Reported progress and real progress
The percentage in a weekly report is a judgement made by the person who will be blamed if it is low. Quantities and binary milestones survive that pressure.
The figure in the weekly report is almost always ahead of the site. Everybody who has run a project knows this, and most organisations respond by discounting the number in their heads rather than fixing the number.
The discount is not a fix. It is an agreement to keep producing a document nobody trusts.
A percentage is a judgement, not a measurement#
A quantity is measured. A percentage complete is estimated. Between those two sentences sits every problem this post is about.
To say a slab is sixty per cent done, somebody has to decide what the remaining forty per cent contains, how long it will take, and how the parts compare in effort. That is a forecast wearing the clothes of an observation. It looks like data on the report because it has a number in it.
And it is a forecast made by a particular person: the one who will be questioned if it is low. Nobody in that position is being dishonest. They are being reasonable in the way people are reasonable under pressure — the reinforcement is tied, the shuttering is nearly ready, the pour is arranged for Thursday, so the honest feeling is that this is mostly done.
Why the number only drifts one way#
Optimism in progress reporting is not random noise. It has a direction, and the direction is structural.
The reporter is the reviewed. The person estimating progress is the person whose performance is read off the estimate.
The invisible work is at the end. Testing, finishing, snagging, the services that only reveal their problems at commissioning. Early work is visible and satisfying. Late work is fiddly and easy to underestimate.
Started reads as half done. Once a trade is on site, the item stops being zero. In practice a great deal of started work sits at very little value for a long time.
Nobody is punished for correcting upward. Reporting more this week than last is comfortable. Reporting less requires an explanation, so a figure that should fall is held flat instead.
That last one produces the plateau everybody recognises. An item reaches ninety per cent and stays there for weeks, because it cannot go down and it cannot honestly go up. The plateau is not a measure of the work. It is a measure of what the reporting system will accept.
Measures that cannot be argued with#
The alternative is not a better estimate. It is a different kind of statement.
Physical quantities against the bill of quantities. Cubic metres poured, tonnes of steel fixed, square metres plastered, against the scheduled quantity. This is a count, not an opinion, and the schedule already contains the denominator — the structure is set out in reading a bill of quantities. It also has the advantage of matching the way the job will eventually be billed, so progress and money stop being two unrelated stories.
Binary milestones. Either the slab is cast or it is not. Either the lift is commissioned or it is not. A milestone that can be half achieved is not a milestone, it is a percentage with a name on it. Define milestones so the answer is yes or no, put a date against each, and record the date it actually happened rather than the date it was supposed to.
Value earned, in outline. Take the scheduled value of the work actually completed and compare it with the value scheduled to have been completed by now. That is the whole idea. There is a large formal apparatus around it and most projects do not need the apparatus; they need the comparison. Two numbers, both derived from quantities, neither of them a judgement about effort.
Photographs with a date and a location. Not decoration on the report — a record. A photograph does not claim a percentage, and it cannot be revised next week to suit the meeting. What makes it evidence rather than a picture is covered in a site photograph is a record.
The common property of all four is that they do not require the reporter to predict anything. They ask what happened, and what happened is a fact somebody witnessed.
The chain that turns a number into fiction#
Even honest estimates degrade as they travel.
The engineer rounds up a little, because the last shuttering will be up by Monday. The site in charge consolidates several such figures and rounds the consolidation up, because he does not want to report a fall on a package he knows is moving. The project head reports upward against a target he committed to. The board sees a summary of summaries.
No one lied. Each step applied a small, defensible optimism to a number that already contained one. By the fourth level the figure has no relationship to the site at all, and — worse — it cannot be audited, because there is no underlying event to check it against. You cannot verify a percentage. You can verify a pour.
This is the deep argument for reporting quantities upward rather than percentages. A quantity survives aggregation. Cubic metres add up honestly at every level and can be checked against measured and docketed quantity whenever anybody doubts them. A percentage cannot even be added up correctly without weights, and the weights are usually nobody's job.
What to change on Monday#
Three changes, in order of how much they cost.
- Report quantities, not percentages. Let the percentage be calculated from the quantity and the schedule, by whoever wants one. Nobody types a percentage into a form again.
- Make milestones binary and dated. Cast, commissioned, tested, handed over. Record the actual date, not the planned one, and keep both so slippage is visible without a meeting.
- Separate reporting from judging. As far as the organisation can manage it, the person who records what happened should not be the person whose appraisal depends on it. Where that is impossible — and on a small site it often is — make the record so factual that judgement has nothing to grip: quantities, dates, photographs.
The daily record is where this either works or does not. If the site's own daily progress report already carries quantities, labour present and plant deployed, the weekly figure can be built from it rather than composed for the meeting. If the daily record is a paragraph of prose, the weekly figure will always be an opinion, however sincere.
The number we refused to print#
We build a site status screen. It reports the floor reached, what is blocking work, what material is awaited, and what falls due this week. It shows no progress percentage at all, and that is a decision rather than an omission.
Nothing in the records measures how much of a slab is done. We hold deliveries, attendance, tasks, bills and photographs, and not one of them answers that question. A figure assembled from them would have been arithmetic performed on an opinion — and it would have been read as a measurement, because a number on a screen always is.
Leaving that space empty is uncomfortable. Every screen wants a headline figure. A screen that says what it knows and then stops is worth more than one that says what it was asked for.
The short version#
Reported progress runs ahead of real progress because a percentage is a forecast made by the person being marked on it, and because every level of the chain rounds the same way.
Report what happened instead: quantities against the schedule, milestones that are yes or no, dates that are actual, photographs that carry a place and a time. Let anyone who wants a percentage calculate one. The estimate will still be optimistic. The record underneath it will not.