Define the metric precisely
Write a plain-language definition before recording the value. State what is counted, excluded, or grouped, and distinguish an output from a rate, percentage, duration, cost, or quality measure. “Requests closed” is different from “requests received,” and “average resolution time” is different from “same-day resolution rate.” If the original source uses a specialized term, preserve that wording and add a short explanation. Avoid turning a broad phrase such as “improved delivery” into a number without evidence. The definition should let another reader understand the measure without relying on memory or a polished application sentence.
Add the period and baseline
Attach a time period to every metric, even when the source seems obvious. Record a month, quarter, project phase, release window, or other supported interval. Then identify the baseline used for comparison, if there is one. A baseline may be a prior period, an agreed target, an initial measurement, or no comparison at all. Do not describe a value as an improvement when the earlier value is absent. Preserve the original dates or labels so a later writer does not confuse a reporting period with the date the note was created. If periods are not comparable, explain why and avoid calculating a change.
Capture source, unit, and calculation
Record where the metric came from and how it was calculated. A source can be a dated report, approved project record, service dashboard, budget document, or another career note that preserves the original evidence. State the unit beside the value: requests, hours, dollars, percentage points, percent, defects, or another supported unit. Keep input values when a result is calculated, and record the rounding approach. Do not replace missing source data with an estimate or treat a percentage-point change as a percentage change. If the source cannot be checked, mark the metric as unverified and retain the uncertainty.
Separate individual attribution from shared results
A team metric and an individual contribution are different records. Write the shared outcome first, then describe the person’s verified role, such as preparing a report, closing a documented subset of cases, changing a process, or coordinating a measured activity. Use attribution only when the source or first-hand record supports it. Avoid assigning the full result to one person because they participated in the project. When the contribution cannot be quantified, preserve the qualitative role and the team measure separately. This produces a more accurate evidence trail and prevents later documents from overstating personal ownership.
Audit gaps before reuse
Review each record as evidence rather than prose. Check that the definition, period, baseline, source, unit, calculation, and attribution agree with one another. Look for hidden assumptions: a percentage with no denominator, a target presented as an achieved result, a total copied from a forecast, or a team outcome described as personal work. Add a question for every missing field and preserve the original wording where interpretation is uncertain. This audit is complete when a reader can tell what is known, what was calculated, and what still requires confirmation. Do not promote an unchecked value into a verified metric.
Use the verified notes in the Radaryno workflow
Once the records are reviewed, keep them as buyer-maintained career notes and use only the supported evidence when preparing application documents. The Radaryno Resume Kit uses the buyer’s career notes, selects relevant experience, reports gaps, and asks rather than inventing. It can produce HTML and PDF resumes and an optional cover letter in the offer language, with a single-column ATS-friendly layout. The kit works with Claude Code or Codex on Windows, macOS, and Linux, but local file storage does not mean offline AI inference; provider access is separate. Do not expect the kit to submit applications, search for jobs, or contact employers.
| Record area | Clear entry | Weak entry | Why it matters |
|---|---|---|---|
| Definition | Resolved support requests | Improved support | The reader can identify what was counted. |
| Period and baseline | Quarterly total; prior quarter used as comparison | Handled many requests recently | The change has a defined time frame and reference point. |
| Source and unit | Service report; requests closed | Internal data; better performance | The value can be checked and its unit is explicit. |
| Attribution | Personally closed 48 of 120 requests; team total was 120 | Reduced the team backlog by 40% | Personal contribution is separated from the shared result. |
Worked example: preserve the evidence before writing
Start with a fictional service record stating that a support group closed 120 requests during a quarter, compared with 90 in the preceding quarter. The definition is requests closed, the period is one quarter, the baseline is the preceding quarter, and the unit is requests. The source is a dated service report. Keep these fields together rather than copying only the increase into a separate note.
Calculate only what the recorded values support. The change is 30 requests, and the relative increase is 30 divided by 90, or 33.3 percent when rounded to one decimal place. Store the unrounded inputs and the rounding rule. Do not infer that the fictional contributor caused the entire change unless the source establishes that connection.
Add attribution separately: the fictional contributor personally closed 48 requests, while the group closed 120. This supports a personal count of 48 and a team outcome of 120. If the source does not show how the contributor affected the remaining 72 requests, leave that relationship undescribed rather than assigning credit.
Record unresolved details as questions, such as whether reopened requests were included or whether the comparison periods had equal operating days. A later document can use the verified fields after those questions are answered, but it should not convert an incomplete record into a stronger claim.