There is a rule that explains most of what goes wrong with construction data: the value of a record decays with the time between the event and the recording. A quantity written down at the pipe is a fact. The same quantity written down on Friday is an estimate. Written down at month end, it is a reconstruction.
Everything in this guide follows from that rule.
The three questions
For every piece of data your company collects, ask three things:
- What decision does this support? If nothing changes based on the answer, stop collecting it. Every unnecessary field is a tax on the person holding the phone, and taxes get evaded.
- Who is closest to the event? The right person to record something is the one who caused it or witnessed it, not the one who will need it later.
- What is the latest moment it is still accurate? For most field data the answer is measured in minutes, not days.
The minimum viable capture set
If you capture nothing else, capture these five. Everything in earned value, job costing, and most compliance is derivable from them.
| Record | Who captures it | When | What it drives |
|---|---|---|---|
| Labor time, by project and activity | Each crew member, or the foreman | At clock in and clock out | Actual cost, payroll, production rates, T&M backup |
| Installed quantity, by pay item | Foreman or crew lead | As work is installed, at minimum daily | Earned value, percent complete, billing |
| Equipment time and assignment | Operator or foreman | At assignment and at day end | Actual cost, utilization, T&M backup |
| Test and inspection records | Whoever witnesses the test | At the test, before backfill | Acceptance, closeout, warranty defense |
| Material deliveries | Whoever receives the truck | On arrival, with a photo | Actual cost, claims, procurement |
Two more earn their place on most jobs: incidents and near misses, captured immediately, and conditions and delays, captured the day they occur, because both are worthless as recollections and valuable as contemporaneous records.
What makes a record defensible
A record that will still be useful in two years, to somebody who was not there, carries five things:
- Time. When the event occurred, not when it was typed. Automatic timestamps beat written ones because they cannot drift.
- Place. Which job, and where on it. A density test without a station is nearly useless.
- Person. Who observed or performed it. Anonymous records invite disputes nobody can resolve.
- Structure. The same fields every time, so records are comparable and searchable. Free text is where data goes to die.
- Evidence. Photos attached to the specific record, not stored in a camera roll.
A photo in a phone gallery is not documentation. A photo attached to a timestamped, located, attributed record is documentation. The difference is retrieval.
The adoption problem, stated honestly
Most field data initiatives fail, and they fail for the same reason. The system is designed by people who need the data, for people who do not. A foreman gets no benefit from the report the office runs. What he gets is another thing to do while the excavator idles.
Three rules that decide whether a rollout survives:
- Every entry must be faster than what it replaces. Not comparable. Faster. A clock-in that takes twenty seconds displaces a paper timecard. One that takes two minutes and asks four questions gets abandoned in week three, and then you have paper timecards plus a software bill.
- Capture must happen where the work is. Any workflow that requires walking to a truck, a trailer, or an office will be batched, and batching is the thing you were trying to eliminate.
- The field has to get something back. Fewer phone calls asking what happened Tuesday. No Friday timecard scramble. Payroll that is right the first time. If the only beneficiary is the office, compliance decays quietly and you will not notice until the data is already fiction.
Cadence: how often is often enough
| Data | Minimum cadence | Why |
|---|---|---|
| Labor and equipment time | Real time, at the event | Memory degrades within hours, and this is most of your cost |
| Installed quantity | Daily | Weekly rollups hide the day that went wrong |
| Tests and inspections | At the test | Backfill makes re-creation impossible |
| Deliveries | On arrival | Discrepancies are only arguable while the truck is there |
| Narrative and conditions | Same day | Late narrative is recollection, which is weak evidence |
Five failure patterns to watch for
- The form that grew. A capture form starts at four fields and reaches fourteen because everyone added "one more thing." Completion rates fall, then quality falls. Audit your forms annually and cut.
- The heroic foreman. One person does all the capture for a crew of twelve. It works until he takes a week off, and then you discover the process was a person.
- The parallel spreadsheet. Somebody in the office keeps the "real" numbers alongside the system. The moment that exists, the system is decorative.
- Silent gaps. A missing record produces a report that looks complete. Systems should flag absence rather than quietly omitting it.
- Data with no consumer. If nobody in the office ever looks at a field the crew fills in, the crew will figure that out and start entering garbage. They are not wrong to.
Where to go next
To see what this data enables, read what is earned value in construction and production rates and how to track them.
PM Axsus is built on the principles above: structured records with timestamps, location, identity, and photos, captured at the point of installation by the person closest to the work. See daily reports, time tracking, and inspection capture.