← DevEval Value & ROI

Profitability per developer, per project, per client.

Computed daily from real delivery data — not reconstructed once a year in a spreadsheet. Five layers connect a merged pull request to money: Effort → Cost → PI → Value → ROI.

The five layers

From a merged PR to money, in five steps.

Import effort and cost data automatically from Jira Tempo, or configure rates and value models directly in DevEval. ROI calculations stay up to date as delivery data arrives.

01

Effort

CU delivered + MD spent

Two measures of the same work: CU (complexity delivered — constant over time, tool-independent) and MD (man-days — estimated for a baseline developer, and actual from worklogs). Works with whatever you have: Jira estimates, worklogs, or just the PRs.

02

Cost

actual MD × internal rate

Actual man-days times internal rate, overhead included. Multi-currency, aggregated honestly: what can't be converted is excluded, never silently blended.

03

Productivity Index

PI = CU delivered / actual MD

The efficiency layer: how much complexity ships per man-day. Same unit for every developer, team, and vendor — see the paradox below.

04

Value

depends on your pricing model

What the work is worth in money — billable hours, a CU-based share of the budget, a retainer base, or an internal proxy. Four models, one per project, below.

05

ROI

ROI = value / cost

1.0 = break-even; 2.0 = value equal to twice the cost — a 100% return on cost; 0.8 = a 20% loss you now see in week 2, not at year end.

The Skill-Time Paradox

Same task, same value — 20× different efficiency.

PI in one example: two developers deliver the identical task, and only the time differs. Hours-based metrics call the slower one more valuable. PI doesn't.

The task 13 CU “Optimize a slow SQL query”
Senior with AI tooling
0.5 MD
actual MD
26.0
PI
Junior without AI
10 MD
actual MD
1.3
PI

Same value delivered — 20× different efficiency.

Why PI survives the AI era

CU stays constant as tools improve while actual MD shrinks — so PI rises exactly when your org genuinely speeds up. Story-point velocity hides this: the same task quietly gets fewer points, and the gain vanishes from the chart.

PI reference bands
BASELINE
1.0
your org's own average
GOOD
1.5–2.0
consistently above baseline
HIGH
2.0–3.0
sustained top efficiency
EXCEPTIONAL
>3.0
rare — worth a closer look
Value models

Value depends on how you sell.

The value layer follows the commercial model of each project — the same delivery data, four ways it turns into money.

Time & Material

Value = billable hours × client rate. The invoice is the value — DevEval attributes it to the people who logged the hours.

Example
150 billable hours × 150 PLN = 22 500 PLN.

Fixed Price

Project budget split by CU delivered, not hours logged. Equal CU = equal share, even at different speeds — rewards value, not sitting time.

Example
100 000 PLN project, a developer delivered 60 of 200 CU → 30% → 30 000 PLN attributed.

Retainer

Fixed monthly base for the committed allocation, plus T&M-style value for anything delivered over it.

Internal / product

Effort-based proxy — MD × internal value per MD — when there's no invoice to point at. Cost centers get comparable numbers too.

Honest by construction

Numbers you can defend in a board meeting.

An ROI figure is only useful if you can trust it — and challenge it. Every number on this layer is labeled, normalizable, and traceable.

A bad quote isn't a bad team
On a badly-estimated project every developer looks unprofitable. DevEval also shows normalized ROI — each developer against the project's own average efficiency — with a raw/normalized toggle, so a bad quote doesn't read as a bad team.
Effort sources are labeled, not laundered
Every ROI figure shows what it's based on — e.g. “65% AI-estimated, 20% imported estimates, 15% worklogs” — with a confidence indicator.
Every number drills down
Click any ROI, value, or cost figure and land on the pull requests and worklogs behind it. No black boxes between the dashboard and the diff.

Know which projects make money — this quarter, not next year.

Connect your repos and worklogs and watch the five layers fill in from real delivery data — per developer, per project, per client.

Start free trial
30 days · no card · cancel anytime
Private beta · early access

Self-service launches soon

We're prioritising enterprise rollouts right now. Leave your email and we'll let you know the moment self-service is open.

We'll only use this to reach out about DevEval. No newsletter, no sharing with third parties.