Frequently asked questions

DevGhost analyzes Git history and changes to estimate each developer’s code contribution. It does not track working time, capture screens or log keystrokes. Hours express effort on a common reference scale.

The time a mid-level developer with 3–4 years of experience, familiar with the codebase and working without AI, would need to create the change. This includes understanding relevant code, designing the solution, implementation, self-testing, and necessary corrections. Self-testing includes appropriate tests, interpreting their results, necessary manual behavior checks, and a final review of the developer’s own change. The estimate does not reconstruct the author’s actual hours or replace an audit or expert review.

DevGhost combines machine learning and its own neural network with structural analysis of Git changes and specialized algorithms. It considers changes in the context of existing code, accounts for reuse, and separately handles recognized mechanical changes and generated artifacts. Substantive AI-assisted code is assessed on the same reference scale. DevGhost does not use LLMs to analyze code.

The method combines engineering experience, structural analysis of changes, and training DevGhost’s own models. The distribution of estimates over time is based on empirical research and practical measurements. DevGhost does not use LLMs to analyze code.

AI-assisted substantive code is evaluated against the same developer reference without AI. If AI helps a developer deliver more estimated output per day, Ghost% can rise. Above 100% means above-baseline output, not proof that AI was used; experience, tools, Share and the selected day basis also affect the comparison.

An estimate of individual code productivity relative to a mid-level developer who knows the project and works without AI. At the default norm, Ghost% = E ÷ (D × 3 × S) × 100: E is reference hours, D is the selected Spread or FTE day count, S is Share ÷ 100. 100% matches the reference; higher is more estimated output, lower is less. Team summaries weight individual percentages by effort. This is not actual working time or a measure of a person’s entire professional value.

Use DevGhost’s common reference scale to examine sustained contribution, recognize strong developers and compare periods. Then review the underlying changes and role context. The estimate is not actual working time, a code quality score or a complete assessment of the person, and should not determine a personnel decision on its own.

The method evaluates substance and difficulty rather than rewarding commit count or lines. However, splitting, squashing and regrouping changes can alter the context seen by the estimator; we do not guarantee an invariant result. Review the underlying changes and sustained patterns, and avoid turning the number into a target people optimize or the sole basis for a personnel decision.

DevGhost assesses the contribution visible in code. A person’s responsibilities may also include mentoring, reviews of other people’s work and other activities outside the analyzed changes. Check the period, Spread or FTE mode, Share and the changes themselves before interpreting a difference. Designing, checking and fixing the analyzed change are already included in reference effort.

Coding Share is a manual allocation from 0% to 100%, default 100%. 100% means a normal full-time development role: the 3-reference-hour baseline already allows for ordinary non-coding work. Reduce Share for a reduced allocation, such as part-time work or additional responsibilities, without deducting ordinary non-coding time again. Designing, implementing, checking and fixing the change are already included in effort. Adjusted Ghost% equals raw Ghost% divided by Share ÷ 100; effort, dates, cost and raw Ghost% stay unchanged. For example, 30 reference hours over 10 days give 100% at Share = 100% and 200% at 50%. At 0%, adjusted Ghost% is unavailable, not zero; an unresolved Share also makes it unavailable.

Use it as one input to an independent human assessment of each developer’s productivity, including sustained low contribution and strong results. Read it across periods with the developer’s responsibilities, code quality, business impact and work outside code. Ghost% must not be the sole or primary basis for reviews, pay, promotion or dismissal, and cannot replace that contextual assessment.

Yes. Every account starts with a batch of credits and gets a fresh free allowance every month — no card required and no feature limits: the free plan is the full product. One credit covers one analyzed commit, and a squash-merged pull request counts as half its internal commits. The exact cost of a run is shown before it starts.

Cost is roughly what the delivered work cost at a standard rate; value is roughly what it would cost to reproduce that volume by hand, without AI. The gap between them is an approximate indicator of leverage (tooling/AI), not a profit-and-loss statement.