Effort

"Build the most valuable thing first" has a hole in it. A six-month project and a three-day fix sit in the same list, and if you only look at what you get, the six-month project wins. The quarter spent on it is the quarter where ten three-day fixes stay untouched.

Pagedeck planning worked out that way once. Two months went into the candidate everyone agreed was most important, while six items left waiting would each have taken two days.

Effort brings that cost into the table as a division. When two candidates return the same thing, the cheaper one moves ahead; when one returns twice as much but costs three times more, it moves back.

The unit is the person-month. Work that takes one person a month is a 1, and two people for two weeks is also a 1.

The number covers more than building. Research, design, development, review and post-release cleanup all count — everything up to the point where the work is finished. One week of research, two weeks of design, four weeks of development and a week of review add up to two months, so the value is 2.

Who fills the column in is a rule, not a preference: the builders. When the product side fills it in alone the number usually lands at half of reality, because they aren't the ones who know what has to be touched.

Keep the scale in half steps. Writing 1.3 doesn't make the estimate more accurate. This is an estimate like reach or impact, so getting the order of magnitude right is the goal.

In the calculation, the other three values are multiplied and then divided by this one. Reach of 2,000 times impact 2 times confidence 0.8 gives 3,200; at three person-months that's 1,067, and at two it's 1,600. Half a month moves the score by 50 percent, so one optimistic entry tilts the whole table.

"Theo wrote that three person-months, right? If product filled it in I'd like it re-done."

Checking the rule about who is allowed to produce this value.

"Same reach, same impact, and effort is three against one. That settles the order."

When the return is identical, the cheaper one goes first.

"This is eight person-months, so it doesn't belong in this table — it needs splitting."

The candidate is too large for one quarter and should be broken up.

At Pagedeck this column belongs to Theo. The permission controls an enterprise account asked for went in at two person-months, then rose to five once the existing account structure turned out to need reworking too. Five person-months means the team spends half the quarter on one item, so Mia asked whether everything else was getting pushed.

What Theo proposed instead was shipping read-only permissions first, leaving the account structure alone. Reach and impact stayed roughly the same while the value dropped to 1.5 person-months, which tripled the score. Without a column for cost, nobody has a reason to look for the smaller version of a feature.