Guide 04
Quantifying impact: find numbers you already have
A bullet with a number in it can be compared with the job you are applying for. A bullet without one can only be admired. The barrier is almost never writing skill: it is that nobody wrote the number down at the time. This guide is about where the numbers already are, how to size work you shared with other people, and how to write an estimate without inventing precision.
On this page
A recruiter reading thirty similar bullets is not looking for the most impressive one. They are looking for the one they can picture themselves inheriting. A number is the fastest way to make a sentence pictureable, because it forces you to say what the work was attached to.
Why the number changes the reading
Consider two candidates describing the same internal systems project. One writes that they improved reporting across the business. The other writes that they replaced a nine-tab spreadsheet with a single report that eleven people in three departments now run themselves. The second candidate has not necessarily done more work. They have just described it in a way that a stranger can hold, and a stranger is the entire audience.
Numbers also do something subtler: they let a reader skip. Without a number, a reader has to read the whole bullet to decide whether it is relevant. With one, they can scan a column of bullets, spot the two that match the job, and read those properly. On a document being read in six seconds, that is the difference between your best work being read and being passed over.
- Supported the implementation of a new booking system across the business.
- Helped the team with training and rollout.
- Worked with several departments to make the transition go smoothly.
- Coordinated the rollout of a new booking system to 40 staff across three sites in a single quarter, with no change to opening hours.
- Trained 38 of those 40 staff in person and wrote the two-page guide the rest used.
- Retired the parallel spreadsheet on day one of the new system, ending a month of double entry.
Every figure in the right-hand version is a count the writer could have produced at the time: how many people, how many sites, how long, how many trained, how many left. None of them required a dashboard. That is the general shape of the exercise, and the rest of this page is a list of places to look.
Four kinds of number worth citing
Not every number earns its place. These four answer four different questions, and between them they cover almost every job.
| Kind | The question it answers | Looks like |
|---|---|---|
| Scale | How big was the thing you worked on | about 1,400 orders a month · 22 sites · a team of 9 · a £3m budget line |
| Change | What moved, and in which direction | from nine days to two · down from 40 tickets a week to 12 · up from 4% to 27% |
| Reach | How widely the effect travelled | three regions · two languages · every branch · 11 departments |
| Rate | How efficiently it happened | four days of manual work a quarter removed · under 2 minutes per case · one person-day instead of five |
Scale and change do most of the work on an experienced resume. Reach is the one people forget, and it is frequently the strongest available signal for someone early in a career: built the same process for all seven sites in my first year tells a reader the work was repeatable, which is exactly what a manager hiring a second person wants to know. Rate is what makes an operations or finance bullet concrete, because it removes the adjective.
One number is enough per bullet. A bullet carrying four figures is a bullet nobody can hold in their head, and it reads as though the writer was substituting precision for thought. Pick the one that best answers so what, and put the rest in the interview where you can expand.
Counts that are not measurements
Plenty of resume evidence is countable without ever being measured, and this is where most people who think of themselves as "not a numbers person" give up too early. None of the following requires a metric, a report, or a memory of a percentage.
- From nothing to something. There was no process, and now there is. Wrote the first onboarding procedure the department had ever used.
- From one to several. One clinic became four. One region became three. One language became a two-language team.
- Frequency. Weekly, daily, twice a term, once a quarter, every Monday for three years. Cadence is a number and a strong one, because it describes reliability rather than effort.
- Breadth of coverage. Every branch, all shifts, the full contract term, all but one of the sites.
- Headcount you were responsible for. Not the department size. Yours.
- Version, edition, or era. Moved three systems from one platform to another. Wrote the copy for the fourth edition of the handbook.
- Duration. Three years, two terms, eighteen months. Length of time sustained is evidence of something.
A support lead who cannot name a single metric can still write that they ran the triage queue for a team of twelve on weekdays and took the out-of-hours rota on Fridays. That sentence contains five quantities and describes a job a hiring manager can picture precisely. It is better evidence than responsible for escalated customer issues, and it took no dashboard to produce.
Where the number already is
You did not write the numbers down, but you worked inside systems that did. Here is a list of places to look, in roughly the order of how quickly they give an answer. The first four take under ten minutes each for most people.
| Technique | What it gives you | Where to look |
|---|---|---|
| Count the artefact you still have | The size of the thing you built or ran, exactly | The spreadsheet, the ticket queue, the CRM export, the ledger, the rota, the shared drive folder, the old inbox |
| Ask the system you already have access to | Anything you were responsible for, in one query | The dashboard, the report builder, the admin screen, the annual review document |
| Take the before-and-after pair | The strongest single figure on any resume | Your memory of how it was before you changed it, dated against how it was after |
| Work backwards from a roster | A defensible scale figure plus your share of it | The team list, the org chart, the staffing plan, the shift pattern |
| Scale up from a pilot | A large figure with a real basis behind it | Whatever you tried in one place first and then rolled out |
| Time to money, with the basis stated | A money figure you can defend in the interview | Hours saved, multiplied by a rate you name out loud in the bullet |
| Count the cycle, not the task | A figure that reads as an operating improvement | Close, billing, payroll, release, rota, reporting, audit, campaign cycles |
| Quote the requirement you met | A capacity or scale figure that is definitionally true | The spec, the tender, the service level, the contract you delivered against |
| Count the thing that replaced you | Evidence of lasting effect rather than effort | The runbook, the checklist, the template, the SOP you wrote that is still in use |
| Ask three people | The figure you would have written down at the time | A former colleague, a former manager, someone who used the thing you built |
Three of these deserve expanding, because they carry most of the weight.
Counting the artefact
This is the most reliable and the most overlooked. If you built a process, the process is usually still there, and it usually has a countable record. Open the spreadsheet and scroll to the oldest sheet. Look at the ticket queue and set the date filter to the week you started. Count the rows in the ledger, the entries in the CRM for your accounts, the documents in the shared folder you own, the entries in the training matrix.
You do not need precision, and precision here is dangerous, because these systems are frequently still being added to. Round to something you can defend. About 600 is true of anything from 500 to 800, and it will still be true in a year.
Asking three people
Memory of a job is far better held by the people around it than by the person who did it. A former colleague remembers the queue size, because they worked it. A former manager remembers the headcount, because they approved it. A customer remembers what it was like before. One message asking roughly how many did we handle a week? will get you an answer in a large fraction of cases, and you can write about in front of it.
This is also the cheapest way to sanity-check a figure you have already found. If you remember 400 and two other people say a few hundred, you have about 400. If they say ten, one of you is wrong, and it is cheaper to find out now than in an interview.
Converting time to money
Turning hours into pounds is the most abused figure on any resume, and it becomes honest the moment you name the basis. "Saved roughly 30 hours a month" is checkable. "Saved £14,000 a year" is not, until you say how you got there. If you want the money version, put the arithmetic in the bullet or be ready to say it in one sentence in the interview: about 30 hours a month, valued at the team's standard rate.
If you cannot name the rate or the basis, use hours. Hours are more honest and frequently more persuasive, because they describe capacity returned to a team rather than a saving that may or may not have been banked.
Three extractions, start to finish
Three people who would all say they have no numbers, and what each of them ends up with.
A support coordinator
Starting point: "Handled customer queries and complaints." No system access, no reporting role, three years in the job.
- The counts she already has: the number of people in the team, the number of sites, the days she covered, the hours of the out-of-hours shift. These are knowable without asking anyone.
- The artefact: she kept a handover sheet. Counting the rows from the first month to the last gives her a monthly figure she can round.
- The before-and-after: the old process was a shared inbox that everyone watched; her process was a rota with named owners and a same-day target. She can state the target and say she met it for a period she can name.
- The result: Ran the customer query queue for a team of 9 across 3 sites on weekdays, and the out-of-hours rota on Fridays, for three years. Replaced a shared inbox with a named rota and a same-day response target, which the team still uses.
A marketing officer
Starting point: "Ran email campaigns and social media accounts."
- The artefact: the mailing platform still has the historical sends. The list size at the start of each campaign is a count, and the open and click rates are figures she can read rather than remember.
- The cycle: campaigns ran every two weeks. That cadence, sustained, is a fact about reliability and it is a number.
- The before-and-after: the subject-line test she set up is the cleanest change on the page, and it has a number attached to it on both sides.
- The result: Sent a fortnightly campaign to a list that grew from 1,200 to about 6,500 over eighteen months. Rewrote the subject-line test, doubling click-through across the next six sends.
A finance analyst
Starting point: "Supported the monthly reporting cycle and helped the team with forecasting."
- The cycle: the month-end close. How long it took, how long it took afterwards, and how many people it involved. This is a figure any finance person can state exactly.
- The basis: the old pack had a known number of tabs and hand-built components; the new one was automated. Hours per close is a rate figure, which is the strongest kind here.
- The confidentiality problem: the totals are not publishable, so she uses the rate instead: cut the month-end pack from six days to two and roughly 20 hours of manual work per close. No sensitive figure, and the improvement is still completely concrete.
- The result: Rebuilt the month-end reporting pack, cutting the close from six days to two and removing about 20 hours of manual reconciliation per cycle.
Note what all three did. None of them started with a dashboard. Two counted something they already had, one used a cadence, and all three used the before-and-after pair, which is the easiest of the four techniques and the one most often skipped because it requires remembering two states of the world rather than one.
Saying about without guessing
An estimated number, marked as an estimate, is entirely credible. An estimated number presented as a measurement is not, because somebody will check it. The whole trick is in the word you put in front.
| Write | Reads as | Right when |
|---|---|---|
| about 400 orders a month | A rounded, defensible figure | Almost always. This is the default for a remembered number |
| roughly 40 staff, across 3 sites | Scale you know approximately, detail you know exactly | You are confident about the sites and unsure of the headcount |
| over 100 a week for two years | A floor you are certain of | You know you never counted the whole period, only that it was busy |
| a third off the response time | A ratio rather than a measurement | The base is confidential, or you genuinely only remember the proportion |
| one of a team of 4 | An honest share of a known total | Shared work. This is better than claiming all of it |
| from 40 a week to 12 | Two rounded figures, same units | The endpoints are the part you are confident about |
| 412 orders in the sample month | A measurement | Only if you read it. If you are remembering, this is the one to avoid |
Two rules make the rest easy. First, round to a number you would be comfortable defending across a range: 400 rather than 437, 6,500 rather than 6,478. Second, never round in the direction that flatters you. If the work reduced complaints by somewhere between a fifth and a third, reduced complaints by about a quarter is honest, and a reader who later sees the underlying figure will find the claim held.
There is one thing not to do, which is to attach precision you do not have. A figure that looks measured invites a follow-up question, and the question is not hostile: it is the ordinary curiosity of somebody deciding whether to spend an hour of their week on you. Being able to say I am estimating, it was roughly 400 is a better answer than a number you cannot source, and it costs you nothing.
What a number cannot do
Quantifying is not a way of making a weak bullet strong. It is a way of making a strong bullet legible, and it has failure modes of its own.
- The biggest number is not the best number. 14,000 accounts touched is less interesting than a process you built that is still in use, if the second one is the relevant thing. Relevance is decided against the posting, not against the size of your old job.
- A percentage with no base is decoration. Improved efficiency by 40% is weaker than cut four days of manual work a quarter, because the reader cannot check the first and can picture the second.
- A number does not fix a job that is not relevant. A quantified role in a different field is still a different field. Edit the relevance first, then quantify what is left.
- Every bullet quantified is exhausting. A page where all fifteen bullets carry a figure reads like a sales report. One figure in each of your three or four strongest bullets does more work than five in a weak one.
- A number you cannot explain is worse than none. If you would be uncomfortable being asked where it came from, it does not go on the page.
Next
Layout is the last variable, and the only one where the honest answer is usually no. The two-column guide sets out the real costs, the three situations where a sidebar genuinely holds up, and a decision rule you can apply in about thirty seconds.
Questions about numbers
What if I genuinely do not know any numbers?
You know more than you think, because a lot of resume evidence is countable without a metric. How many people, how many sites, how many languages, how many days a week, how many products, which version of which system, how many reports a cycle, how many years. A support lead who cannot name a metric can still write that they ran the triage queue for a team of twelve on weekdays and the out-of-hours rota on Fridays. That is a specific claim, it is checkable, and it is a number. Only after that list is exhausted is a bullet genuinely unquantifiable.
Should I put a percentage on every bullet?
No, and a percentage with no base is weaker than the raw count it came from. Cut the close from nine days to two and cut the response queue from 40 tickets a week to 12 both say something a reader can picture. Improved efficiency by 40 per cent says nothing, because 40 per cent of what, and measured how. Use a percentage when the base is large enough to be uninteresting on its own, such as a headcount or a budget, and use counts when it is not.
Is it acceptable to round a figure down?
Yes, and it is the honest way to write an estimate. Rounding to a number you can defend is not the same as inventing precision, so about 400 is a good line to write and 412 is a bad one if you are remembering rather than reading. Rounding down costs you nothing and survives a follow-up question, because roughly 400 is true of anything between 350 and 450. A number that looks measured but is not will be checked, and the whole page becomes less credible with it.
How specific does the number need to be?
Specific enough that a reader can picture the operation behind it. That usually means pairing the figure with what it counted: 1,400 orders a month reads very differently from 1,400, and so does a 30 per cent cut in something you have not named. A figure attached to the wrong noun is decoration. If the number is the whole point of the bullet, the noun is doing the work and it needs to be there.
Do numbers help with an applicant tracking system?
There is no documented evidence that they do, and anyone claiming otherwise is guessing. No vendor publishes matching logic, and Greenhouse documents that Talent Matching works from skills, job titles, years of experience, dates and company names, with no mention of figures. Numbers are worth having because they make a sentence concrete for the human reading it, and because the process of finding them makes you describe the work properly. They are not a trick for a filter. What genuinely affects parsing is covered in the ATS formatting guide.
What if the figures are commercially confidential?
Say the shape instead of the value, and the shape is usually more informative anyway. A specific supplier spend becomes the largest non-payroll budget line, with the change described as a share rather than a sum. Revenue becomes a named segment, a growth rate, or the number of months a trend held. Percentage and ratio framings survive confidentiality because they disclose the direction and the magnitude without disclosing the base, and they are usually what a hiring manager actually wants to read.