LockRadar

Idea

LockRadar is a GitHub-native CI check that scores every database migration pull request against the real, current state of a team's production database, not just the migration's SQL text. It connects read-only to a Postgres or MySQL replica, pulls the actual table size, row count, and current lock contention, and blocks a migration before merge if it would lock or rewrite a table that has grown far past what the migration's SQL looks safe for. It is built for growth-stage engineering teams that cannot yet justify hiring a dedicated database reliability engineer.

Market gap

Every existing migration safety tool, free or paid, reads the migration's SQL text and matches it against known-bad patterns: adding a NOT NULL column without a default, creating an index without CONCURRENTLY, changing a column type that forces a rewrite. None of them ask the one question that actually determines whether a migration is safe: what does the real production table look like right now. A migration that is instant on a 10,000-row table can lock a 40 million row table for hours.

Atlas, the most schema-aware of the static tools, paywalled its own migration linter behind a $100 a month Atlas Cloud plan in October 2025, which is itself a signal that teams already pay for this category. Bytebase, the broadest platform in the space, shows only 2 reviews on G2 despite being open source since roughly 2022, suggesting the enterprise-platform approach has not reached the smaller teams who need it most. The teams in the 20 to 500 engineer range are big enough to run schema migrations weekly and small enough that a $150,000 a year DRE hire is not yet justified. That gap is where LockRadar sits.

Total Addressable Market (TAM)

Bottom-up:

• Roughly 39,000 to 73,000 companies verified running PostgreSQL in production, and 150,000 plus running MySQL (Landbase and 6sense technology-adoption data, sources below).
• Narrowing to the addressable segment (funded or growth-stage companies, roughly 20 to 500 engineers, on Postgres or MySQL, without a dedicated DRE) is an assumption: call it 20,000 companies globally. This is the weakest link in the math and the first thing to validate against real signup data.
• Average contract value: positioned well below a DRE hire (average reported salary $156,093 a year per ZipRecruiter data) and below Atlas Cloud or Bytebase enterprise minimums. Assume a $750 a month team plan, roughly $9,000 a year average across tiers.
• TAM: 20,000 companies x $9,000 = $180 million.
• SAM (companies that have already had a migration incident, or are actively hiring for DRE-adjacent roles): roughly 30 percent of TAM, 6,000 companies x $9,000 = $54 million.
• SOM (realistic 2 to 3 year target for a small team): 150 to 400 customers x $9,000 = $1.35 million to $3.6 million ARR.

Monetization strategy

Flat, tiered SaaS subscription billed per repo or per connected database, not per engineer seat, so pricing does not require an intrusive headcount audit. A free tier ships the static rule engine only, matching what Squawk and MigrationPilot already give away, so the funnel starts where developers already are. Paid tiers unlock the live-database risk enrichment, which is the part nobody else offers. Customers keep paying because every quarter of growth makes their tables bigger and the risk of an unreviewed migration higher, not lower.

Pricing strategy

• Free: static rule linting only (GitHub Action), unlimited repos, no live-database connection. Anchors against Squawk and MigrationPilot's free tiers.
• Team, $149 a month per repo: live risk scoring against up to 3 connected databases, PR risk comments, Slack alerts, 90-day dashboard history.
• Growth, $399 a month per repo: unlimited connected databases, Postgres and MySQL in one plan, longer dashboard retention.
• Enterprise, custom: VPC or on-prem deployment, SSO, long-term audit log retention. Anchor tier, priced to compete with Atlas Cloud and Bytebase enterprise contracts.
• Entry point is the $149 Team plan: a fraction of one incident's cost and far below Bytebase's roughly $20 a user a month enterprise pricing once a team is large enough to need seats for it.

Lead magnet

A free, no-signup Migration Blast Radius Calculator: paste a migration's SQL plus a rough row count and disk size for the table it touches, and get an instant estimate of lock duration and downtime risk. This is the exact live-data-aware wedge offered as a free taste, unlike competitors' calculators which only lint the SQL text.

Social proof that the problem exists

• GitLab's own public incident tracker logs a production deploy failure caused by a migration: gitlab.com/gitlab-com/gl-infra/production/-/issues/20494. Shows even a company with a dedicated infrastructure team ships migration incidents into production.
• A widely read technical postmortem on GitHub's November 2021 outage, caused by a schema migration that deadlocked replicas: arpit.substack.com/p/dissecting-github-outage-downtime-242. Shows the failure mode at a company operating at massive table scale.
• A curated, ongoing public collection of company postmortems including multiple migration-caused incidents across different companies: github.com/danluu/post-mortems. Shows this is a catalogued, recurring pattern industry-wide, not a one-off.
• Database Reliability Engineer job descriptions that explicitly list migration review and incident-prone schema migration handling as a core responsibility: handbook.gitlab.com/job-description-library/engineering/infrastructure/database-reliability-engineer and resources.workable.com/senior-database-reliability-engineer-job-description. Shows companies pay a dedicated salary specifically to catch this before it ships.
• Atlas paywalling its migration linter behind a $100 a month Atlas Cloud plan in October 2025, and the developer community reaction to it: dev.to/mickelsamuel/atlas-paywalled-their-migration-linter-here-are-your-free-alternatives-4god. Shows real willingness to pay already exists in this exact category.

Competitors

• Squawk (squawkhq.com): free, open-source, Rust-based static linter for Postgres migrations, roughly 600,000 downloads a month. No live-database awareness, no dashboard, no paid tier.
• strong_migrations (github.com/ankane/strong_migrations): free, open-source Rails gem, static pattern matching only, Rails-specific.
• safe-pg-migrations (github.com/doctolib/safe-pg-migrations): free, open-source, built by Doctolib, static Postgres checks for Ruby migrations.
• MigrationPilot (marketplace.visualstudio.com/items?itemName=migrationpilot.migrationpilot): a GitHub Action and VS Code extension, Postgres-only, 77 static rules with lock classification and auto-fix suggestions. Newer and less established than Squawk, still purely static.
• Atlas / Ariga (atlasgo.io): schema-as-code tool that treats the whole database schema as declarative infrastructure. Paywalled its migration linter behind Atlas Cloud starting at $100 a month in October 2025. Still a static SQL-diff planner, not aware of live production table state, and requires buying into a full schema-as-code workflow.
• Bytebase (bytebase.com): a broad "database DevSecOps" platform covering change management, SQL editor, and governance. Cloud pricing around $20 a user a month, self-hosted requires an Enterprise license. Only 2 reviews on G2 despite years on the market, suggesting limited reach into smaller teams.
• PlanetScale (planetscale.com): ships genuinely safe, non-blocking schema changes, but only for databases migrated onto its own Vitess-based platform. Not a bolt-on for teams already on RDS, Cloud SQL, self-hosted Postgres or MySQL, or Supabase.

What competitors offer now

Every tool above except PlanetScale reads the migration's SQL text and checks it against a fixed list of known-dangerous patterns. That is genuinely useful and free-tier competitors have made it a commodity. None of them connect to the actual running production database to check whether the table the migration touches is 10,000 rows or 40 million rows, whether it already has replication lag, or whether there is an active long-running transaction that would turn a normally-instant lock into an hours-long stall. Atlas and Bytebase go further into workflow and governance, but both ask a team to adopt their platform as the system of record for schema, not to bolt onto whatever the team already runs. PlanetScale solves the underlying problem, but only by requiring a full platform migration.

What can be done differently to attract customers

Ship a read-only connection to a team's existing production replica (RDS, Cloud SQL, self-hosted, or Supabase, no platform migration required) that pulls real table size, row count, index bloat, and current lock contention, and combine that live data with the same static rule checks every free tool already offers. Score every migration PR against what the database actually looks like today, not what the SQL text implies in the abstract. Support Postgres and MySQL in one tool, since most existing free linters pick only one engine. Enter through a free static-only tier and a free Blast Radius Calculator so the funnel starts exactly where Squawk and MigrationPilot already sit, then convert on the live-data feature nobody else ships.