Writing
The camp and the spreadsheet
I did field sampling out of a remote camp in Nunavut. I now write estimates for transmission line work. The distance between those two jobs is where most companies quietly lose money, and it is not a software problem.
How data left the camp
On a fly-in exploration project, nothing leaves except by air. That includes information.
We sampled, we logged, we recorded. Somebody wrote it down — on paper, outdoors, in weather, wearing gloves. The paper went into a folder and the folder waited for the next plane. Then it waited for somebody in an office to type it up.
Nobody was careless about it. The work was good. But between a person describing what was actually in front of them and that description becoming a number somebody made a decision with, there were four or five pairs of hands and often a couple of weeks. Every one of those steps was a chance for the record to get thinner.
What the office saw
Cleaned up. Summarised. Late.
That is the whole problem in three words, and I did not properly understand it until I started sitting on the other side of it.
Estimating transmission line work means building a number out of other people's records. Production rates, crew sizes, equipment hours, what the ground actually did last time against what the plan said it would do. When those records are good, an estimate is arithmetic and judgement. When they are thin, it is judgement wearing a costume.
And the records are usually thin. Not wrong, exactly. Smoothed. Somebody reconstructed the week on Friday afternoon from memory and a notebook, and everyone downstream treats that reconstruction as measurement.
Why the two sides drift
It is tempting to explain this as carelessness in the field or impatience in the office. It is neither. The two groups are measured on different things.
A superintendent is judged on whether the work got done. Paperwork is the tax he pays at the end of a day that already ran long. An accountant is judged on whether the numbers reconcile, and has no way to know which of them were measured and which were remembered. Both behave sensibly. The drift is structural.
Which is why buying software rarely fixes it. Most systems sold into this gap add a step to the end of the superintendent's day and produce a nicer-looking report for the office. The report was never the problem. The distance between the work and the record is the problem, and a prettier report does nothing to it.
What actually closes it
Two things, and only one of them is technical.
The technical one: capture has to happen where the work happens, on the day, by the person who did it. Not reconstructed later, not re-typed by somebody in an office from a photograph of a notebook. Every hand the record passes through costs you accuracy, and the cost is invisible because nobody logs what got lost.
The other is harder. The record has to be worth something to the person filling it in. If a foreman gets nothing back from the system he feeds, he will feed it the minimum that stops anyone complaining. That is not dishonesty. It is a rational response to a one-way arrangement. Give him back something he actually wants — his crew's production rate, which machine is costing him a shift a week — and the quality of what goes in changes without anyone being asked to care more.
What I do differently, and what I do not
I test systems with the person doing the work, not the person most enthusiastic about technology. That is the whole method and it is less impressive than it sounds. It mostly means sitting with a foreman, watching him try to use the thing, and removing whatever made him hesitate.
What I do not do is promise that better software fixes an organisational problem. Sometimes the gap between field and office is not a data-capture failure at all. Sometimes nobody has decided who owns a number. No system makes that decision for you, and anyone who says otherwise is selling you a report.
I still think about that folder, waiting for the plane.