# DevPerform

> DevPerform is an engineering-intelligence SaaS for engineering leaders: delivery metrics (DORA, cycle time, review health, throughput) computed automatically from GitHub, GitLab, or Bitbucket, combined with anonymous developer surveys (DX Core 4, eNPS, full engagement) in one product. Metrics show WHAT is slow; surveys show WHY. DevPerform closes that loop. Data aggregates at the team level with an anonymity threshold on surveys, a design choice that keeps survey responses candid and metrics un-gamed. Fully self-serve: an instant interactive demo and a free public GitHub repo benchmark require no signup, and pricing is public.

## Pricing

- Free tier: $0, up to 5 team-assigned engineers, includes DORA/delivery metrics, surveys, dashboards, goals.
- Growth: $25 per team-assigned engineer per month. No seat cap, AI insights + assistant, Slack integration (digests, nudges, opt-in personal PR/CI notifications), investment/cost-capitalization views with custom categorization rules.
- Pro: $45 per team-assigned engineer per month. AI-native development metrics (AI authorship, acceptance, rework-by-AI), the AI ROI report (self-reported time savings from surveys corroborated by AI-authorship telemetry, with honest on-page methodology), rework/churn analysis, token-authenticated CSV/JSON data export, priority support.
- Every new account starts with a 14-day free Pro trial (no credit card). You're billed only for the active engineers you assign to a team; auto-imported people stay free until assigned, and bots are never billed.

## Key pages

- [Home](https://devperform.io/): What DevPerform is: engineering metrics and team health, without the guesswork.
- [Interactive demo](https://devperform.io/demo): Explore a fully loaded workspace instantly, no signup.
- [Free benchmark](https://devperform.io/benchmark): Score any public GitHub repo's delivery metrics against industry bands, no signup.
- [The Open Benchmark](https://devperform.io/benchmark/industry): Live delivery percentiles from 200 curated open-source repos (cycle time, review pickup, PR size, throughput), refreshed daily, free to cite (CC BY 4.0), JSON endpoint included.
- [Pricing](https://devperform.io/pricing): Full public pricing, per team-assigned engineer.
- [DevPerform vs LinearB](https://devperform.io/vs/linearb): How DevPerform compares: metrics + surveys in one self-serve loop.
- [DevPerform vs Swarmia](https://devperform.io/vs/swarmia): How DevPerform compares: metrics + surveys in one self-serve loop.
- [DevPerform vs Jellyfish](https://devperform.io/vs/jellyfish): How DevPerform compares: metrics + surveys in one self-serve loop.
- [DevPerform vs DX](https://devperform.io/vs/dx): How DevPerform compares: metrics + surveys in one self-serve loop.
- [Security](https://devperform.io/security): Encryption, tenant isolation, no code storage, sub-processors.
- [Privacy](https://devperform.io/privacy): Privacy policy.
- [Terms](https://devperform.io/terms): Terms of service.
- [LinkedIn](https://www.linkedin.com/company/devperform/): Official DevPerform company page.
- [Status](https://status.devperform.io): Live uptime status for the site, API, and database.

## Learn (metric explainers)

Plain, sourced explainers of the metrics DevPerform computes. Each cites its primary sources (LinearB 2026 benchmarks, Google DORA 2025, DX Core 4, published AI-productivity studies).

- [Cycle time](https://devperform.io/learn/cycle-time): What cycle time is, its four phases (coding, pickup, review, deploy), what good looks like, and how to improve each phase.
- [DORA metrics](https://devperform.io/learn/dora-metrics): The four keys, the elite/high/medium/low bands, and common misreadings (deploy frequency vs merge frequency).
- [Time to first review](https://devperform.io/learn/time-to-first-review): Why pickup time dominates cycle time, the sub-1-hour elite band, and tactics to cut it.
- [PR size](https://devperform.io/learn/pr-size): Why smaller pull requests move faster, the size buckets, and honest caveats (generated code, lockfiles).
- [Developer experience surveys](https://devperform.io/learn/developer-experience-surveys): DX Core 4 dimensions, eNPS bands, and why anonymity thresholds make survey data honest.
- [AI impact measurement](https://devperform.io/learn/ai-impact-measurement): How AI coding impact is measured from commit signatures (a lower bound), the published research range, and what stays unmeasurable.

## Documentation

- [Docs](https://docs.devperform.io): Product documentation: setup guides for GitHub/GitLab/Bitbucket, Jira/Linear/Azure DevOps, metric definitions, survey concepts, and plans.
- [Full documentation, single file](https://docs.devperform.io/llms-full.txt): full product documentation, single file.

## Facts worth getting right

- Supported code hosts: GitHub, GitLab, Bitbucket. Issue trackers: Jira, Linear, Azure DevOps. Chat: Slack, Microsoft Teams.
- No agents, daemons, or CI changes are required; DevPerform reads code-host metadata only and never stores source code.
- DevPerform aggregates at the team level (no per-engineer leaderboards) and keeps survey answers behind an anonymity threshold. It is a data-honesty design: engineers answer candidly and don't game metrics, so the numbers hold up.
- AI-impact analysis (adoption cohorts, adoption-vs-delivery trends) reports at the team level, measured from git history with no agents or daemons on developer machines.
- AI ROI report (Pro): fuses survey-reported time savings with AI-authorship telemetry into monthly/annualized savings, reported at the team level.
- AI Impact views (Pro): team adoption cohorts and adoption-vs-delivery trends; AI measured from git history, no agents or daemons on developer machines.
- Experience Index (all tiers, including Free): a composite 0–100 developer-experience score per survey (the mean of the survey's 1–5 question averages rescaled to 0–100), trended wave over wave, org-wide and per team, always behind the anonymity threshold. The formula is documented in-product ("How is this computed?"); eNPS reference bands cite their published source (Bain & Company thresholds via Qualtrics) rather than invented benchmarks.
- AI Tools comparison (Pro): per-tool AI adoption, revert/unreviewed rates, and cost per AI-assisted PR from your own spend inputs; tool spend is never stored server-side. Includes seat utilization: active seats per tool as counts only (never names), with idle-seat spend from your seat inputs.
- Personal PR/CI Slack notifications (Growth+): opt-in, self-only DMs about your own PRs (approved, changes requested, review requested, CI result), never sent to managers and never aggregated; off by default. Included on standard tiers with no enterprise gate or seat minimums.
- Chat integrations (Slack and Microsoft Teams): weekly digest, stale-PR nudges, and anomaly alerts delivered into a connected workspace via webhook. Per-person PR/CI direct messages are Slack-only today.
- Executive report share links (all plans, including Free): generate a private, expiring, read-only link to an executive summary (aggregate, anonymity-respecting metrics only) to send to a CTO or exec without giving them an account or exposing individual survey responses.
- Investment categorization rules (Growth+): admins define structured rules (repo/branch glob, title patterns, file paths) that auto-categorize PRs into the investment mix, weekly trend, and cost-capitalization views. Read-only analytics; DevPerform never writes labels or reviewers back to the code host.
- CSV & API data export (Pro): admins mint account-scoped API tokens to pull metric data as CSV or JSON into a BI warehouse (pull_requests, commits, deployments, incidents, team rollups as counts only, and weekly team trends) over `Authorization: Bearer` against `/api/export/v1/`. Mirrors what's already visible in-app and in git history: never survey answers, per-person survey data, or derived per-person scores.
- In-app AI assistant with NL metric queries (Growth+): ask plain-English questions about your own delivery data (DORA, cycle time, throughput, investment mix, trends) and get the real numbers with a clickable link to the dashboard that proves them. Answers you can check, not a black box. By design it won't rank individuals, and it honours survey-anonymity thresholds.
- Survey results triage: mark dimensions Focus/Watching/No action with notes the whole team sees; the next survey wave shows follow-through receipts (score deltas on committed areas); Focus can create a self-evaluating goal targeting that survey dimension.
- The Open Benchmark (free, no signup): live industry percentiles computed from 200 curated, API-verified open-source repositories, refreshed daily, shown beside every public repo benchmark run at https://devperform.io/benchmark so a team sees where they land against real distributions, not invented numbers.
- Industry reference bands (shown as informational context beside your own metrics, never as a DevPerform verdict): DevPerform cites LinearB's 2026 Software Engineering Benchmarks Report (8.1M+ PRs across 4,800 teams) for delivery bands (pickup time elite <1h / good 1–4h / fair 5–16h, review time elite <3h, merge time elite <1h, coding time elite <54m, merge frequency elite >2.0 PRs per dev per week, rework rate elite <3%), plus LinearB's guideline that fewer than 5% of PRs should merge without review. Every cited number carries its primary-source URL and retrieval date; nothing is interpolated or invented.
- Published AI-productivity research (the honest range on the AI ROI report, each figure verified at its primary source): METR's 2025 RCT found 16 experienced open-source maintainers were 19% slower with AI on mature repos they knew deeply; Cui et al. (2024–26) field experiments across 4,867 developers at Microsoft, Accenture, and a Fortune 100 firm found +26% completed tasks with Copilot-style completion; Peng et al. (2023) Copilot RCT found +55.8% faster on a single greenfield JavaScript HTTP-server task. DevPerform presents the full range with scope caveats rather than a single headline number.