Grafá

What the owner put in, and what came out

A People team with 100 engineers runs its yearly pay review on a workbook. This is everything they did, and everything Grafá did. The company, its data, the brief and the answers are made up for this example; everything Grafá produced is its real output.

1
workbook, uploaded as it was
72
words on what it is for
8 of 10
questions answered, a sentence each
4
comments on the summary
THE OWNER · 1

Uploaded the workbook, no clean-up

7 sheets and 2,539 formulas, with a short README sheet written by its author. The first rows of “Engineers”:

ABCDEFGHI
1IDNameLevelTechnologyStart dateFTEMonthly gross (EURStatusTenure (years)
2E001Georgi KolevMidJava28.08.20220.51320Active=ROUND(YEARFRAC(E2,AsO…
3E002Martin PopovMidPython19.06.202012700Active=ROUND(YEARFRAC(E3,AsO…
4E003Viktoria VasilevLeadFrontend03.03.201615610Active=ROUND(YEARFRAC(E4,AsO…
5E004Ivan KolevSeniorFrontend25.08.202213980Active=ROUND(YEARFRAC(E5,AsO…
6E005Raya HristovaMidPython02.10.202412580Active=ROUND(YEARFRAC(E6,AsO…
7E006Nadia PetrovMidJava20.04.202512800Active=ROUND(YEARFRAC(E7,AsO…
THE OWNER · 2

Wrote what it is for

People team of a software company with about 100 engineers. We use this workbook in the yearly compensation review: it compares every engineer's salary with colleagues of the same technology and level, with the market salary survey and with what comparable contractors cost us, and lists the outliers: people who are underpaid and at risk of leaving, and people who are overpaid. HR and the CTO read the Outliers and Summary tabs.
THE OWNER · 3

Answered what they knew

Grafá asked only where the workbook was unclear or contradicted itself. Skipping is fine.

Should colleagues who are leaving count when we work out the peer median? The guide says "everyone"; the file leaves them out.
Leavers stay out of the peer median. The README is outdated, the formula is right.
Should people who are already leaving appear in the Outliers list and the Summary counts?
Keep them in for now, we still want to see them in the review.
Is the "Overpaid" rule right? It is 15% or more above peers, or 20% or more above the market middle (P50), and the guide never mentions it.
Yes, those are our thresholds: +15% vs peers or compa-ratio 1.20.
Do you have a market figure for QA Leads? The survey sheet has none.
We have no survey figure. From now on QA Leads should use the QA Senior P75 as their market reference.
The 61k market reference for Maria Vasilev (E037): is it a standing exception, and should it also be used in the contractor comparison?
Yes, a standing exception agreed with the CTO. Only for her own compa-ratio and market position, not for the contractor comparison.
When comparing contractor cost with an employee at the market middle, should the employer's social security be capped, as it is for our own staff?
Yes, it should be capped exactly like for our own staff. Please change that.
For part-timers, should "cost vs contractor" use what a full-time version of the person would really cost (with the cap)?
Is the yearly salary exactly 12 monthly payments, with no 13th month or fixed bonus?
Yes, exactly 12 monthly payments, no 13th salary and no fixed bonus.
Should someone with, say, 1 year and 11.5 months count as having the 2 years needed for "retention risk"?
Should the Summary's headcount, costs and median salary keep including people who are leaving?
Keep all statuses in the Summary.
THE OWNER · 4

Read the summary, said what was off

A plain description of what the workbook does, line by line. Their comments, and what Grafá did with each:

For QA Leads use the QA Senior P75 as their P50, and leave P25 and P75 empty: show 'P50 reference only' instead of a band. Same for any QA Lead contractor.
This should include the 13th salary we pay in December.
People on leave should not count as colleagues either, only active ones.
I don't see the contractor whose contract has a fixed 200 days a year instead of 220. Is that handled?
GRAFÁ DID THE REST

The same workbook with an empty brief and no answers at all also came out identical on 120 of 120 cases: answers make the app more right; they are not needed to make it work.

Open the app it made   Try it with your workbook