Opportunities › BM25 full-text search for Postgres databases › Proposed app
Rankbench
Tune and test keyword plus vector search on your own Postgres database
- Score
- 61
- confidence 67
- Evidence
- 13
- items on the page
- Rivals
- 2
- named in sources
- First version
- 35 d
- estimate
- Type
- Desktop app
- PWA: runs in the browser, installs on Mac and Windows
- Market since
- 2023
- Emerging market · AI estimate
ScreensConcept
Desktop app: Engineering teams tune search rankings by comparing dense performance data and adjusting parameters at a desk over sustained work sessions. Sample screens on a computer, in the proposed brand. Their numbers are sample data.
Compare
Weights 60 / 40hybrid
About this appAI proposal
Rankbench is a workbench for teams who run search inside Postgres. Connect a read-only role, pick a table, and see how keyword (BM25-style) ranking, vector similarity and a blend of both order your real rows for the same query.
Save queries with the results you expect, change weights and filters, and see at once which queries got better or worse. Share a link to a comparison so a teammate can review it without a database login.
It does not replace your search extension or your database. It works with whatever ranking functions your Postgres already exposes, so you can test open-source options against each other and avoid lock-in to one engine.
- Tagline
- Know why your search ranks what it ranks.
- Voice
- PlainPreciseCalm
- Why this name
- A bench is where you test parts side by side, and rank is the thing being tested. Two plain words joined, easy to say and spell.
In the first version
Members say which features they would use. It tells PWAS what people want; it changes no score.
Side-by-side ranking
Run one query as keyword, vector and hybrid and see the three result lists next to each other, with the score behind each row.
No member yet
Expected-result sets
Save queries with the documents that should appear. Rankbench scores each run against them so tuning is measured, not guessed.
No member yet
Hybrid weight sliders
Change the keyword and vector balance, boosts and filters, then replay your whole saved set and see what moved.
No member yet
Before and after diff
Compare two configurations or two extensions on the same data. Wins, losses and ties are listed per query.
No member yet
Brand identityAI proposal
Check the name against trademark registers before using it.
Palette
#12363ADeep Spruce · primary#F5F2EAPaper · surface#16232AInk · text#F2A33ASignal Amber · accent#3E8E96Lake · data#D8D3C5Stone · border
Typography
Aa
Rankbench
PricingAI proposal
The material gives no competitor prices: ParadeDB's pricing page lists a free plan and prices on request. These tiers are a starting point to test, not a market claim.
Free
Free
- 1 database connection
- 1 saved test set of 25 queries
- Side-by-side compare
Team
$49 a month
- 5 connections
- Unlimited test sets
- Before and after diff
- Shareable review links
- Column masking
Business
$199 a month
- Unlimited connections
- Scheduled re-runs with alerts
- SSO
- Audit log
Would you pay Team · $49 a month?
Your answer tells PWAS which prices people accept. It changes no score.
No answers yet.
What it could be worthEstimate
What a software business making this much sells for today, from real deals. Pick a monthly revenue.
If it makes
Worth about
$190,000
Sale range $132,000 to $263,000
$120,000 a year in revenue
From 87 software businesses listed and sold on Empire Flippers in the last year, revenue vetted by the broker: they sell for about 19× their monthly revenue, most between 13× and 26×.
Measured on profit, the middle deal is 32× monthly net profit: about 2.7 years of profit.
An estimate, not an offer or financial advice. A real price depends on profit, growth, churn, how much of the work is the owner's, and the buyer.
How it would growAI proposal
How the product would reach new customers through its own use, and how the first ones would be found. A plan to test, not a result.
- Built-in loop
- Shared outputs. Users send read-only comparison links to teammates, vendors or in open-source issues and blog posts. The viewer sees ranked results with the Rankbench name and a link to try it.
- Do they become customers?
- Viewers are mostly engineers who run search too, so some could sign up. Test how many open a shared link and then connect their own database.
- First customers
- Backend engineers at small SaaS companies of 10 to 50 people who run Postgres and are adding hybrid search to an AI feature.
- Where to reach them
- Hacker News and Postgres forums: Post a worked comparison of keyword, vector and hybrid on a public dataset, with a link to try it.
- Open-source extension communities: Offer the compare view to maintainers and users who share benchmark questions in issues and chats.
- First milestone
- Five teams run a saved test set on their own database, and two of them pay for the Team tier within 60 days.
Why people already pay for this
- VerifiedHacker News post received 203 points and 56 comments
- Public claimPostgres lacks native BM25 search capability, forcing teams to choose between limited built-in search or external solutions
- Public claimBuilding a competitive BM25 extension from scratch requires substantial engineering effort
- Public claimTeams need hybrid search (keyword + semantic) to serve emerging AI-centric workloads
- Public claim
Tech stack
Every proposal builds on the same PWA base, installable from the browser on any phone or computer. Dashed boxes are what this app adds.
- AppNext.js and TypeScript
- PWAManifest and service worker
- UITailwind CSS
- Data and authSupabase Postgres
- PaymentsStripe
- EmailResend
- HostingVercel
- This app · Database connectorServer-side Postgres client with a connection proxyRuns read-only queries against customer databases from a server, since a browser cannot reach them safely.
- This app · Job runnerBackground queue for test-set replaysReplaying 50 to 500 queries and timing them takes longer than a web request.
RisksAI inference
- CommercialTeams may tune search once and stop. Recurring value must come from re-runs after data or index changes, and that needs testing.
- TechnicalEach ranking extension exposes different functions and score scales. Adapters for each need upkeep, and scores are hard to compare across engines.
- DataCustomers must trust a hosted tool with a connection to production data. Many will want a self-hosted runner or a copy of their data.
- CommercialThe extension makers or Postgres vendors may add their own tuning tools, leaving little room for a separate product.
- PlatformThe evidence shows interest in the extension itself, not in a tuning tool. Demand for this layer is unproven.
Everything marked AI proposal or AI inference was written by a model from the evidence on the opportunity page. The proposed name must be checked against trademark registers before it is used.