Where we store the bets: hypotheses, target accounts, wedge, messaging, insights. We fill the open boxes together.
What are we trying to learn?
Learn at what company size maintenance/WO automation is a real, budgeted pain — not close revenue.
No means to deliver an enterprise rollout today. So every touch is discovery: find the pain, the buyer, the budget. Logos are for learning, not for invoicing — don't burn the good ones on a pitch we can't yet fulfill.
The bets we're testing. Click a status pill to cycle it (testing → validated → killed → queued).
Who we target — and just as important, who we exclude. "Any enterprise" is the answer this section exists to kill.
Prometheus (~28k) sits at the top edge of the band — "big, but not too big." That edge is the bet.
How we go from one name to a real list: where to source operators, then how to rank them.
Goal is discovery — so weight reachability + fit over "ability to buy" (we're not closing yet).
| Criterion | What we score | Weight |
|---|---|---|
| Fit | Units in ~1–30k band, multifamily, regional | high |
| Pain | Decentralized vendors / legacy systems → more follow-up pain | high |
| Reachability | Exec findable on LinkedIn, email pattern known | high |
| Strategic value | Marquee regional name, reference-able later | med |
| Urgency | Growth, recent bad reviews — squishy | low |
| Ability to buy | N/A — discovery, not a sale | skip |
Where we hunt first.
Bay Area — most ideal. Prometheus is here, so you can meet execs in person; the Bay is a dense cluster of right-sized operators; and the existing SMB list already gave you Bay Area muscle. Expand to broader Northern California only after the Bay is swept.
Ranked best-first. Tier rolls up fit + pain + reach into one call; Why is the reason. Click a status pill to advance it.
| Tier | Company | Units | Geo | Segment | Status | Why this rank |
|---|
Tiers: S flagship · A strong · B caveated · C parked (affordable/Sec 8, deprioritized per H3). Add accounts by editing the source; statuses persist locally until we re-promote.
The one specific problem we lead with.
Follow-up to completion — tracking every open WO and chasing the vendor until the work is verified done.
Andrew's biggest manual time sink. The bet: this pain is ~linear for SMB (a few jobs, held in your head) but super-linear at scale — hundreds of jobs across many vendors/properties is where the loop breaks. That gap is the enterprise hypothesis. It's also a better "pay a lot" story than triage: a dropped job = re-work, tenant churn, liability.
The real question: if a work order slips, what do they actually lose? Hard ROI vs the bear case.
| Cost bucket | What happens when follow-up fails | ~$ / event |
|---|---|---|
| Avoidable turnover | Resident doesn't renew over bad maintenance | $4–8k make-ready + Bay Area vacancy + concession |
| Damage escalation | Small leak sits → water / mold remediation | $3–20k the pure "nobody confirmed it was done" cost |
| Habitability / legal | Unaddressed → rent abatement, fines, suit | $10k+ CA is tenant-friendly; tail risk, low frequency |
| Vendor leakage | Dup dispatch, paid-not-done, emergency premiums | 10–20% of spend recovered by tracking alone |
| Follow-up labor | Coordinator time spent chasing | ~$18k / 1k units / yr reclaim ~half with a tool |
Back-of-envelope (per 1,000 units, then ×28 for Prometheus — all assumptions, verify)
At scale these losses are real, recurring, CFO-legible. A $1–3/unit/mo tool ($336k–$1M/yr at 28k) clears 1.5–4× ROI even discounted.
The expensive failures already get caught — the tenant screams about the leak. What truly slips silently is the cheap stuff (squeaky hinge, ~$0). If so, the marginal WO we save is worthless → ROI collapses → H1 confirmed.
Discovery resolves this. Don't ask "is follow-up annoying" (yes, mildly). Ask: "tell me about the last dropped or delayed WO that actually cost you real money." War stories → real pain. Blank stares / all small → H1 confirmed.
Goal is discovery, so these are research asks, not pitches — founder voice, low ask, give-to-get. A/B the angle; we learn which framing makes an exec reply. Swap [Name]/scale per account.
How the touches actually go out.
Are we validating pain, or selling a solution?
Discovery. Every call is to learn whether the pain is real and budgeted at this scale — not to pitch a product we can't yet deliver. This follows directly from the Goal.
Goal is learning and n is small — so weight absolute counts + qualitative signal over noisy rates. Batch 1 = ~10–15 execs across ~8–10 operators.
| Metric | Batch-1 target | Why |
|---|---|---|
| Reply rate | ≥ 25–30% | Founder research ask, hand-personalized — should crush the 1–5% cold norm |
| LinkedIn connect accept | ≥ 40% | Founder profile, warm-ish ask |
| Discovery calls booked | 3–5 | The real near-term KPI |
| $-loss war stories | ≥ 2 | The crux: real, specific dollar losses from dropped WOs — or a confident "they all get caught" → H1 confirmed |
Rates are noisy at n≈10 — treat them as directional. The war-story count is the signal that actually moves H1.
Running record of what we decided and learned.
Local edits (statuses, new insights) live in your browser. Hit Copy as JSON and paste it back to Claude to persist into the source + re-deploy.