If you originate mortgages or you run lead-gen for someone who does, you already know the funnel that paid for the last decade of refi outbound just stopped working. H.R. 2808, the Homebuyers Privacy Protection Act, was signed into law in September 2025 and the operative restriction took effect March 5, 2026. The bureaus can still generate prescreened consumer reports from a mortgage inquiry. They can no longer hand those reports to a third party who isn't already in a defined relationship with the consumer. That single sentence is the whole story, and it's the reason every independent trigger-lead aggregator in the country watched their best channel go to zero in a quarter.
I'm writing this from the data side of the table. We run public-record property scoring for lenders who pivoted off bureau leads, and over the last 60 days we've had the same conversation with maybe 40 LOs and broker shops. The pattern is consistent enough to be worth writing down.
What HBPPA actually changed
The Fair Credit Reporting Act has always allowed prescreened offers of credit. A lender pulls a credit report, the bureau identifies other consumers who meet the credit criteria, and a list ships out under FCRA's prescreening exception. That was the legal frame trigger leads lived under for twenty years. The mortgage trigger-lead variant worked off the inquiry signal: the moment a consumer applied for a mortgage, the bureau detected the hard pull, packaged the consumer record (name, phone, partial credit profile, address, sometimes equity-derived overlays), and resold it to anywhere from 6 to 30 competing lenders inside 24 hours.
HBPPA amended FCRA section 604(c) to gate that resale. Post-HBPPA, a consumer reporting agency can only furnish a prescreened mortgage consumer report to a third party if the recipient meets one of four conditions. The recipient must have originated the consumer's current mortgage, be the current servicer, be an insured depository institution or credit union with an existing account relationship, or have a documented written opt-in from the consumer. Everyone else is locked out by statute, not by industry guideline.
The carve-outs read narrow on purpose. The National Association of Mortgage Brokers spent two years lobbying for this exact language because the carve-outs preserve relationship-based competition (your existing bank, your current servicer) while closing the open-market resale lane that funded the cold-dial industry. The opt-in carve-out is theoretical for outbound purposes. No consumer has ever clicked a checkbox saying "yes, please sell my mortgage inquiry to twenty strangers." The lane is closed.
Voices from the floor
What practitioners are saying right now
It will change from 50 calls literally in minutes to 50 calls over several weeks.
It has been a channel with a great deal of fallout: angry customers, a potential for more complaints, a race to bottom for pricing.
That garbage stain on our industry is about to go bye-bye. Can’t wait to see trigger-dependent LOs get their well-deserved karmic justice.
What broke
The dominant pre-HBPPA refi playbook had three layers and all three depended on bureau resale being legal. Layer one was the data layer: an independent aggregator paid the bureau roughly $5 to $15 per raw trigger record, often on a real-time API feed. Layer two was the worked-lead layer: that record got phone-appended, scrubbed against DNC, sometimes voice-verified, sometimes routed through a qualifying call center, and resold to lenders at $80 to $250 each. Layer three was the dialer floor: a refi shop bought 300 to 1,500 of those worked leads per week, drove them through a scripted "your existing rate is X, today's rate is Y" outbound script, and converted at the well-worked-lead industry benchmark of 3 to 8 percent application-to-funded.
The economics worked because the credit-pull signal was a real intent signal. A consumer who just applied for a mortgage is, by definition, in market. Connect rates on bureau-sourced lists ran 18 to 28 percent because the timing was right. The cost per funded loan came out somewhere between $1,800 and $4,500 depending on the shop, which was tolerable against an average refi gross of $4,000 to $9,000 per closed loan.
All three layers depended on the resale being legal. When the resale stopped on March 5, layer one had no product to sell, layer two had no input to enrich, and layer three had no list to dial. Several large trigger-lead aggregators publicly pivoted to "property-data lead products" inside Q1 2026. LeadPops, which tracks the broader mortgage lead market, reported the shift candidly: the bureau lane closed, the property-data lane filled the gap, and nobody in the industry expects the bureau lane to come back. The MBA's October 2025 forecast projected $737 billion of refi origination volume for 2026. Most of that pipeline will be sourced off public-record property data this year, not bureau trigger leads. That's a structural channel change, not a cyclical one.
What didn't break
Public-record property data is untouched. County recorder filings, assessor tax rolls, MLS-derived sale and listing history, deed-of-trust recordings, lien recordings, court records (divorce, probate, bankruptcy, foreclosure), and the blended automated-valuation layer derived from those sources are all outside the FCRA scope. HBPPA amended FCRA. FCRA governs consumer reports, which are produced by consumer reporting agencies under a specific statutory definition. County records are not consumer reports. They are public records published by the county. ATTOM, CoreLogic, and the smaller property-data shops aren't consumer reporting agencies for these data feeds. The lane stays open, and the lane is what mature operators are already running on.
What this means operationally: the equity computation, the refi-fitness scoring, the rate-strike flag, the lien-position read, and the why-now recency signal can all still be derived. They just have to be derived from property data instead of bought pre-derived inside a bureau trigger lead. That derivation is non-trivial. The bureau used to do it for you and charge you $80 to $250 per record for the convenience. Now you do it yourself or you pay someone with the engine to do it for you.
Two more things didn't break. First, the existing-customer carve-out. If you're a depository or a servicer, your own portfolio is still legally targetable for refi outbound and you can still pull soft inquiries through the bureau for your own book. Second, the prescreened-offer machinery for credit cards, auto, and personal lines is unaffected. HBPPA is mortgage-specific. If your shop runs cross-product, the mortgage lane is the only one that changed.
The replacement playbook
The post-HBPPA refi outbound funnel runs on three inputs and one engine. The inputs are a property-data list (ATTOM, PropertyRadar, BatchLeads, CoreLogic, or county-recorder pulls), today's rate strip (Freddie PMMS or Mortgage News Daily), and the borrower's amortized balance computed from the original loan amount, origination date, and back-derived note rate. The engine turns those three into a ranked outbound list with five working columns: refi strike-rate flag, current LTV, cash-out capacity, why-now recency, and a DNC-scrubbed phone with channel preference.
The strike-rate flag is the column every LO asks for first. It's the answer to "is this borrower's note rate above today's market by enough to justify a call." The standard threshold is 75 basis points, which is where the rate-and-term refi math typically clears closing costs inside 24 months. A 2022-vintage origination at 6.50% against a 2026 market of 6.45% is not a today refi. A 2023-vintage origination at 7.25% against the same market is. The flag is computable from public data alone, no credit pull required, because the original loan amount and the recorded deed date are both on the recorded deed of trust at the county. Freddie PMMS gives you the rate for that origination week. Subtract today's rate. Compare the delta to 75 bps. That's the flag.
LTV and cash-out capacity come from a blended AVM (a three-source midpoint usually works: Zillow Zestimate, Redfin Estimate, and the zip-level Zillow ZHVI applied to the recorded sale price). Subtract the amortized balance from the blended value. Divide by the blended value. That's current LTV. If LTV is under 80%, conventional cash-out is mechanically available. Under 90% for VA cash-out. The capacity number is the maximum new loan minus the existing balance minus closing costs. Most LOs are still doing this in Excel one row at a time. A property-data engine does it for 200 rows in the time it takes to brew coffee.
The why-now column is the one that replaces the "this borrower just got pulled" signal trigger leads used to carry. The replacement signals are public-record events: a new permit pulled at the parcel, a divorce filing on the owner, a probate filing, a recent listing pulled (signals the homeowner explored selling and didn't), a 5-to-7-year hold-ripeness window, a recent loan modification recorded, or a new junior lien recorded. Each of these is a why-now sentence in plain English, sourced from a specific public record with a date. Combined with the strike-rate flag, this is your dialer-floor priority order. The ICP this scales for is the LO running 200-plus property pulls per week, the broker shop running 500-plus, or the depository running its servicing book.
One workflow note. The output that matters is the CSV, not the dashboard. The LO doesn't want to log into another tool. They want a ranked file that drops into Five9, RingCentral, HubSpot, or whatever dialer they already have. One row per property. Sorted by call priority. Phone, status, equity, LTV, refi verdict, opening line, lien check, all in named columns. That's the file we ship and that's the file replacing the bureau worked-lead drop.
What a deep-dive actually looks like
Here is the math layer worked out for a real (redacted) example. Subject: a 2022-vintage VA purchase in Cobb County, GA. Original loan $357,538 recorded October 2022. Borrower a veteran owner-occupant. Public-record only, no credit pull.
| Field | Value | Source |
|---|---|---|
| Original loan amount | $357,538 | Recorded deed of trust, Cobb County |
| Recorded date | Oct 2, 2022 | Recorded deed of trust |
| Estimated note rate | 6.50% | Freddie PMMS week of Sep 29, 2022 (6.70%) minus typical VA 15-25 bps |
| Amortized balance (43 mo elapsed) | $343,000 | Standard amortization on inputs above |
| Blended current value | $377,000 | Zillow ($381.4K), Redfin ($372.6K), ZHVI-adjusted sale ($376.8K) midpoint |
| Current LTV | 91.0% | Computed: 343,000 / 377,000 |
| Equity dollars | $34,000 | 377,000 - 343,000 |
| Today's VA 30-yr | ~6.50% | Mortgage News Daily / Freddie PMMS |
| Strike-rate delta | ~0 bps | Below 75 bps threshold, watch list not today |
| Why-now recency score | 9/10 cold | Cobb recorder: no refi, no second, no HELOC, no junior since Oct 2022 origination |
That row took the engine four minutes to assemble end-to-end. The verdict on this property is "watch list, IRRRL becomes profitable on a sustained 50-75 bps drop, nurture lane until then." That's a defensible call you can make from public data alone, with every cell sourced. The same row done by an LO in Excel takes 12 to 18 minutes, skips the recorder check most of the time, and never has an audit trail. The full sample output walks through the eight sections of a complete deep-dive for one property and the ranked-CSV format for a 200-row pull.
What to actually do this week
If you're a loan officer or you run a refi shop, three concrete steps are worth running through before the next pipeline review.
- Audit your current top-of-funnel source. Pull the last 90 days of leads by source and tag which were bureau-sourced (directly or laundered through a worked-lead vendor). If that bucket is more than 30% of your funded volume, you have a structural revenue gap forming. Replace it deliberately, not by buying more of whatever your aggregator pivoted to.
- Test a property-data scoring pull against your top-of-funnel cohort. Pick 200 properties from your existing geo and run them through a refi-fitness engine. Compare the ranked output to your senior LO's gut rankings of the same 200. If the engine surfaces the same top-quintile your senior would surface, the workflow is calibrated. If it doesn't, the data feed or the thresholds need work. Either way, you learn fast.
- Scrub DNC and tighten phone-channel routing. Post-HBPPA, your top-of-funnel cost just moved up, which means every reachable phone matters more. Run the list through DNC first and segment by best-channel (cell vs landline vs VoIP) before the dialer floor gets near it. The connect-rate uplift here is real and it's where the post-HBPPA economics actually clear.
One more. If you're still uploading raw, unmapped property pulls, the cleanup tax compounds. The file-prep guide covers the column shape that lets a scoring engine ingest cleanly. Five minutes of prep saves an hour of cleanup at the engine end.
Closed beta, first deep-dive free
We built the engine that runs all of the above on public-record property data. It's in closed beta with lender and broker shops right now, priced as a subscription that scales with weekly pull volume. If you read this far and you're running a refi book in the post-HBPPA world, request a beta seat at /real-estate-lead-scoring/ and the first deep-dive is free. One property, the full eight-section workup, defensible cell by cell. If the math holds up against what your senior LO knows about that parcel, the engine is calibrated for your geo and we can talk volume.
Written 11pm-ish, sent the way I'd Slack-DM it. If anything in here is wrong, mail me back and tell me which row.
Evo
Skip Trace Depot
Score your first 200 properties this week
Refi strike-rate, LTV, why-now recency, ranked CSV. Closed beta, first deep-dive free.
Request beta seat →