively unused. At the top of that range, nearly nine services in ten depart from the book.
measuredinvestigationOvertime is material and recorded inconsistently
585,829 overtime hours are recorded across the group, while one operating company records almost none despite running the longest shifts in the group. Its job costs therefore exclude premium pay that its peers include, which flatters its whole region.
measuredlight buildThe route optimizer is licensed, and used on a minority of days
The optimizer in the incumbent platform demonstrably works. A fifth of crew-days are exactly optimally sequenced, which cannot happen by chance. It is simply not applied consistently, and never on the largest days.
measuredprocess onlyEstimating accuracy varies by 27 points between operating companies
Actual hours run at 120% of estimate in one operating company and 93% in another, on similar work. For the worst performer this is roughly five times larger than its entire scheduling opportunity.
measuredinvestigationProperty map data exists in five operating companies, not across the group
The platform holds 18,943 property map attachments, but only five operating companies have a map attachment type configured at all. The rest return zero because the type does not exist in their configuration, so there is nothing to upload into. A configuration gap, not an adoption gap.
measuredinvestigationReview is the only real control, and it is barely used
Completed tickets are reviewed between 4% and 61% of the time depending on the operating company, while approval sits at 99.8% almost everywhere. Approval is a formality. Review is the only step that checks work before it bills.
open questioninvestigationUnbooked time at the start of the shift
Crews are clocked on for 3.4% to 7.9% of the shift before any work is booked. It is the largest unbooked block of the day in almost every operating company, and it is a morning-start question rather than an end-of-day one.
measuredinvestigationCustomer, account-manager, property, route and crew-leader identities are replaced with stable pseudonyms or removed throughout. Operating company names are not shown. Every figure is unchanged. Reproduced with the client's permission.
What we would move, and how you would check
The register tells you what is wrong. This is what we would aim at, in numbers a branch manager already tracks rather than in features. The starting points below are typical for a mid-sized operator, and get replaced with your own last twelve months before anyone commits to a target.
| What moves | How | Today | Target | Measured by |
|---|
Every one of these can be measured from the first day, rather than after some later phase of work. That is the only reason the targets are worth printing at all.
Only one item in the register was both measured and fixable by process alone. The rest needed something built, or someone to go and ask a branch manager a question.
Why we lead with the diagnostic
An operator who has been sold software before is right to be skeptical of a demo. The register is the antidote. It is their data, their totals, reconciled to their financials, and they can check any line of it without us in the room.
It also settles the sequencing argument before it starts. In this group, scheduling was the most precisely measured opportunity and nowhere near the largest. Retention losses and uncaptured labor were each several times bigger. Without the register, the schedule is what everybody would have worked on, because it is the thing you can see from the truck.