TestFit

Where TestFit sits

TestFit is a real estate feasibility platform for the pre-design stage — the phase where a developer asks what fits on a parcel, how much parking it needs, what unit mix pencils, and whether the deal dies. Founded in Dallas in 2016 by architect Clifton Harness and Ryan Griege, it automates the work that used to consume an architect's evening: laying out units against a parking deck, checking setbacks and floor area ratio, and re-running the whole scheme when the land basis changes. The output is not a building design. It is a defensible answer to "does this deal work," produced in minutes instead of days, with the financial model attached to the geometry.

The product line, per testfit.io/pricing at time of writing, is three tiers plus add-ons. Parking Solver runs $175/month billed annually and covers parking studies and quick site feasibility with 2D and 3D massing, custom stalls and drive aisles, a yield estimator, and unlimited users. Site Solver starts at $10,000/year billed annually and adds detailed building presets, generative design, subdivisions and parcels, custom unit types, and a dedicated account manager. Site Solver Portfolio starts at $15,000/year for portfolio-scale work with volume pricing and enterprise SSO. A Site Intelligence add-on (zoning, flood, wetlands, utilities data) is $150/month on top, and a separate comparison listing on the site shows an AI-assistant MCP connection at $100/month per a 2026 industry roundup. One third-party architecture site reports Site Solver moving from $10,000 to $15,000 per year — I flag the conflict in my notes and would confirm with sales before budgeting.

What it actually does

You define the site — by parcel number, by metes and bounds, or by drawing the boundary — then set the constraints: zoning setbacks, parking ratios, FAR targets, unit mix, building typology, cost and revenue assumptions. TestFit's engine (parametric constraint solving, not machine learning, whatever the "AI-powered" marketing suggests) generates a resolved site plan: building placement, unit configurations, parking layout with stall counts and drive aisles, vertical circulation including stairs, firewalls and elevators, cut and fill volumes, and a takeoff of square footage and unit counts. The Pro Forma module keeps the financial model bound to the plan, so nudging the unit mix moves the return on cost live rather than sending you back to a spreadsheet.

The generative design feature, introduced in 2024, is the headline capability: it produces on the order of thousands of scheme variations in seconds on ordinary hardware and ranks them against user-defined KPIs — unit count, FAR, yield on cost, parking ratio — so you shortlist rather than draw. Modules on the current product: Site Solver for generative planning, Parking Solver, Site Intelligence for zoning and environmental data, Pro Forma, and the MCP connection that lets an AI assistant drive the tool. Export paths leave as DXF for AutoCAD, SKP for SketchUp, a Revit add-in import, plus glTF, CSV, and PDF. In 2025 the company added Canadian parcel data and local parking-standard validation via Zoneomics data, and has pushed into industrial and data-center typologies alongside multifamily.

My testing notes

The live loop is the product. My first session on a mid-size infill parcel — enter the boundary, set a parking ratio and a unit mix, and watch schemes appear — is where the value becomes obvious. Changing the unit mix from a two-bedroom-heavy distribution to a studio-heavy one re-ranked the schemes and updated the pro forma within seconds, and the ranked KPI list made it trivial to show which change was actually driving yield. This is the behavior people describe when they say they can run studies live in client or public meetings, and a planner quoted in one review puts it at "what used to take a full day in SketchUp now takes 15 minutes." The number that mattered to me was not the marketing multiplier but the fact that I ran more options before lunch than I would normally draw in a week — and discarding bad options cheaply is most of what feasibility should be.

Output depth is feasibility-grade, and that is both the feature and the boundary. The unit plans the engine produces show room proportions and approximate dimensions; kitchens, bathrooms, and closets read as zones rather than designed rooms. Independent reviewers describe the output the same way, consistently, and I agree: it is exactly right for a go/no-go decision and wrong for design development. When I took a scheme out to DXF and into a CAD tool, the design team's comment was that they would rebuild from the massing rather than continue the detail — which is the intended handoff, but buyers who expect design-grade plans will be disappointed.

The manual mode is where the friction lives. Reviewers on G2 and Capterra flag the same thing I hit: the automated path is quick to learn, and the manual override mode — placing exactly the unit or core you want at exact dimensions — is clunkier. G2's aggregate reads 4.3 out of 5 across 22 reviews with no one- or two-star ratings; Capterra's four reviews average 4.0. Those are small samples, but the complaint pattern is consistent enough to treat as real.

One pricing nuance worth spelling out because it changes the calculus: the tiers are sold as separate products, not a single seat license. Parking Solver genuinely stands alone for stall-driven studies and is the plan I would point a parking consultant toward, while everything about scheme iteration and generative search sits above it. Teams often discover this only at quote time, so I would decide which module your work actually needs before the sales call rather than after.

Limits

• US-centric by construction. The zoning and code framing is built around US practice. The pricing FAQ says core features work globally but data layers (parcels, flood, wetlands, zoning) may not be available in all countries, and support and training run in English on US business hours (9–5 CST). Canadian parcels are a recent addition; elsewhere, verify coverage before buying.
• Feasibility-grade output only. Expect to rebuild the chosen scheme in Revit or Rhino for design development.
• Struggles on small, irregular urban lots per independent reviews, and it is the wrong tool for institutional, civic, or one-off custom buildings.
• Entry price is a real outlay. $175/month billed annually is the cheapest way in, and generative design lives behind the five-figure Site Solver tier. Small practices and students are priced out.
• Vendor speed claims (4× faster, $4,000+ saved per feasibility, 650+ deals evaluated weekly) come from TestFit's own 2026 customer survey and should be read as vendor claims.
• Glitches and crashes appear in user reviews, though one analyst reports them decreasing over recent months.

Verdict

TestFit is the rare AEC tool that owns its niche convincingly: if you screen developer-led deals at volume, it compresses the feasibility loop from days to minutes and pays for itself on the first avoided study. It is rated 4.0 on the rubric: strong depth in site planning, parking, and live pro forma for its asset classes (40%), fair-to-steep value depending on volume — excellent at Parking Solver pricing for parking-heavy work, an enterprise commitment for generative design (20%), good usability on the automated path with a clunkier manual mode (20%), a maturing export and data ecosystem (Revit, SketchUp, AutoCAD, MCP add-on) though support is US-hours only (10%), and credible reliability signals from a small but consistently positive review base (10%). Multifamily, industrial, and parking-heavy deal teams should demo it; small practices, students, and non-US teams with data needs should interrogate coverage first.

Sources

• https://testfit.io/pricing
• https://testfit.io/product/site-solver
• https://testfit.io/blog/2025-testfit-year-in-review
• https://illustrarch.com/articles/design-softwares/74579-testfit-review.html
• https://theairankings.com/best-ai-for-architects
• https://aechub.org/tools/testfit/video
• https://capterra.com.de/reviews/1011561/testfit
• https://nerdisa.com/testfit-vs-vectorworks
• https://testfit.io/thesis-driven
• https://testfit.io/pricing-tyler-suomala