Effort is not duration: when adding people helps

Effort (UK «трудомісткість») is the amount of work, in person-days. Duration (UK «тривалість») is the calendar time from start to finish, including waiting, briefing and the parts only one person can do. They're related, but duration is never simply effort divided by the number of people.

Leo's arithmetic (ten days of review, five reviewers, two days) treats the review as a pile of identical, independent pieces. It isn't. Part of it can be shared: reading and checking units. Part can't: someone has to hold all five units in one head for the final pass, and that's Vira. And every new reviewer costs time before they help, because they must learn what counts as an error, or you get five different standards to reconcile.

So before estimating how extra people change a date, take the task apart: what can be split, what can't, and what each newcomer costs to bring in.

Any task can be described by three parts, and duration follows from them.

• Work that can be split: shared among people. For Vira's review, 8 days of reading and checking units, about 1.6 days per unit.
• Work that can't: done by one person whatever happens. Vira's 2-day final pass across all five units, checking that terms, notation and difficulty line up.
• Coordination: 1 day per extra reviewer, to brief them and agree what counts as an error.

With n reviewers: duration = 8 ÷ n + 2 + (n − 1) × 1. Alone, Vira needs 8 + 2 = 10 days. With two reviewers: 4 + 2 + 1 = 7 days, three days off the critical path.

Notice the floor: Vira's 2 days stay the same however many people join. The splittable part shrinks towards zero, but never below what only one person can do. Danylo sees the same in software: two developers on one feature spend the first day agreeing how it should work.

Push the same formula to Leo's five reviewers: 8 ÷ 5 + 2 + (5 − 1) × 1 = 1.6 + 2 + 4 = 7.6 days. That's slower than two reviewers (7), and nowhere near the 2 days Leo expected.

Why? Each extra person shrinks the splittable part less than the one before (from 4 days with two reviewers to 1.6 with five), while coordination grows by a full day each time (from 1 to 4). The part only Vira can do never moves. Past a certain point, every new person adds more briefing than they remove reading.

The pattern is general. Adding people helps most when the first one or two join; after that the returns shrink and can turn negative, especially when newcomers need the attention of the very person who's on the critical path. The best number of people for a task is often small, and you find it by working the numbers, not by assuming “more is faster”.

The review formula charges one briefing day per extra reviewer. In real teams the cost of coordinating often grows faster than the number of people, because every pair of people may need to agree on something.

The number of pairs in a group of n people is n × (n − 1) ÷ 2. Two people: 1 pair. Three: 3. Five: 10. Eight: 28. People are added one at a time; possible conversations multiply. Five reviewers checking units against each other's standards could mean ten separate disagreements to settle.

That's why small groups with one clear standard-setter work better. Vira briefs each reviewer herself and owns the final pass, so the reviewers don't need to agree with each other; each agrees with Vira. The same idea runs through the whole project: fewer hand-offs, clear owners and written agreements keep coordination close to linear instead of letting it explode.

Adding people shortens the calendar but raises the effort. With two reviewers the review takes 7 calendar days, but the work is 8 + 2 + a briefing day for each of the two = 12 person-days, against 10 for Vira alone. You buy calendar time with extra effort, and someone pays for it: Hanna plans 8,000 UAH for the extra reviewer, as the plan's own assumption, from the 154,550 UAH left under the ceiling.

That trade is worth making only on the critical path. Three days saved there move the opening date; three days saved on a task with float buy nothing and still cost the extra effort.

How the person joins also matters. Vira can live with one extra reviewer: someone she knows, briefed by her, early in the review. A newcomer arriving in the last days spends the remaining time being briefed, and the briefing comes from the busiest person on the project.

An extra person shortens a project's schedule only when several conditions hold together:

• The task is on the critical path. Off it, finishing sooner doesn't move the end date. A task with 30 days of float gains nothing.
• Most of its work can be split into independent parts, like reading separate units. A task that is mostly one person's judgement, like Vira's final pass, barely shrinks.
• The newcomer joins early, so the briefing pays back over enough remaining days.
• The briefing doesn't come from the critical person's time. If a newcomer needs a week of training from the busiest person on the project, the critical path may get longer.
• The duration is work, not waiting. Days spent waiting for replies don't split.

Miss any one and the extra person costs effort and money without moving the date. That's why the one extra reviewer on Vira's review works: critical, mostly splittable, early, briefed briefly.

A gap between effort and duration is often waiting, not slow work. Roman's screening shows 9 person-days of effort but 18 days of duration: candidates take days to reply, and they're free only evenings and weekends. Half of the duration is waiting.

Waiting doesn't split between people. A second recruiter would send messages to different candidates and then mostly wait as well; the 18 days would barely move, while the effort and cost would grow. Asking Roman to “work faster” misreads the problem: he isn't slow.

What does shorten waiting: start earlier, so the waiting happens in parallel with other work, and ask differently, so there are fewer rounds of back-and-forth. Roman already asks candidates for their free slots in the first call, which saved a round of messages. Before adding anyone to a long task, check how much of its duration is work and how much is someone else's reply.

Before adding a person to any task, ask four questions, in this order:

1. Is the task on the critical path? If not, stop: a faster task doesn't move the end date.
2. Can its work really be split? Count the part only one person can do.
3. Who briefs the newcomer, and is that person critical too?
4. Is the duration work or waiting? Waiting doesn't split.

The first question comes first because it filters out most bad ideas cheaply. Availability, cost per day and working style matter only after it.

And if every answer says no? Then change the plan, not the headcount: start something earlier, run it in parallel, or cut scope on the critical path.

How it fits together: effort is work, duration is calendar time; extra people help only on the critical path, for splittable work, joining early, and they raise effort. The case next puts all seventeen tasks together and asks whether 1 February really holds.