Percent complete reporting lies about briefs and here is the arithmetic
Percent complete reporting fails on event briefs because a brief ships only when every section is approved, so partial credit across sections aggregates into a number that no edition can be launched from. Report the count of fully approved sections out of the total, plus the named blocker on each unapproved one.
The slide says the autumn portfolio is 92 per cent complete. Eight editions, thirteen sections each, and a bar chart with almost no white left in it. Somebody in the room asks how many of the eight are ready to hand to design. The answer is none.
Nobody fabricated anything. Percent complete reporting produced a technically correct number that describes work performed and says nothing about output produced, and the gap between those two is the whole problem.
The arithmetic that makes ninety per cent worthless
A brief ships when every section is approved. That makes the brief a conjunctive deliverable, and conjunctive deliverables do not average.
Take one brief of twelve sections, each reported at 90 per cent. If you read that as work done, you get 90 per cent and a comfortable feeling. Read it as a probability instead, treating 90 per cent as a nine in ten chance that each section is finished on time, and multiply: 0.9 raised to the twelfth power is 0.282. Under that reading the brief has about a 28 per cent chance of being ready.
Both readings use the same input and they disagree by a factor of three. The second one is closer to what teams observe, which is that portfolios sitting at ninety per cent for three weeks are common and portfolios finishing three weeks early are not.
The cleaner way to see it needs no probability at all. Twelve sections at 90 per cent each is zero shippable briefs, because 90 per cent of an approval is not an approval. There is no partial state in which design can begin, sales can quote, or the site can go live. The only number with a unit attached is the count of approved sections, and on that brief it is nought.
Why does the number keep rising while nothing ships?
Because the remaining work is not like the completed work, and the percentage assumes it is.
The first 80 per cent of a section is writing, which the author controls and can do on a Tuesday. The last 20 per cent is getting somebody else to agree, which the author does not control and cannot schedule. Those are different activities with different distributions, and a single percentage smears them together.
Goldratt made the general version of this argument in Critical Chain in 1997. His account is that individual task estimates carry substantial safety inside them, that the safety gets consumed rather than saved, and that the two mechanisms doing the consuming are work expanding to fill the time available and people starting late because the deadline looks distant. His sharpest observation for reporting purposes is the asymmetry: delays on one task pass straight through to everything downstream, while a task finishing early usually banks nothing, because the next person is not ready to start.
Applied to briefs, that asymmetry is why a portfolio percentage is a ratchet. Sections that finish early do not pull the number forward, because the brief still waits for its slowest section. Sections that slip push the number back for everyone. The average conceals a maximum, and the maximum is what governs the ship date.
The software version of the same observation is older and shorter. In the September 1985 issue of Communications of the ACM, volume 28 number 9, Jon Bentley printed a collection of one line maxims under the heading Bumper-Sticker Computer Science, and one of them, credited to Tom Cargill of Bell Labs, appears under the label Rule of Credibility: "The first 90 percent of the code accounts for the first 90 percent of the development time. The remaining 10 percent of the code accounts for the other 90 percent of the development time." The joke works because the arithmetic is deliberately broken and everybody recognises the shape.
What do you report instead?
Count approved sections. That is the whole change and it is unwelcome the first time you do it, because the number is much smaller and much less flattering.
For one brief, report approved out of total. Eleven of thirteen. For a portfolio, report two numbers: the count of briefs where every section is approved, and the distribution of the rest.
Take the eight edition portfolio from the slide. Under percent complete, it is 92 per cent. Under section counting, suppose the approved counts are 13, 13, 12, 12, 11, 11, 10 and 9 out of thirteen. Two briefs are complete. The total approved sections are 91 out of 104, which is 87.5 per cent, so even the honest percentage is close to the original figure. But the shippable count is two of eight, which is 25 per cent, and 25 per cent is the number that predicts what happens next month.
That gap, 87.5 against 25, is the thing to put on the slide. It is not a criticism of anybody's work. It is a statement that thirteen missing approvals are distributed across six briefs in a way that blocks all six.
The distribution is worth showing as well as the count, because it changes the decision. Thirteen missing approvals spread one or two per brief across six briefs is a very different problem from thirteen missing approvals concentrated in two briefs that are barely started. In the first case you are close, and a fortnight of chasing clears four editions. In the second case two editions are in trouble and the other six are fine, and chasing everybody is the wrong response.
A short histogram does the job. Two briefs at thirteen of thirteen, two at twelve, two at eleven, one at ten and one at nine. Eight bars, no percentages, and a reader can see in a second that six briefs need between one and four approvals each. That is a picture of a portfolio a fortnight from done, and it is the same underlying data that produced a comfortable 92 per cent.
Adding blockers turns a report into a decision
The count alone still leaves the reader with nothing to do. Attach two fields to every unapproved section and the report becomes actionable.
The first is a named owner, one person, for the next action on that section. Not a team, not a function.
The second is the state the section is in, from the fixed set your workspace uses, so a reader can distinguish a section nobody has drafted from one sitting with a reviewer for nine days. Those two situations have nothing in common and both currently render as not approved. This is why the status definitions have to be written down first; the report inherits whatever precision the states have and no more.
With those attached, the eight edition report becomes thirteen lines, one per missing approval, each with a name against it. That fits on a slide, and in a status meeting it converts a discussion about percentages into thirteen short conversations about specific people.
The one place partial credit belongs
There is a real case for percentages, and drawing the boundary honestly keeps the argument credible.
Where output is divisible, partial completion has value and the percentage means something. Six hundred exhibitor profiles for the show directory is divisible: four hundred finished profiles are four hundred usable profiles, and 67 per cent is a true statement about deliverable value. The same goes for translating a site into three languages where each language ships independently, or writing session descriptions where a finished description can go live on its own.
Briefs are the other kind. So is a rate card, so is a floor plan, and so is a registration form. Anything where the parts are only useful assembled should be reported as a count of finished units, and the unit is the assembled thing.
The test is one question: can somebody use this at its current state of completion? If yes, percent complete is fine. If no, the only defensible number is the count.
Where this stops
Counting approved sections has its own failure mode, and it is worth naming before somebody else does.
The count is only as good as what approved means. If a reviewer can approve a section by opening it and clicking, then eleven of thirteen is a count of clicks. That is why the count has to sit on top of written completion criteria for each section, and why adopting the count without the criteria makes the reporting worse rather than better, because it replaces a soft number with a hard number that is equally uninformed.
The count also flattens a real distinction. A brief missing two sections might be a week from done or a month, depending on which two, and the count treats them the same. Sequencing matters, because the positioning section blocks four others and the venue logistics section blocks nothing. If you want that visible, mark two or three sections as gating in the template and report the gating count separately, which is cruder than a dependency graph and about as useful in practice.
Finally, the count says nothing about elapsed time. A portfolio holding steady at two of eight for six weeks and a portfolio that reached two of eight yesterday look identical, and they are not. Age in state is the missing dimension, and measuring cycle time from request to live is where that belongs rather than here.
Do this at the next status meeting. Take whatever percentage is on the slide, and beside it write the count of briefs where every section is approved. Two numbers, no commentary. If the second number surprises the room, the first one has been doing damage for a while, and you now have the cheapest possible argument for changing what the brief workspace reports by default.
Questions people ask about percent complete reporting
- Why is percent complete misleading on an event brief?
- Because the sections are conjunctive. An edition needs all thirteen sections approved, so the value of twelve approved sections is zero editions, the same as the value of two. Averaging partial progress across sections produces a number that rises smoothly while the shippable count stays at nought, which is why the graph looks healthy for weeks.
- What should replace a percent complete figure in a status report?
- Two numbers and a list. The count of fully approved sections out of the total, the count of sections that have not moved in fourteen days, and a named owner against every section that is not approved. That fits on one line per brief and tells a reader what to do, which a percentage never does.
- Is percent complete ever useful for production work?
- It is useful where work is genuinely divisible and partial output has value, such as writing three hundred exhibitor profiles where each finished profile is usable on its own. It stops being useful the moment the deliverable is conjunctive, meaning every part must be finished before any of it can be used.
Related reading
- Section status definitions that stop draft and in review meaning the same thing
- A definition of done checklist for each section of an event brief