Getting the Hour Onto the Job
The mechanics of attribution: when hours get booked, by whom, and the four conditions that decide whether the result means anything.
Where one week's hours went in a team that thought it booked everything
Accurate within the grain.
Reconstructed; rounded to halves and wholes.
Estimated by a person who was not there.
The catch-all; nobody chose it deliberately.
Appeared only in the payroll reconciliation.
Less than half the week was booked at the time it happened. Everything else is a reconstruction of some kind, and the job costs built on it carry that. One employer's own week, not a benchmark.
Getting an hour onto a job is a mechanical act, and its accuracy is decided by four conditions: how soon after the work it happens, how long it takes, how short the list is, and whether the person booking it was there.
The record examined in “Getting the Hour Onto the Job” is strongest when it is created close to the work and remains open to correction. A team evaluating consult the current feature overview for managing freelance project time can connect time with projects and work categories, provided people can review entries before they influence payroll, billing or cost reports.
None of those is about diligence. All four are design decisions somebody made, usually by default.
As an outside benchmark for “Getting the Hour Onto the Job”, see Atlassian project-management guide. Keep the source record, calculation rule, approval and resulting decision visible as separate steps.
Soon
Accuracy falls sharply after about a day. Same-day booking is substantially better than next-day and in a different class from Friday.
The single largest available improvement in most operations is moving entry from weekly to daily, and it is a habit change rather than a system change.
Quick
Under two minutes a day or it will not happen reliably. Every extra required field costs compliance.
A system that demands a narrative per entry gets narratives that are copied from last week.
Short
A list the person can hold in mind. Covered in its own note; the mechanical point is that list length and accuracy trade off directly.
Present
Hours booked by somebody who was not there are an estimate. Supervisors booking on behalf of a team is common in field work and it should be labelled as what it is.
It is not wrong; it is a different quality of data and the job cost inherits that quality.
The default code
Every system has one and it should be examined rather than tolerated. Hours that default are hours nobody attributed.
Making the default code blank — forcing a choice — moves some of that into real codes and some of it into non-compliance, and both are more informative than a silent default.
Booking in advance
Planned hours entered before the work. Useful for scheduling, dangerous for costing, because the plan then becomes the record unless somebody corrects it.
Where a system pre-populates from a plan, the costing is measuring the plan.
Approval
A supervisor checking a week before it posts. Catches gross errors and introduces a second rounding, and it delays the data by days.
Worth it where the data goes to a customer; questionable where it does not.
Correcting after posting
Should be possible and should leave a trace. Jobs get re-costed, errors get found, and a system that forbids correction gets corrected in a spreadsheet instead.
What matters is that the correction is visible, not that it is prevented.
Mobile and field entry
Where the work happens away from a desk, entry has to happen there or it happens on Friday. This is the single biggest determinant of data quality in field operations.
A good mobile entry path is worth more than any amount of reporting built on bad data.
What good looks like
Same day, under two minutes, from a list of twenty or fewer, by the person who did the work, with a kind-of-work field and a visible daily total.
Six properties. Most operations have two or three and wonder why their job costs are disputed.
What to do about the Friday problem
Where entry has always been weekly, moving to daily is a change of habit rather than of system, and it fails if it is announced rather than supported.
What works is reducing the entry to under two minutes first, then moving the timing. Asking people to do a twenty-minute task daily instead of weekly is asking for five times the effort, and they will correctly decline.
Booking for somebody else
Legitimate in field work and in operations where the people doing the work are not at screens. It should be recorded as third-party entry so that the data's quality is known.
A job cost assembled from supervisor estimates is a different object from one assembled from first-hand entries, and both are useful as long as nobody confuses them.
The entry that is never corrected
Most systems allow correction and most people never use it, because the correction path is harder to find than the entry path.
Where a week's data is wrong and cannot easily be fixed, people stop trying and the quality settles at whatever the first attempt produced. A visible, quick correction route is worth as much as a quick entry route and gets a fraction of the attention.
Measuring the booking itself
Share booked same day, share defaulted, share booked by a third party, and the payroll residue. Four percentages, published monthly.
They describe the quality of every job cost downstream, and they are the only honest answer to "how reliable is this number".