Blog

Data centers need quarterly roof + cooling-loop thermal sweeps — go win that contract

7 min read SkySuite Team

Data-center facilities teams already have a software stack that watches their cooling. PRTG, Datadog, ServiceNow alerts. The CRAC outlet temperatures, the chilled-water loop ΔT, the rooftop unit on the southeast corner — all of that is on a dashboard and someone is paged at 3am when it drifts.

What they don’t have, and what they’ll buy from you on a recurring schedule, is the thermal view from above the roof. The dashboard tells them the loop is running 2°F warmer than baseline. It does not tell them that two condenser coils on row C have fin damage and the ambient hotspot above the chiller bypass is creeping. That’s the photo you can deliver, on a quarter you can write into a service contract.

The contract you’re pitching for

You’re not selling drone flights. You’re selling quarterly thermal inspections of the roof, condenser deck, and cooling-loop intake/exhaust zones, delivered to their existing monitoring stack. The pitch length is two minutes:

  1. Quarterly flight, same time window, same flight plan.
  2. Per-asset thermal record: each rooftop unit, each condenser bank, each CRAC return. You key your photos to their asset IDs.
  3. Anomaly classification: hot spots above ambient baseline, fin damage signatures, hardware-failure thermal patterns.
  4. Delivery in the format their dashboards already ingest — JSON for PRTG/Datadog, ServiceNow tickets pre-filled, raw GeoTIFFs for the facilities engineer who still wants the source data.

The facilities team writes that contract once. You fly it every 90 days for four years. They never re-RFP it because switching costs hurt them more than they hurt you.

What the facilities engineer will ask you in the first meeting

These are the questions, in order. Have the answers ready.

“How do you key photos to our asset IDs?” You ask them for the facility’s CMMS asset list (typically a CSV export from IBM Maximo, Hippo, Brightly, or homegrown SQL). You import that as the project’s asset key. Every flight’s photos auto-link to the asset they covered.

“What if a unit is replaced between flights?” Their CMMS already tracks asset lifecycle. Your tooling reads the asset list at the start of each flight — new assets get linked automatically, retired assets stop appearing on the deliverable.

“Will this integrate with [their monitoring tool]?” PRTG and Datadog both accept webhook ingestion of structured anomaly records. ServiceNow has a REST API for ticket creation. You hand the engineer a sample JSON payload from a past flight; they map it to their schema in 15 minutes. You do this once per customer, not per flight.

“What’s the time from flight to alert in their queue?” Honest answer: ~45 minutes. Flight lands at 10:00, photos sync 10:05, GPU processes the thermal mosaic by 10:35, anomaly classifier runs, your delivery hook posts to their stack by 10:45. They like that number because it’s faster than their own sensor-debounce window.

Pricing this contract

Raptor Maps and similar utility-pole platforms charge per megawatt or per pole. That model doesn’t fit data centers — the megawatt count doesn’t map to inspection complexity, and per-pole is meaningless for rooftop work.

Price it as a flat quarterly fee + per-anomaly handoff. The flat fee covers your flight time, mission planning, and the standard deliverable. The per-anomaly handoff covers ServiceNow ticket creation, thermal evidence packaging, and your engineer (or your customer’s) doing the diagnostic write-up.

For a single hyperscale facility, that’s typically a $4-8K quarterly base + $50-150 per confirmed anomaly. Your customer’s facilities team pays that out of their preventive-maintenance budget, not their drone budget. The fact that it’s drone-collected is incidental to them.

Why this is your contract to lose

Three reasons the facilities team will choose a small pilot over a national vendor:

  1. Local response when their alert fires. If a roof RTU drops overnight, they want a re-fly in 24 hours, not 14 days. A regional pilot is closer than a national vendor’s nearest dispatch.
  2. Same crew, every flight. Continuity means the operator who flew last quarter knows what looked normal. Pattern recognition you can’t write into a checklist.
  3. No vendor-lock on data. Their GeoTIFFs are theirs. If they switch to a competing platform two years from now, the source data ports. Big-vendor lock-in is a real reason internal champions get overruled by procurement; lead with this.

What to bring to the second meeting

  • One pre-existing flight’s thermal deliverable in their preferred format
  • A 30-second screen recording of your tooling auto-keying photos to a sample asset list (use a fake list — they’ll get it)
  • The proposed service contract in plain markdown, not a 14-page PDF

That’s the close. The facilities engineer will fight procurement for you if you make the first meeting feel like the work is already half-done.

Run the same thermal stack on your data-center contract: SkySuite’s solar inspection vertical — the radiometric mosaic + anomaly classifier was built for solar but works on any thermal-capture deck.