From a brittle UiPath snapshot to a sheriff-sale database you can actually trust
Right now, Sheriff Sale Genie is a bot that breaks every time Zillow renames a field, and a spreadsheet that's only true the moment you run it. Phase 1 turns that into a real plan: a documented pipeline, a working collector pulling live data into a real database, and a level-of-effort estimate for the full product — so you know exactly what this costs before you commit further.
What we heard on our call
- The current system is a snapshot, not a database. Your UiPath bots scrape the county sheriff-sale site, enrich it with Zillow and Realtor.com, and produce a same-day spreadsheet — nothing persists, and nothing updates itself.
- It breaks on a letter. When Zillow renames a field or selector, the whole pipeline goes down, and finding and fixing it has cost more than the original build.
- You want it updated roughly daily, matching how often the county records actually change, and stored once rather than re-scraped every time.
- You want the comparison view, not a single lookup. You're looking at every property in a sale at a glance — the bank's upset price next to Zillow's estimate — to spot the best margin, not searching for one address.
- Sales are small and predictable. Every other Wednesday, typically under 20 properties, public records with no login required.
- You're open to paying for reliability — Zillow's official API or a licensed MLS/IDX aggregator — if it means the product stops breaking.
- This is a feasibility study for you too. You want real costs and a clear go/no-go before committing further, and you liked the idea of a working collector as proof, not just a plan.
- The market may be bigger than investors. New Jersey's owner-occupant priority program at sheriff sales opens this up to realtors and buyers too, not just investors like yourself.
The plan
Three things have to happen before anyone writes a line of production code: understand exactly what the current bots do, confirm what's realistic to collect and at what cost, and prove it on one real source.
Document & map
Every RPA activity becomes a row in a dependency table — source, selector, field, transform, output — starting with salesweb.civilview.com, the sheriff-sale listing site you sent us.
Price the alternatives
Zillow's official API, a licensed MLS/IDX aggregator, and the free public-records path — each priced out so the on-demand, scrape-only-what-changed model stays cheap as you scale.
Prove it
One working collector, built in your own repository, pulling real sheriff-sale records into a versioned Postgres database — never overwritten, a new dated row per change.
What's included
- Workflow dependency table — every source, field and transform your current UiPath bots touch, documented as a maintainable spec.
- Source-by-source legal/data report — what's safely public, what needs a provider agreement, and the per-source terms-of-use review this kind of resale needs.
- Architecture & stack recommendation — Python collectors, Postgres, a TypeScript API, and a Next.js + Stripe customer platform for the subscription model you described.
- Initial database schema, built for versioned history, not overwritten snapshots.
- Retention, failure-detection & county-expansion plan — so the next county added doesn't mean starting over.
- Risk register covering the data-licensing and site-fragility risks flagged below.
- Phase 2/3 estimate for the full pipeline and customer platform, plus a monthly cost calculator so you can see exactly how Zillow-API-vs-aggregator costs scale with usage.
- One working collector, built end-to-end in your own repository against the sheriff-sale site — live proof, not a slide deck.
- Handover document and a walkthrough call.
A first look
A concept of the comparison view you described — every upcoming sale side by side, upset price against Zillow's estimate, best margin first. Concept only, with sample numbers — the real view is shaped once Phase 1 confirms what data we can actually pull.
Sample data for illustration only — not real sheriff-sale listings.
What we checked before quoting
We did an initial pass on feasibility before putting a number on this, so there are no surprises mid-project.
UiPath → Python/Postgres rebuild
This is a normal rebuild pattern. Documenting the existing RPA steps as a dependency table is the standard way to turn an automation script into a maintainable pipeline.
No statewide sheriff-sale API
Each NJ county runs its own listing page, often through a shared platform (SalesWeb/CivilView). That means a collector per county, with upkeep whenever a county's site layout changes.
Statewide parcel data, one source
NJ publishes statewide parcel/MOD-IV data through official ArcGIS feeds, with a Rutgers-hosted historical database — likely faster and more reliable than scraping property details county by county.
Reselling public-records data
Scraping genuinely public pages is generally lawful, but each source's own terms of use still apply. This needs real per-source legal review time before scaling resale — budgeted into Phase 1, not waved away.
Milestones and pricing
Dependency table & source mapping
- Document the existing UiPath workflow field-by-field
- Research salesweb.civilview.com for scraping complexity
- Zillow API and MLS/IDX aggregator pricing memo
Architecture, schema & legal review
- Stack recommendation: Python collectors, Postgres, TypeScript API, Next.js + Stripe
- Initial database schema (versioned, never overwritten)
- Source-by-source legal/data report
Working collector
- One collector built end-to-end in your own repository
- Pulls live sheriff-sale records into the database
- Retention & failure-detection plan
Risk register, cost calculator & handover
- Risk register and county-expansion plan
- Phase 2/3 level-of-effort estimate + monthly cost calculator
- Handover document and walkthrough call
Ad hoc work outside this scope is billed at $40/hr. Phase 2 (pipeline) and Phase 3 (customer platform) are priced as a $25,000–$45,000 range today, fixed exactly once the Phase 1 report is in hand.
After Phase 1
Phase 1 is a feasibility study on purpose — it's built so you can stop here with real answers, or move straight into the build with a number you trust.
| Path | What it covers | Terms |
|---|---|---|
| Stop here | You keep the dependency table, legal report, architecture plan and working collector either way | No further cost |
| Phase 2 – Pipeline | Full per-county collectors, versioned database, scheduled refresh | Fixed after Phase 1 |
| Phase 3 – Customer platform | Next.js + Stripe subscription site, buy-once or recurring plans | Fixed after Phase 1 |
| Ongoing maintenance | Monitoring, fixes when a county site changes, data provider upkeep | ~4–8 hrs/mo + provider fees |
When is it done?
What we need from you
- The sheriff-sale site — already sent: salesweb.civilview.com
- Priority NJ counties and which data fields matter most to your subscribers
- Today's data sources — whether you already pay for any licensed feed, or everything is free/public today
- Subscriber pay model — plain listings vs. enriched with valuations, one-time vs. subscription
- Who signs off on the Phase 1 report — just you, or other reviewers/investors
- A short weekly check-in so questions don't sit for days
Relevant work
A few of our recent data- and workflow-driven builds: Accio Data, NextPlay Nutrition, Raise Forward, ServicePost, and Junior Fit Golf — spanning scraping/automation rebuilds, subscription platforms, and mobile apps.
How we work
- One point of contact. I lead the project and stay on every call; our collector/backend engineers and QA are based in Pakistan, with US-hours overlap for critical periods like launch.
- Full visibility. An Asana board with every task, so you can see progress any day — not just at the weekly call.
- Proof over promises. Milestone 3 ships a real collector against real data, not a slide about one.
- Pay as it's delivered. Fixed-price milestones on Upwork, released only when you've reviewed and approved each one.
Next steps
- Confirm priority counties and today's data sources so Milestone 1 starts scoped correctly.
- Kick off Phase 1 on Upwork — Milestone 1 lands within a week.
- Review the Phase 1 report together and decide, with real numbers, whether to move into Phase 2/3.