Bullets Per Job: Distribute Them, Don't Count Them
The commonly repeated answer is a fixed number of bullets for every role, and it is the wrong shape of answer. Detail on a resume is a budget you allocate, and giving every role an equal share tells a reader that every role is equally relevant — which is almost never true. The decision worth making is distribution: which role gets the most space, which gets a line, and which gets only its title and dates. Get that right and the per-role count takes care of itself.
Why an even count misleads
A reader uses visual weight as a signal before reading a word. Four bullets under each of five roles says: these five things are comparably important. If one of those roles is the reason you are a credible candidate and another was a stopgap a decade ago, you have flattened the most useful information you had.
The even-count habit also produces a specific kind of filler. Once a role owes you four bullets and you have two real things to say, the remaining two get invented out of duties — sentences that restate the job title in longer form. Those lines actively cost you, because a reader who hits two generic bullets starts skimming the rest.
And the reverse happens on your strongest role: a fixed cap forces you to cut the third and fourth genuinely interesting thing you did, precisely where a reader would have wanted more.
Allocate against the posting, not against the calendar
Recency is a decent default because recent work is usually more relevant, but relevance is the actual criterion and the two come apart often.
Start from what the role is about. Two or three things, taken from the posting.
Rank your roles by how much evidence each provides for those things. Not by how much you enjoyed them or how senior you were.
Give the top-ranked role the most space, and let the drop-off be visible. A reader should be able to see, from the shape of the page, which experience you think matters most.
This is why the allocation changes between applications. The same career, applied to two different roles, should not produce the same distribution — and if it does, you have not tailored anything. Keeping one long source document and cutting down from it per application makes this cheap rather than exhausting.
The four sizes a role can be
Full treatment. Your most relevant role or two. Enough bullets to cover scope, the work itself, and what came of it. No arbitrary ceiling.
Reduced. Relevant but not central. A small number of lines covering the most transferable part.
One line. Older or tangential roles where the fact of the role matters and the detail does not. A single sentence, or even a phrase.
Title and dates only. Roles that exist to keep your timeline continuous. This is a legitimate and underused option, and it is the answer to most “but I have too many jobs” problems — it also avoids creating the appearance of a gap, which is the trade-off discussed in which roles to list at all.
There is no rule that all roles get the same treatment, and using all four sizes in one document is a feature. It is the clearest way to show a reader where to look.
What makes a bullet worth a slot
Length is not the test. The test is whether the line says something a stranger could not have guessed from your job title.
It names something. A system, a domain, a scope, a constraint, a decision.
It has a consequence attached, or an honest scope statement instead. Either what changed, or what you were responsible for. Both are informative; the second is often more defensible.
It is about you. The most common failure is a bullet describing a team’s outcome under a first-person implication, which is a claim you may not be able to hold under questioning — the distinction is worked through in your contribution versus your team’s result.
The verb is accurate. “Led”, “owned”, “supported” and “contributed to” describe genuinely different relationships to a piece of work, and picking the flattering one is a claim, not a style choice. That is the argument in the verb you choose is a claim.
A line that fails these is not fixed by rewording. It is fixed by finding a different thing to say about that role, or by accepting that the role only needs a line.
Symptoms of a bad distribution
Every role looks the same on the page. You have hidden your best material in a uniform block.
Your oldest role has the most bullets. Usually because it was written first, years ago, and never re-cut.
A bullet appears under two roles in slightly different words. You did the same kind of work twice and described it twice. Keep the better instance and use the second role’s space for something else.
Sub-bullets. Nesting is almost always a sign that a role has been given more structure than its content justifies. Flatten it.
A role with one bullet that repeats its own title. Either find something real to say or drop to title and dates only.
Where AI help goes wrong here
Asking a tool to “add bullets” to a role is asking it to generate claims, and it will. The output is fluent, plausible, and not sourced from anything you did — which is the exact drift pattern described in where AI-drafted bullets drift.
Asking a tool to “make each role consistent” is worse, because consistency is the defect. It will even out the distribution you deliberately created.
The useful version is narrow: give it your own raw notes for one role and ask it to propose tighter phrasings of what is already there, then check each proposal against your record. Anything that arrived with new information in it did not come from you, and checking a bullet against your own record is how you catch it.
A pass you can do in twenty minutes
Print the document. Without reading, look at the shape: where is the visual weight? Is it where you want a reader’s attention?
Then read only the first bullet of each role. Those are the lines most people actually read. Are they your strongest, or are they whichever one you happened to write first?
Then find the three weakest bullets in the document and delete them. Do not replace them. Reread. In most drafts the page is better, and you have space you did not have before.
Finally, check that each surviving specific is something you could answer a follow-up question about, which is what reading your resume as a sceptic is for. A bullet that survives a distribution pass and fails a factual one still has to go.