FurnishWatch

1. One-liner

Watches your books for the first material or labor charge on a job and warns you before lien rights die.

2. Trend signal — why now?

Three things moved at once.

The money at stake got documented. Rabbet’s 2024 Construction Payments Report put the cost of slow payments to US construction at $280 billion in 2024 — roughly a 14% add-on to total construction spending. In the same report, 82% of contractors now wait more than 30 days for payment, up from 49% two years earlier. Only 5% of subcontractors get paid on time per CCFG Credit. This is not a niche annoyance; it is the defining cash-flow problem of the trade.

And the legal remedy is barely used. NCS Credit counted roughly 90,000 mechanics liens filed in the trailing 12 months across an industry with hundreds of thousands of specialty trade establishments running many jobs each. Preliminary notices covered projects worth $22.7B (Lienser) — a rounding error against total US construction spend. The tool exists. Almost nobody pulls it.

Because the trigger is invisible. Here’s the mechanic that makes this a product. The preliminary notice clock does not start at contract signature, PO receipt, or project kickoff. It starts at first furnishing — the first date you deliver material or put a body on site. California is 20 days from first furnishing. Florida is 45. Utah is 20 with an electronic registry. Washington 60. Texas runs monthly notice deadlines where missing one month voids lien rights for that month’s work with no cure. Industry practitioners describe missing the preliminary notice as the single largest cause of construction-payment recovery losses nationally, and the documented failure mode is clerical, not legal — one published example describes a sub-subcontractor missing a 45-day notice because “administrative staff missed that deadline and totally forgot to send it.”

Nobody forgets on purpose. They forget because the event that started the clock — a $340 pipe delivery, one guy on site for four hours — never crossed the office manager’s desk as a legal event. It crossed as a line item in QuickBooks.

That line item is now machine-readable in real time. Accounting APIs, payroll APIs, and supplier invoice parsing are commodity in 2026. The clock’s starting gun is already sitting in the sub’s own ledger. Nobody is listening for it.

Provenance:

• Signal 1 (demand): Missed preliminary notice described as the single largest cause of construction-payment recovery losses nationally; documented clerical-failure case where staff “totally forgot to send it”; 50-state deadline variance (CA 20d, FL 45d, TX monthly, WA 60d) — https://mechanicslien.com/preliminary-notice/ and https://www.licensesetc.com/2019/11/08/forgot-to-send-notice-to-owner/ — observed 2026-08-07
• Signal 2 (feasibility): First furnishing legally defined as the first date materials arrive or labor first appears on site — not contract date — making the trigger a cost event already recorded in the sub’s accounting/payroll system, now readable via commodity APIs — https://www.ncscredit.com/education-center/blog/so-when-was-your-first-furnishing/ — observed 2026-08-07
• Signal 3 (economic): Rabbet 2024 Construction Payments Report — $280B annual cost of slow payments, 82% of contractors waiting 30+ days (up from 49%); only ~90,000 liens filed in trailing 12 months against that base (NCS Credit) — https://rabbet.com/reports/construction-payments-2024 and https://www.docjoist.com/reports/construction-payment-statistics — observed 2026-08-07 Category: Workflow automation

3. The opportunity

The incumbents sell filing. Nobody sells detection.

Levelset (now Procore-owned), SunRay, Handle, Lienser, CNS, NCS Credit, RCS — all of them are excellent at the moment after you’ve decided to send a notice. You log in, you type the job, you pay $30–$50, they serve it certified mail. SunRay charges $35 per Florida NTO. That machinery works fine.

The failure happens 20 days earlier, and it’s a failure of noticing. The sub’s office manager is not sitting in a lien portal waiting for jobs to appear. She’s in QuickBooks, coding a supplier invoice to job #4471. That coding event is the first furnishing. It is the legal starting gun. And in every incumbent product, the workflow requires a human to independently remember that a legal clock just started, leave their accounting system, and go declare it.

Levelset does have a QuickBooks integration — that’s a real fact and I’m not pretending otherwise. But it pulls jobs and invoices for tracking, built around the assumption that you’ve already set up the job and know you want it monitored. It’s a system of record for jobs you’ve enrolled. It is not a tripwire that screams when a cost lands on a job you never enrolled, which is precisely the job that kills you. The jobs that lose you money are the ones nobody thought to enter.

The 10× is narrow and specific: invert the workflow from declare-then-track to detect-then-confirm. The sub does nothing. The system watches the ledger. The moment a material invoice or a payroll hour hits a job code that has no notice on file, it computes the state deadline from that furnishing date, and sends one message: “Truck delivered to 1420 Harbor Blvd on Aug 3. California clock says you have until Aug 23. Send notice? Yes / Not lienable.”

That’s it. One-tap. The filing itself can be resold through an existing service bureau — I don’t need to reinvent certified mail.

Second reason this is fresh: Procore’s ownership of Levelset pushed it upmarket. Procore’s pricing is volume-based and contractors report $10K–$60K/year, with one paying $55K; for a 15-person crew doing $2M in annual volume, that math is absurd. The bottom of this market — the 5–75 person electrical, plumbing, HVAC, drywall and concrete subs — got structurally orphaned. They are exactly the firms whose office manager is one person wearing four hats.

4. Target market

• Primary customer: Owner or office manager at a US specialty trade subcontractor (NAICS 238 — electrical, plumbing/HVAC, drywall, concrete, roofing, glazing), 5–75 field staff, $1M–$15M annual revenue, running QuickBooks Online/Desktop or Sage 50, working commercial or multi-family jobs as a sub to a GC. Highest-value beachhead: California, Florida, Texas, Arizona, Washington — the states where notice is mandatory, the deadline is short, and the penalty for missing it is total forfeiture.

• Why they buy: Because they have already eaten one. Almost every sub of this size has a story about a $40K–$120K receivable that went unrecoverable because the notice window closed while the invoice sat in a folder. The pain is not abstract: 82% of contractors wait 30+ days for payment, and by the time the payment is visibly late — day 45, day 60 — the California window has been shut for a month. The moment you realize you need the leverage is the moment it’s already gone. That asymmetry is the whole emotional core of the sale.

• Rough TAM reasoning: The US Census counted roughly 497,875 specialty trade contractor establishments (2020 figure). Filter to firms large enough to have an office manager and small enough to be priced out of Procore — call it 5–75 employees — and to the mandatory-notice states. A conservative serviceable slice is 40,000–80,000 firms. At $150–$400/mo, capturing even 1% of the low end is a real business. I don’t need to win the category; I need 700 customers.

• Why now for them: Payment delays roughly doubled in two years (49% → 82%). Their working capital is thinner than it’s been in a decade, GCs are slower, and 100% of subs now report checking a GC’s payment history before bidding. Lien leverage went from a legal nicety to a cash-flow instrument — and simultaneously, the affordable tool that used to watch it for them got absorbed into a $10K+/yr enterprise platform.

5. Product sketch (MVP)

• Ledger tripwire. Connects to QuickBooks Online (then Desktop, Sage 50) read-only. Watches for the first cost — supplier invoice, material receipt, or payroll hour — coded to any job with no notice on file. That timestamp becomes the candidate first-furnishing date.
• State clock engine. Given furnishing date + job site state + your role (sub vs. sub-sub vs. supplier), computes the actual statutory deadline and shows a countdown in days. Covers the mandatory-notice states first: CA, FL, TX, AZ, WA, UT, MN, NV, MA, MI.
• One-tap triage. Every detected job arrives as a single decision: Send notice / Not lienable (residential exempt, public job, direct-to-owner) / Snooze. Dismissals are logged with a reason — that log is your audit trail.
• Party resolution. Pulls likely property owner, GC, and lender from public property and permit records so the office manager isn’t hunting for the legal owner name and mailing address — the single most tedious part of preparing a notice.
• Filing hand-off. One click packages the notice and routes it to a partnered filing service for certified-mail service. FurnishWatch charges the subscription; the bureau charges its per-notice fee.
• Texas monthly mode. For Texas jobs, tracks the rolling monthly notice obligations per month of work rather than a single deadline — the state where the most rights quietly die.
• Exposure dashboard. “You have $312,000 of work in progress. $84,000 of it is currently unprotected.” One number the owner actually cares about.
• Weekly digest. Friday email: what started, what’s closing, what’s protected.

6. AI angle — what’s load-bearing

Remove the AI and this product does not function. Three places it does real work:

1. Deciding what a cost event actually means. A QuickBooks line reading INV 8821 — FERGUSON — 4471 has to be classified: is this a first furnishing on a lienable commercial job, or is it a shop supply restock, a warehouse purchase, a tool rental, or a rebill? Chart-of-accounts conventions are idiosyncratic per contractor — every shop names jobs differently. A rules engine written for one contractor breaks on the next. Classification over messy, per-tenant, free-text ledger data with no shared schema is exactly the job that got cheap and reliable in the last 18 months.

2. Reading supplier invoices and delivery tickets. The furnishing date is frequently on a scanned PDF or a photo of a delivery ticket from the field, not in a structured field. Extracting the delivery date, site address, and material description from a photographed ticket is load-bearing document AI.

3. Resolving the legal parties. Matching a job-site address to the recorded property owner, the GC of record, and any construction lender means reconciling messy address strings across permit and property records that disagree on formatting. Fuzzy entity resolution at volume.

What is not AI, and must not be: the deadline math. State statutory deadlines are deterministic rules maintained in a table by a human with legal review. An LLM guessing a lien deadline is a lawsuit. The AI decides what happened and when; hard-coded rules decide what that means legally.

7. Localization angle

N/A — this is a US-only play, and the US-ness is the entire moat. Mechanics lien law is state-by-state statutory law with no analog that transfers. The 50-state deadline matrix, the Texas monthly-notice quirk, the Utah electronic registry — that accumulated rule set is the durable asset. Attempting to generalize this internationally destroys the value. Within the US, the “localization” that matters is state depth: get California, Florida, Texas, Arizona and Washington genuinely right before touching the other 45.

8. Business model — path to $1M–$5M ARR

• Pricing: Three tiers by active job volume, not seats. Starter $149/mo (up to 25 active jobs, 1 state), Pro $299/mo (up to 100 jobs, multi-state, Texas monthly mode, exposure dashboard), Shop $599/mo (unlimited jobs, multi-entity, priority party resolution). Filing fees pass through to the bureau at cost or with a small margin.
• ACV: Blended ~$3,600/yr. Most of this market lands in Pro.
• Rough math to $1M ARR: 280 customers × $299/mo × 12 = $1.0M. Against a serviceable base of 40,000–80,000 firms, that’s under 1% penetration.
• Rough math to $5M ARR: ~1,200 customers at a blended $3,600, plus filing-fee margin. Requires all mandatory-notice states covered, a Sage/Foundation integration, and at least one distribution partnership (trade association or supplier). Realistic in 30–36 months; not in 18.
• Expansion path: Job-volume tiers grow naturally with the contractor’s revenue — this is the good kind of usage pricing, because it expands when the customer is winning. Then: per-notice filing margin, lien filing (higher value, lower frequency), notice-of-intent-to-lien as a dunning tool, and eventually a receivable-risk score on GCs built from observed payment behavior across the customer base.

Honest note on gross margin: party resolution requires paid property/permit data. Shovels-class permit APIs and property data are a real COGS line. Budget it as a genuine cost center, not a rounding error — this is not a 95%-margin SaaS.

9. Go-to-market wedge — first 100 customers

• Mine the public lien records for people who already learned the hard way. State and county recorders publish mechanics lien filings. Every filing names a claimant — a sub who got burned badly enough to escalate. Pull 12 months of filings in CA, AZ and FL, filter to firms with 5–75 employees, and you have a list of several thousand subs with demonstrated, documented payment trauma and proven awareness that lien rights matter. This is the highest-intent cold list in the entire vertical, and it’s free. Message: “You filed a lien on the Harbor project in March. How many other jobs didn’t you file on because you found out too late?”
• Sell through the material suppliers. Electrical and plumbing supply houses (independent distributors, not just the big chains) have deep, trusted relationships with exactly these subs, and they hold the same credit risk on the same jobs. A supplier branch manager introducing this to their top 40 contractor accounts is a warmer channel than any ad. Offer a revenue share, or white-label it as a value-add on their contractor credit accounts.
• Ride the state-specific panic calendar. Texas monthly notice deadlines and California’s 20-day rule create recurring, dateable urgency. Publish a genuinely useful free 50-state deadline calculator, then run targeted outreach to subs in one state at a time. Free calculator is a lead magnet with real utility, not a content-marketing shrug.
• Construction bookkeepers and outsourced CFOs. A specialized construction bookkeeper serves 15–40 contractor clients and is the person actually sitting in the QuickBooks file. Land 10 bookkeepers, get access to 200+ contractors, and the integration is installed by someone who already has admin credentials. This is the highest-leverage channel and probably where customers 40–100 come from.
• The specific first move: 500 lien claimants scraped from CA + AZ county recorder filings, personalized email naming the actual project they filed on, offering a free retrospective audit — “connect QuickBooks read-only, we’ll show you every job in the last 12 months where your window closed unprotected, and what it was worth.” That retrospective audit is the demo. It quantifies the loss in the customer’s own dollars, on their own jobs, before they pay anything.

10. Build complexity — justification

Medium. The accounting integrations (QuickBooks Online API, then the genuinely unpleasant QuickBooks Desktop) and payroll connections are standard but fiddly integration work. The 50-state deadline rule engine is not technically hard but demands real legal review — budget for a construction attorney on retainer to build and sign off on the initial 10-state matrix, and that is a non-negotiable cost. Party resolution against permit and property data is licensed-data plumbing plus fuzzy matching. Ledger classification is off-the-shelf model work with a per-tenant learning loop.

Estimate: 4–5 months to a credible v1 for two people covering 5 states and QuickBooks Online only. The temptation to launch all 50 states must be resisted — depth in California alone is a sellable product.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Deadline tracking and document preparation. Must ship with clear “not legal advice” positioning and route actual service through a licensed filing bureau. Avoid anything resembling unauthorized practice of law.
Ethical — no harm / dark patterns	✅	Helps small firms collect money they already earned. The one ethical guardrail: never auto-file. Notices strain GC relationships and that judgment must stay with the contractor — always human-confirmed.
Market exists (evidence above)	✅	$280B slow-payment cost, 82% waiting 30+ days, seven-plus incumbents charging money for the filing half of the workflow.
1–5 person team can build this	✅	Two engineers plus a fractional construction attorney.
Launchable with <$50K / ₹40L	✅	Main costs are property/permit data licensing and legal review of the deadline matrix. Comfortably under $50K for a 5-state launch.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	17/20	Total forfeiture of a five-figure receivable, caused by a clerical miss. Practitioners call it the largest single cause of payment-recovery loss. Not 18+ because it’s a low-frequency catastrophic event, not daily friction — which makes it easy to defer buying.
Demand evidence	15	12/15	Strong indirect evidence: $280B, 82%/30+ days, multiple incumbents charging real money for adjacent filing services. Docked because I could not source direct verbatim complaints from subs specifically about detecting first furnishing — the searches for contractor-forum quotes came back empty. That’s the weakest link in this proposal and I’m not going to paper over it.
Build feasibility	15	11/15	QuickBooks Desktop and the 50-state rule matrix with legal review push this past a quick MVP. 4–5 months, not 6 weeks.
Distribution clarity	15	12/15	The public-lien-records list is genuinely excellent and free — named firms with proven trauma. Supplier and bookkeeper channels are plausible but unproven.
Revenue mechanics	15	11/15	$299/mo is defensible against a $35/notice bureau and a $10K+ Procore. 280 customers to $1M is achievable. Docked for real data COGS and unproven willingness to pay a subscription for insurance against a rare event.
Time to first revenue	10	8/10	The retrospective audit closes fast because it quantifies past losses in the customer’s own numbers, but a 4–5 month build delays the start line.
Defensibility	10	4/10	Honest assessment. The rule matrix is a real 12-month asset and ledger integration creates workflow lock-in, but Levelset already owns QuickBooks integration and could reposition. This wins on focus and price, not on a moat. The eventual GC payment-behavior dataset is the only durable moat and it’s years out.
Total	100	75/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required

You need someone who can build reliable accounting integrations, and you need real construction-lien domain knowledge — either a founder from the trade or an attorney co-founder/advisor with equity. Attempting this with neither will produce a product that is wrong in ways that get a customer sued.

Key assumptions to validate (3–5)

1. Assumption: A first-furnishing event is reliably detectable from a typical small sub’s QuickBooks file. How to test: Get read-only access to 10 real contractor QuickBooks files (offer the retrospective audit free). Manually establish ground-truth furnishing dates for 100 jobs, then measure detection precision and recall. Below ~85% precision, the alert becomes noise and the product dies.
2. Assumption: Subs will pay a recurring $299/mo to prevent an event that hits them maybe twice a year. How to test: Run the retrospective audit for 25 subs, show them the quantified historical loss, and ask for a card on the spot. Watch whether the number converts or just impresses.
3. Assumption: The gap between “notice window closed” and “we realized we weren’t getting paid” is real and large. How to test: In the same 25 audits, measure the actual interval between first furnishing and the first internal collections activity on jobs that went bad.
4. Assumption: Filing bureaus will partner rather than compete. How to test: Pitch three regional bureaus on a referral arrangement. If all three refuse and say they’ll build detection themselves, the moat problem is worse than scored.

Risk flags

1. Incumbent response: Levelset already has a QuickBooks integration and Procore’s resources. If they ship first-furnishing detection at the low end, the wedge closes. Mitigation is speed and the deliberately downmarket price point Procore is structurally unwilling to serve.
2. Liability: If the product misses a furnishing event or miscalculates a deadline and a customer loses a $200K lien right, you are the named defendant. Requires E&O insurance, aggressive contractual limitation of liability, and framing the product as an assistant rather than a guarantor. This is the single largest business risk and it is not small.
3. Platform dependency: Deep reliance on the QuickBooks API. Intuit changes terms and pricing on its own schedule.
4. Data COGS: Property and permit data licensing scales with usage and compresses margin. Also note that permit-data vendors refresh on roughly biweekly cycles with a further processing lag — permit feeds are useful for party resolution but are far too slow to serve as the furnishing trigger. The ledger has to be the trigger. Building on permit data as the primary signal would be a fatal architectural mistake.
5. Relationship risk as sales objection: Many subs deliberately don’t send notices because they fear annoying a GC they want repeat work from. In some regions notices are routine and expected; in others they’re read as hostile. This objection will come up in every sales call and the answer must be regional evidence, not a shrug.

14. Structured verdict

Score:                  75/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder who can ship accounting integrations,
                        paired with a construction-lien attorney as co-founder or
                        equity advisor. A former trade-shop office manager as first
                        hire would be worth more than a second engineer.
Time to revenue:        4–5 months to v1, first paying customer within 6–8 weeks
                        of launch via the retrospective-audit motion.
Capital to launch:      $30–45K (legal review of the deadline matrix, property/permit
                        data licensing, E&O insurance, incorporation)
Top 3 assumptions to validate first:
  1. First-furnishing detection from real QuickBooks files hits >85% precision —
     test on 10 donated contractor files against manual ground truth.
  2. Subs convert on a quantified retrospective loss — run 25 free audits,
     ask for the card, measure close rate.
  3. Filing bureaus partner instead of competing — pitch three regional bureaus.
Kill criteria:
  - Abandon if detection precision stays below 75% after tuning on 10 real
    QuickBooks files — noisy alerts make this worse than a spreadsheet.
  - Abandon if fewer than 4 of 25 retrospective audits convert to paid.
  - Abandon if Levelset or Procore ships ledger-triggered furnishing detection
    at a sub-$200/mo price point before your v1.

15. Next step — 1-week validation sprint

• Day 1–2: Pull 12 months of mechanics lien filings from Los Angeles, Orange, Maricopa and Miami-Dade county recorders. Filter to specialty trade claimants with 5–75 employees. Build a list of 300+ named firms with confirmed payment trauma. This list is itself an asset regardless of the outcome.
• Day 3–4: Call 30 of them. One question set: “Walk me through the last job where you found out too late that you couldn’t lien. When did you first realize? What was it worth?” Log the dollar figure and the detection lag every time. Simultaneously, secure read-only QuickBooks access from 3 friendly contractors and hand-check whether first-furnishing events are actually identifiable in their real chart of accounts.
• Day 5: Decide. Go requires: (a) ≥12 of 30 subs describe a specific job lost to a missed window with a dollar figure attached, (b) median stated loss ≥$25,000, and (c) first-furnishing events are manually identifiable in at least 2 of the 3 real QuickBooks files without heroic interpretation.

Falsifiable failure: if subs consistently say “we knew about the deadline, we chose not to send the notice because we didn’t want to upset the GC” — then this is a relationship problem wearing a detection costume, and no amount of tripwire software fixes it. Kill it and move on.