WageFloor

1. One-liner

Tells an electrical contractor what their crew actually costs this month, before they sign a fixed-price bid.

2. Trend signal — why now?

Three things moved at once, and they moved against the same people.

The data-center boom is bidding electricians away from everyone else. Electrical work is 45–70% of data center construction cost, which pulls experienced electricians toward those megaprojects and away from standard commercial jobs. Microsoft president Brad Smith called the electrical talent shortage the number one problem slowing data center expansion in the U.S. The industry needs 300,000+ additional electricians while roughly 20,000 retire every year. Northern Virginia electricians now clear $120K+. Contractors increasingly cite labor availability — not project demand — as their primary growth constraint.

Burden rates have drifted, and nobody re-derived them. Labor burden adds 38–55% on top of base wages once workers’ comp, payroll taxes, benefits and insurance are counted. An electrician earning $45/hr costs $62–70/hr before overhead or profit. The trade press is blunt about the consequence: “If your burden was 35% in 2023 and it’s actually 45% now, you’re losing 7-8% on every labor dollar.” Set that against NECA’s Financial Performance Report, where median electrical contractor net margin is roughly 5–6%. A 7-point error on labor doesn’t dent the margin. It eats the whole thing and starts on the equity.

Wage floors became machine-readable this year. New Jersey’s prevailing wage rules for data centers take effect April 1, 2026, covering facilities at 250kW+ electrical capacity. New York’s DOL announced electronic certified payroll submission in December 2025. Published, per-classification, per-county wage determinations are now structured data instead of PDFs in a filing cabinet — which means a small tool can finally triangulate published floor against what the shop next door is actually paying against what this contractor booked in their estimate.

The shape of the pain is not “I need compliance paperwork.” It’s: the number in my estimating software is a number I typed in 2023.

Provenance:

• Signal 1 (demand): Contractors using 2-3 year old burden rates lose 7-8% on every labor dollar; burden now 38–55% of base wage — https://www.constructioncostaccounting.com/post/2026-construction-bidding-material-labor-cost-trends-to-price-jobs-profitably — 2026-08-04
• Signal 2 (economic): Data center boom created 300K+ electrician shortfall, made labor the binding growth constraint; electrical = 45–70% of DC cost — https://build.inc/insights/data-center-construction-labor-shortage-2026 — 2026-08-04
• Signal 3 (feasibility): NJ data-center prevailing wage effective 2026-04-01; NY electronic certified payroll launched Dec 2025 — per-classification wage floors now structured/queryable — https://vensure.com/employment-law-updates/new-jersey/new-jersey-prevailing-wage-rules-for-data-centers/ — 2026-08-04 Category: Tech-unlock

3. The opportunity

Every estimating tool on the market treats labor burden as a constant the contractor configures once. ServiceTitan, Simpro, BuildOps, the countless Excel price books — they all have a burden field. They all assume you know what to put in it. The category’s own guidance admits the failure: “recalculate labor burden rates every year with actual insurance quotes and benefits costs.” Which is advice, not software. Nobody does it, because doing it means calling your insurance broker, re-pulling your comp mod, and reconciling against payroll — a half-day of CFO work at a shop that has no CFO.

The incumbents are not bad at this. They’re not playing. Estimating software sells itself on takeoff speed and bid volume — how fast you can get a number out the door. The number’s accuracy is defined as the user’s problem. That’s a defensible product decision when wages are flat. It is a catastrophic one in a market where a megaproject 40 miles away just repriced your entire labor pool.

The wedge: be the layer that keeps the burden number honest, and prove it in dollars. Not another estimating tool — an input to whatever estimating tool they already own. A contractor who learns their true burden drifted from 42% to 51% doesn’t need to switch platforms. They need to know before signing the next fixed-price contract, and they need to know which of their last six jobs were priced with the stale number so he can decide whether to chase a change order or eat it.

Ten-times-better is not a better UI. It’s the difference between finding out at bid time and finding out at year-end close.

4. Target market

Primary customer: Owner or estimator at a US electrical subcontractor doing $2M–$25M in annual revenue, 10–80 field employees, mostly commercial/institutional bid work, operating within commuting distance of a metro where data-center or large-industrial construction is pulling labor. Northern Virginia, Columbus, Phoenix, Dallas–Fort Worth, Atlanta, Salt Lake, central Iowa. Extends naturally to mechanical, HVAC, and low-voltage subs with the same cost structure.

Why they buy — in their words: The industry’s own commentary is unsentimental about the dynamic. On wage competition: “Underpay for your market and your best people leave for the shop paying $2 more an hour.” On bidding behavior: “If there were only two electrical contractors in the entire world and neither one could keep up with all the work they currently had they would still keep trying to underbid each other.” On the squeeze from below: “The trend in my neck of the woods seems to be that contractors are hiring younger guys and illegals to do almost all residential and commercial work, paying them half or less per hour than what i and most of my work buddies require in order to survive.” And the direct framing of the miss: contractors who estimate on direct cost alone “may systematically underbid work by 35-60%.”

Note what these quotes have in common. Nobody is complaining about their software. They’re describing a market that repriced underneath them. That’s the buying trigger — and it’s why this sells as insurance, not as a productivity tool.

Rough TAM reasoning: There are on the order of 70,000 electrical contracting establishments in the US; the $2M–$25M commercial-bid band is a minority of that, call it 8,000–12,000 firms. Add mechanical/HVAC subs with identical economics and the reachable base is roughly 25,000–35,000 firms. At $400/mo average, capturing 1% of the electrical-only slice is ~$500K ARR. This is comfortably a $5M-ceiling business and not remotely a venture-scale one — which is exactly the profile.

Why now for them: Their labor pool was repriced by projects they didn’t bid on and can’t compete with. A contractor in Loudoun County isn’t losing electricians to another electrician — he’s losing them to a hyperscaler’s general contractor with a schedule penalty clause and no price sensitivity. Historically, wage drift was slow enough that an annual burden update was fine. It isn’t anymore.

5. Product sketch (MVP)

• Burden rebuild from source documents — upload the workers’ comp declaration page, benefits invoice, and a payroll export; get a per-classification burdened hourly cost, rebuilt monthly instead of annually
• Market wage read — for the contractor’s counties and classifications, show the published prevailing-wage floor alongside observed market rates, so “what I pay” can be compared to “what the job across town pays”
• Bid guardrail — enter the labor hours and crew mix for a bid in progress; get the cost floor, the margin at the intended price, and a red flag when the bid’s implied burden is below the contractor’s actual
• Stale-bid sweep — point it at the last 12 months of won jobs and get a ranked list of which were priced on an out-of-date burden and by how much, in dollars
• Escalation clause helper — for jobs running 6+ months, produce the labor-index language and the threshold math to justify it to a GC
• Crew retention flag — highlight classifications where the contractor’s own pay has fallen below the local floor, i.e. who is most likely to leave next
• Monthly one-page cost letter — a PDF the owner can hand to their bookkeeper, banker, or bonding agent

Explicitly not in v1: takeoff, scheduling, invoicing, certified payroll filing. Those are crowded, and losing to them isn’t the goal — being the honest number that feeds them is.

6. AI angle — what’s load-bearing

Remove the AI and this product dies at the first customer, because the input is a pile of inconsistent documents nobody will hand-key.

Burden inputs arrive as a workers’ comp dec page (carrier-specific layout, experience mod buried in a table), a benefits invoice (per-employee, per-plan, mid-year tier changes), and a payroll export whose job codes are whatever the office manager invented. Turning that into a per-classification burdened rate is a document-extraction-plus-reconciliation problem: read heterogeneous PDFs, map the shop’s idiosyncratic job codes onto standard classifications, and reconcile hours across systems that disagree. That’s precisely the work that got cheap and reliable in the last 18 months and was genuinely not tractable before.

The second load-bearing use is classification mapping against published wage determinations. The single most common prevailing-wage violation is using the wrong classification — “a worker performing journeyman-level work classified as an apprentice is underpaid under prevailing wage law, and classification needs to match actual work performed, not crew payroll categories.” Matching a shop’s internal labels to the correct published classification is fuzzy, judgment-laden semantic work. A lookup table doesn’t do it. A model reading the actual scope descriptions does.

What is not AI: the arithmetic. Burden math is arithmetic and must be auditable to the penny. The model reads and maps; deterministic code computes. Any product that lets a language model do the multiplication deserves the lawsuit it gets.

7. Localization angle

N/A — this is a US-only play, and deliberately so. The wedge depends on published per-county, per-classification prevailing wage determinations (Davis-Bacon and state analogs) plus a fragmented small-contractor market with no CFO function. The EU and UK have collective-bargaining wage floors that are national and stable — the drift this product detects barely exists there. India and SEA have neither the published-determination infrastructure nor the burden structure. Porting this outside the US means rebuilding the entire data foundation for a market where the pain is milder. Not worth it.

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

• Pricing: $299/mo (under $5M revenue, 1 county), $599/mo (under $15M, up to 5 counties), $1,200/mo (multi-state, unlimited classifications, bonding-agent report pack). Annual prepay at 10x.
• ACV: ~$4,800 blended.
• Math to $1M ARR: 208 customers at $4,800. Against a reachable base of ~10,000 electrical firms in the target band, that’s ~2% penetration.
• Math to $5M ARR: ~1,000 customers, which requires expanding beyond electrical into mechanical/HVAC/low-voltage (same cost structure, same buying trigger) and adding the multi-state tier. Realistic in 30–36 months, not 18.
• Expansion path: counties → classifications → trades within a multi-trade firm. A contractor who opens a second branch doubles naturally. The bonding-agent report is the upsell with the best story: surety underwriters care about cost discipline, and a contractor who can show a monthly rebuilt burden gets treated better at renewal. That’s a real dollar benefit that justifies the $1,200 tier without new engineering.

The honest number here: this is a mid-hundreds-per-month product for a business doing millions in revenue on 5% margins. It’s cheap relative to a single mispriced job, which is the entire sales argument. It is not cheap relative to their current software spend, which is the entire sales objection.

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

• The free stale-burden audit, targeted by geography. Build a list of electrical contractors within 50 miles of announced data-center projects — these are public, tracked in trade press, and county-permit searchable. Start with Loudoun County VA, New Albany OH, Mesa AZ, and Cedar Rapids IA. Offer: send your comp dec page and a payroll export, get your true burden back in 48 hours, free. The deliverable is the sales pitch, because for most of them the number will be materially worse than what they’ve been bidding. Target 400 firms, expect 10–15% to take a free audit, expect a third of those to convert once they see the gap.
• NECA chapter meetings and state electrical associations. Local NECA chapters run monthly meetings with exactly this audience and are perpetually short of programming that isn’t a vendor pitch. “What your labor actually costs in 2026” is a talk, not an ad. Ten chapters, 20–40 owners each, in the metros above.
• The bonding and insurance channel. Surety agents and construction-specialty insurance brokers already have the trust relationship and already ask contractors uncomfortable questions about cost control. They have a direct interest in their book being priced correctly. Recruit 5–10 agents with a referral arrangement; each has 30–100 contractor relationships.
• Mike Holt forums and trade-specific communities. Not Reddit — the electrical trade’s real online center of gravity is Mike Holt’s forums and similar trade boards, where threads like “keep getting undercut” and “how do I keep from being underbid by contractors who don’t follow the specs” run for years. Participate honestly with the burden math; do not pitch. This is slow and produces the highest-quality inbound.

The first channel is the one that matters and it’s a two-week sprint to build the list. The others compound.

10. Build complexity — justification

Medium. The document-extraction pipeline (comp dec pages, benefits invoices, payroll exports across a handful of common providers) is off-the-shelf model work but needs real handling for messy layouts and a human-verification step, because a wrong burden number is worse than no product. Wage determination data is publicly available but requires ingestion and normalization per state — start with 4 states, not 50. The burden arithmetic and bid guardrail are deterministic and straightforward.

The genuine risk to the timeline is not code, it’s correctness: you need a construction accountant to validate the burden methodology before you show a single customer a number. Budget 14–18 weeks to a v1 that a paying contractor will trust, with a domain advisor on retainer throughout. A solo technical builder can do this; a solo technical builder without an accounting advisor will ship something confidently wrong.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Uses public wage determinations plus customer-supplied documents. Not giving legal or accounting advice — outputs are cost calculations the contractor owns. Terms must say so plainly.
Ethical — no harm / dark patterns	✅	Pushes contractors toward paying market rates and pricing honestly. The crew-retention flag surfaces underpaid classifications. Aligned incentives throughout.
Market exists (evidence above)	✅	300K+ electrician shortfall, 5–6% median net margins, documented 7–8% loss per labor dollar from stale burden.
1–5 person team can build this	✅	One technical builder plus a construction accounting advisor.
Launchable with <$50K	✅	Inference, wage-data ingestion, advisor retainer. Well under $30K to first revenue.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	17/20	A 7–8% labor-dollar error against a 5–6% median net margin is existential arithmetic. Felt on every bid. Docked 3 because the pain is invisible — it shows up at year-end close, not at the moment of bidding, which is exactly why nobody’s fixed it and also why it’s harder to sell than a bleeding wound.
Demand evidence	15	12/15	Strong convergent evidence: documented margin loss, documented shortage, published margin benchmarks, active trade-forum complaints about underbidding. Docked because the complaints are about symptoms (getting undercut, losing crew) — I found no one asking for this specific product. Demand is inferred from economics, not from stated want.
Build feasibility	15	11/15	Document extraction is solved; multi-state wage ingestion and classification mapping are grind. The correctness bar is high and non-negotiable. 14–18 weeks, not 6.
Distribution clarity	15	12/15	The free-audit-by-geography play is concrete, list-buildable in two weeks, and the deliverable sells itself. NECA chapters and surety agents are named, reachable channels. Docked because contractor cold-outreach response rates are genuinely poor and the audit requires them to send financial documents to a stranger — real friction.
Revenue mechanics	15	12/15	$299–$1,200/mo is defensible against a single mispriced job, and 208 customers to $1M is achievable. Docked because $5M requires cross-trade expansion that isn’t proven, and because contractors in this band are famously tight on software spend.
Time to first revenue	10	7/10	14–18 week build, then a free audit that converts over weeks. First dollar realistically 5–6 months. The audit can be run manually before the product exists, which pulls revenue earlier for a founder willing to do services work first.
Defensibility	10	5/10	Honest assessment: this is an execution moat, not a structural one. Normalized wage data and validated burden methodology accumulate into something real by month 12, and per-customer cost history creates switching friction. But an incumbent estimating platform could ship a competent version of this as a feature. The bet is that they won’t, because it makes their customers’ bids higher.
Total	100	76/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required

This needs someone who can build a document pipeline and has — or rents — real construction accounting knowledge. A pure technical founder will get the burden methodology subtly wrong and never know until a customer’s CPA catches it. A pure domain expert can’t build it. The pairing is the requirement.

Key assumptions to validate (4)

1. Assumption: Contractors in the $2M–$25M band are actually running materially stale burden rates — the 7-8% figure is real and common, not a vendor’s scare statistic. How to test: Run 15 free audits. Measure the actual gap between their configured burden and their rebuilt one. If the median gap is under 2 points, the product has no story.
2. Assumption: Owners will hand a stranger their comp dec page and payroll export. How to test: In the same 15 outreach attempts, measure document-submission rate separately from interest rate. Interest without documents means the free audit doesn’t work as a wedge and the whole GTM needs rebuilding.
3. Assumption: Knowing the true burden changes bidding behavior rather than just producing anxiety. How to test: Follow up with audit recipients at 30 days. Did they reprice anything? A contractor who learns he’s underbidding and bids the same way anyway will not renew.
4. Assumption: $299/mo clears the bar for a business that already resents its software spend. How to test: Price the audit-to-paid conversion directly. Offer the paid tier at close and measure. Do not ask “would you pay” — ask for a card.

Risk flags

1. Invisible-pain risk. This is the big one. The cost of a stale burden shows up months later in aggregate, not at the moment of decision. Products that solve delayed, aggregate pain sell far worse than products that solve immediate, visible pain, regardless of the dollar magnitude. The free audit exists specifically to make the invisible visible, and if it doesn’t work the idea is in trouble.
2. Incumbent feature risk. ServiceTitan, Simpro, or BuildOps could add dynamic burden as a feature. The counter-argument is genuine — a feature that raises customers’ bids is a strange thing for a platform selling bid velocity to build — but it’s an argument, not a moat.
3. Correctness liability. A contractor who bids on a wrong number from this tool will blame the tool. Methodology must be validated by a credentialed accountant, every number must be traceable to its source document, and the terms must be unambiguous about who owns the bid. This is a real constraint on shipping fast.
4. Market timing risk. The wedge rides the data-center construction boom. If hyperscaler capex slows sharply, wage pressure eases, drift slows, and the urgency deflates. The underlying burden-drift problem persists, but the “why now” weakens considerably.

14. Structured verdict

Score:                  76/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder with a construction accountant as advisor or
                        co-founder; ideally someone with prior exposure to trade contracting
Time to revenue:        5–6 months (or 6–8 weeks if run as a manual audit service first)
Capital to launch:      $25–30K (inference, wage data ingestion, accounting advisor retainer)
Top 3 assumptions to validate first:
  1. Real burden gap is material — run 15 free audits, require median gap >2 points
  2. Contractors will submit financial documents to a stranger — measure submission rate
     separately from stated interest
  3. Knowing changes behavior — 30-day follow-up, did they actually reprice anything
Kill criteria:
  - Abandon if median burden gap across 15 audits is under 2 percentage points — no story
  - Abandon if fewer than 3 of 15 interested contractors actually send documents — the
    wedge doesn't work and there's no second GTM idea of comparable quality
  - Abandon if an incumbent estimating platform ships dynamic burden as a native feature
    before v1 launch

15. Next step — 1-week validation sprint

• Day 1–2: Build the target list. County permit data plus trade-press announcements for data-center projects in Loudoun County VA, New Albany OH, Mesa AZ, Cedar Rapids IA. Extract electrical contractors within a 50-mile radius from state licensing boards. Target 150 named firms with an owner name and phone number. This list is reusable regardless of outcome.
• Day 3–4: Call — do not email — 40 of them. Single offer: “I’ll rebuild your true labor burden from your comp dec page and a payroll export, free, in 48 hours.” Track three numbers separately: reached, interested, actually sent documents. The third number is the only one that matters.
• Day 5: Hand-compute the burden for every submission received, using a construction accountant to check the methodology. Send each contractor their real number against what they’ve been bidding.

Go / no-go: Proceed only if ≥5 of 40 contractors submitted documents AND the median gap between configured and actual burden is ≥2 percentage points. Fewer than 5 submissions means the wedge doesn’t open the door. A sub-2-point median gap means the problem is real but too small to charge for — and in that case the correct move is to kill it, not to lower the threshold.

This is falsifiable in a week, costs nothing but phone calls and an accountant’s afternoon, and the failure mode is informative either way: a high-interest, low-submission result specifically indicts the free-audit wedge rather than the underlying thesis.