Numbers in your bullets: the heaviest single evidence check there is
The short answer
Numbers in your bullets is a 15 point check in our published rule set. Outside the two total parse failures, which are worth 35 points each because they mean the software gets nothing at all, it is the heaviest single check there is.
It is worth more than fixing a two column layout, which is 10. It is worth more than tables and text boxes combined at 8 each would suggest relative to their reputation. And unlike almost everything else on the list, it has nothing to do with file formats. It is about what your sentences say.
A bullet without a number is a claim. A bullet with one is evidence. That is the entire idea, and it is worth 15 points because it is the difference between a resume that asserts and a resume that demonstrates.
Why this one carries so much weight
Evidence is worth 25 of the 100 points, second only to parse at 35. Quantification takes 15 of that 25 on its own, with overlong bullets at 6 and enough bullets to judge at 4 making up the rest.
The weighting reflects something simple about how resumes are read. A reader arrives at your bullets already knowing you are going to say positive things about yourself. That is what the document is for. So the positive statements themselves carry very little information. What carries information is anything a reader could in principle check, or compare, or be surprised by.
Consider two versions of the same fact.
Responsible for improving the customer onboarding process.
Redesigned customer onboarding, removing four handoffs and two manual approvals. Median time to first value fell from eleven days to three.
The first is a job description. It tells a reader what you were assigned. The second tells them what changed because you were there, and it does it in a way that invites a follow up question in an interview, which is exactly what you want a bullet to do.
The number does not have to be a percentage
The most common objection to this advice is that the work did not produce a metric. Usually it did, and the number in question is not the one people reach for first.
Percentages get all the attention and they are frequently the weakest option, because a percentage without a base is close to meaningless. Improving something by a large percentage means one thing at scale and another thing entirely when the starting number was tiny.
Numbers that are usually available and usually better:
Counts. How many people, teams, systems, customers, markets, products. How many spreadsheets you replaced with one view.
Durations. How long something took before and how long it took after. How long a process ran. How quickly you shipped.
Scale. The size of the thing you were responsible for. The number of users it served. The size of the budget.
Frequency. How often something happened before and after.
Money, where you can say it. Revenue, cost, savings, budget owned.
Almost every piece of work has at least one of these attached to it, and the ones people forget to use are counts and durations, which are the easiest to remember accurately and the hardest to argue with.
What to do when there is genuinely no number
Sometimes the honest answer is that no number exists. The work was qualitative, or you left before the result landed, or the measurement was never taken.
Do not invent one. A fabricated metric is a bad trade at every level: it is the kind of thing that gets tested in an interview, and the person testing it will be the person who does the job you are applying for.
What works instead is specificity of a different kind. Name the concrete thing rather than the category.
Led cross functional collaboration is a category. Ran the weekly review that product, legal and operations all attended, and which set the launch order for the quarter is a specific thing, and a reader can picture it. It carries no number and it is still evidence, because it names something that either happened or did not.
The instinct to reach for is the same one either way: replace the abstraction with the particular.
The two checks either side of it
Overlong bullets is 6 points. A bullet that runs to four lines has usually buried its own result somewhere in the middle. The number that justifies the whole sentence arrives after the reader has stopped.
The fix is usually to split rather than to cut. If a bullet contains two achievements, it should be two bullets. If it contains one achievement and a lot of context, the context can go.
Enough bullets to judge is 4 points, and it points the other way. A role with a single bullet under it gives a reader nothing to assess. It reads as a placeholder, and it makes the role look like something you would rather not discuss.
Together the three describe a shape rather than a rule. Enough detail to evaluate, not so much that nobody finishes the sentence, with a checkable fact in as many of them as you honestly have facts for.
Where to spend your time
This is the check to start with once your file is readable, and the order matters.
First confirm the file parses at all, because a locked file or a file with no text layer costs 35 points each and both look completely normal on your own screen. That takes about ten seconds to test and is covered in the two resume files that score zero.
Then come here, because 15 points is the largest single thing left and it is entirely within your control. It does not require a template, a tool or a redesign. It requires going through your bullets one at a time and asking, for each one, what changed and by how much.
Then move to the layout checks, which are real and smaller: multi column layout at 10, tables and text boxes at 8 each, contact details in the header at 6. Those are covered in two column resumes and tables and text boxes in a resume.
The reason for that order is that most people do it backwards. Reformatting feels productive and finishes in an evening. Rewriting bullets is slower, harder, and worth more than all the formatting put together.
A note on keywords
One check in the consistency category is worth mentioning here because it cuts against a lot of published advice: keyword stuffing is penalised, not rewarded. It is a 3 point check, and it exists because modern screening flags density rather than treating it as relevance.
That does not mean the language of the posting is irrelevant. Using the employer's own words for the things you genuinely did is sensible, and knowing which words those are is a skill worth having, which we cover in pulling keywords from the job description.
It means the goal is to describe your real work in the posting's vocabulary, not to append a block of terms. A quantified bullet written in the employer's language does both jobs at once.
For the full ranked list of checks and weights, see what actually breaks a resume in an ATS. To see which findings apply to your own file, the free ATS checker runs the whole published rule set in your browser and uploads nothing.
iapplyai.app scores this check on your own file and shows you which bullets are carrying a claim without evidence, on the free ATS checker. Fixing those is the highest return hour available to most resumes, and it needs no template, no tool and no redesign.
