NJ surplus funds county land records scraper

i am trying to build is essentially a New Jersey sheriff-sale surplus funds lead qualification and discovery system that works county by county and relies on the actual recorded deed images as the source of truth. ( note for image scraping i do not want to pay anything using a third party source ) There are two separate workflows you want the system to handle. 1. Qualify your existing ATTOM leads in a csv file i already have CSV file containing potential leads, generally with the former owner's name and other property information. with i think around 70,000 names and a lot being duplicates For each person/property on that list, the system should: Determine the appropriate New Jersey county. Open that county's land-records website. Search the former owner's name. Find the corresponding Sheriff Deed / SHER DEED. Open the actual scanned deed. Read the deed and extract: Previous owner / defendant Property address Sheriff sale date Final sale/purchase price Foreclosure debt/judgment amount Purchaser/new owner Instrument/document number where useful Calculate: Estimated surplus = Final sale price − debt Determine whether the purchaser is an acceptable third-party bidder. Mark the ATTOM record as either Qualified or Not Qualified. A record qualifies only when both of these are true: Sale price − debt ≥ $35,000 The purchaser is a third party, meaning it is not the foreclosing bank, mortgage lender, mortgage servicer, or creditor. Individuals, investors, and ordinary/non-lender LLCs can qualify. Importantly, the qualification information needs to come from the actual recorded deed/document images, rather than relying on ATTOM, an index search result, or the land-record metadata. part 2 of this system is different . Generate completely new/raw leads from county records This is separate from my ATTOM list. Instead of searching an owner's name, the system searches the county database itself for Sheriff Deeds, starting from approximately 2019 through the present. For Camden County ( https://camden.newvisionsystems.com/SearchAnywhere) , for example, searching the deed type can produce tens of thousands of records. The system would need to process those records in batches, open the underlying documents, and apply the exact same qualification rules. So the pipeline would essentially be: County records → Sheriff deeds → Open document → Extract values → Calculate surplus → Check purchaser → Export qualifying lead Because Camden showed 95,744 results and only the first 1,000 are displayed, the automation cannot simply run one unrestricted search. It would need to divide the search into smaller ranges, most logically by recording/sale date, and process the results batch by batch so that no records are lost because of the site's 1,000-result display limit. Exact qualification rule A lead is: QUALIFIED Final sale price − stated foreclosure debt ≥ $35,000 AND Purchaser is a third party real Example camden county land records name, ( finley, richard ) Sale price: $194,000 Debt: $117,923.14 Difference: $76,076.86 Buyer: DJM Properties Buyer is not shown as the foreclosing lender QUALIFIED That is exactly what your Richard Finley / Camden County example demonstrates. The recorded sheriff deed identifies the purchaser as DJM Properties, and the deed gives a true consideration of $194,000. The writ/deed materials state approximately $117,923.14 as the amount to be satisfied. Therefore: $194,000 − $117,923.14 = $76,076.86 That exceeds your $35,000 threshold by a substantial amount. NOT QUALIFIED For example: Sale: $180,000 Debt: $160,000 Difference: $20,000 Not qualified, even if an investor purchased it. Or: Sale: $250,000 Debt: $150,000 Difference: $100,000 Buyer: Bank of America Not qualified, because the creditor/lender acquired the property. What should be exported For every qualified record, I would make the final dataset contain at least: Field Purpose County Camden, Essex, etc. Previous Owner Person potentially entitled to surplus Property Address Foreclosed property Sale Date Actual sheriff sale date Recording Date Useful separately from sale date Final Sale Price Amount property sold for Debt Amount Amount stated in recorded foreclosure documents Calculated Difference Sale − debt Purchaser Winning bidder/new owner Purchaser Type Individual / LLC / lender / etc. Third-Party Buyer? Yes/No Qualified? Yes/No Instrument Number Allows deed to be found again Book/Page If available Document Link/ID Allows manual verification Source County County record system Review Status Automated / manually reviewed One thing I would specifically keep separate is sale date versus recording date. For example, Camden's index may show when the sheriff deed was recorded, but you want the actual sheriff-sale date contained in the document whenever it is available. What the automation actually needs to do Conceptually, the system has five components: County website automation → document retrieval → document reading/OCR → qualification logic → CSV/database export The difficult part isn't the $35,000 calculation. That is straightforward. The more complicated part is reliably navigating the different county websites and correctly interpreting the scanned deeds. For each county, the automation will need its own configuration because New Jersey counties don't necessarily use identical land-record systems. Some will be public without authentication, while others may require credentials that you provide. It should also recognize that wording varies. For example, the sale amount might appear as: "true consideration for this conveyance ... is $194,000.00" while the debt might appear as: "the sum of $117,923.14 being the principal, interest and advances..." The scraper therefore shouldn't depend on one exact phrase.