Evidence Center
Five levels of evidence. Pick yours.
Construction software vendors love a hero metric. We don't. Every claim we make is gated to an evidence level you can verify. Here's the ladder, here's what's available now, and here's what lands as pilots ratify.
The ladder
What we'll show, and when.
- L1
Live sanitized outputs
Available nowShow, don't tell. Run a representative plan set in front of you with PII redacted, share the takeoff and proposal artifacts, no edits applied.
- Schedule a live bid review on a plan set we provide
- Bring your own plan set under a one-line NDA
- Walk through the takeoff edit table, evidence graph, and review queue
- L2
Workflow demonstrations
Coming soonPre-recorded demos walking through end-to-end runs on representative trades, with timestamps and the source plan-set redacted versions linked.
- Electrical bid run-through (~12 min)
- GC multi-trade rollup (~18 min)
- Plumbing fixture-schedule extraction (~9 min)
- L3
Pilot artifacts
Post-pilotReal bids run by real customers, with the customer's consent, sanitized of pricing and proprietary scope. Includes win/loss outcome.
- Pilot project A — pending publication
- Pilot project B — pending publication
- Pilot project C — onboarding
- L4
Validated outcomes
Post-pilotPost-bid analysis: predicted vs actual, with the customer's reported variance. Names appear only after written approval; the artifacts themselves are public.
- Quarterly variance reports per pilot customer (anonymized until approved)
- Specific trade accuracy claims backed by reproducible test fixtures
- L5
Independent benchmarks
Coming soonThird-party head-to-head benchmarks against named competitors on a shared plan-set corpus. Methodology open-sourced.
- Open-source benchmark corpus + harness on GitHub
- Annual head-to-head report
Our evidence policy
Customer logos appear after written approval. Period.
- We never publish a customer logo, project value, or screenshot without written approval
- All public artifacts redact PII, pricing, and proprietary scope unless the customer waives
- Reproducible fixtures live on GitHub once the corpus stabilizes
- We publish what didn't work, too — calibration misses, model drift, addendum-delta gaps
- Outcomes are reported variance-from-actual, not hero metrics
Read the privacy policy for how we handle workspace content.
Frequently asked
Evidence questions we hear.
- Can I see a real customer's takeoff?
- Once the customer has approved publication and signed a release, yes. Until then, we run a representative plan set live for you under NDA.
- Why don't you publish a hero accuracy number?
- Because hero numbers without a fixed corpus and a fixed model are marketing, not evidence. The benchmark methodology is in development; when it ships, the numbers ship with it.
- How do you handle customer NDAs?
- We sign mutual NDAs by default. Customer content stays in the customer's org. We retain the right to publish redacted aggregates only with written approval.