What a free 90-day building analytics pilot should prove

Photo: Industrial plant-room piping and machinery. Photo by Dave Meckler on Unsplash.

A free building analytics pilot sounds cheap. The scarce resource is your team's time. Facilities, energy and property operators who say yes still have to grant access, answer mapping questions, triage findings and own a day-90 call. That is real work. The only honest reason to spend it is a defined proof: what must be true at day 90 for the pilot to earn a continue, and what you will do if it does not.

This note is the operator checklist. It is written for people deciding whether a free 90-day window on one building is worth the calendar, not for ranking pages that invent search volume. FrostLogic Explore publishes a free 90-day pilot on one building. The rules below are what that offer should be measured against, whether you run Explore or another stack.

Why most building analytics pilots fail

Pilots fail for boring reasons. Success is never written down, so day 90 becomes a feeling. Data access stalls for weeks while someone waits on a vendor login or a firewall ticket. The product lands as another dashboard, and operators tour charts without a ranked queue of work. Nobody owns the continue-or-stop decision, so the window expires into polite silence.

Dashboard tourism is the quiet failure mode. Your BMS already has graphics. A pilot that only adds more tiles has not proved anything an operator can act on. A pilot that never gets point history or stable tags cannot tell signal from mapping debt. A pilot without a named decision owner will not produce a commercial call; it will produce a meeting that ends with "let's keep looking."

None of that is about algorithm quality. It is about scope, access and governance you can state on day 0.

Scope the building, the systems and the decision owner

Keep the first window to one building. Multi-site rollouts belong after a single site has earned its place. Pick a building where the BMS, meters and the important sensors you already trust are live. The point of an Explore-style pilot is connection to equipment on site, not a hardware purchase.

Name the systems in scope before week 1. Typical first cut: HVAC and related setpoints from the BMS, main electrical or thermal meters, and any zone or plant sensors already feeding the head-end. Leave exotic IoT sidecars out until the core feed is clean. Write who owns the day-90 decision: facilities lead, energy manager, or a joint ops/property pair. If that name is missing, stop and fill it before you open a VPN.

Also write who grants write scopes later. Reading points is the default start. Permissioned writes into Automation are a separate decision, not an automatic upgrade on day 30.

Data you need before week 1

Treat data access as a gate, not a nice-to-have. Before the clock matters, confirm:

• Point or tag access to the systems in scope (BACnet, Modbus, OPC UA, oBIX, Niagara station traffic, or an Edge Agent path that can reach them).
• Enough recent history to baseline, if the BMS or historian already holds it. Missing history does not kill a pilot, but it stretches the observe phase.
• A named contact who can approve firewall rules, service accounts and, later, write scopes.
• Mapping ownership: someone on your side who can confirm AHU, FCU, meter and zone labels when tags are ambiguous.

Explore connects to equipment already in place. No start fee and no hardware for the public pilot. That only helps if the existing points are reachable. If access will take eight weeks, say so and delay the start date rather than burning calendar against a dark feed.

Success criteria to write on day 0

Write the bar before the first anomaly lands. Categories matter more than vendor slides. Operators set their own thresholds; do not accept a savings percentage as the proof, and do not ask Explore (or anyone else) for an invented ROI figure as the pass/fail.

Useful categories:

1. Data completeness and connection quality. Are the scoped points live, mapped and stable enough that findings are about the building, not about broken tags?
2. Actionable findings in a prioritised queue. Not another dashboard. Can ops open a short ranked list with evidence and decide what to chase this week?
3. Operator time. Can the team triage and act inside normal hours without a side project that starves the rest of the estate?
4. Evidence quality for continue/stop. At day 90, can the decision owner point at a pack of findings, actions taken and open questions without reconstructing the story from screenshots?
5. Optional comfort and ticket side-effects. If you open permissioned writes, do comfort complaints and ticket noise stay inside the envelope you already accept for that building?

Refuse savings-% promises as the pilot's north star. Cost impact may appear in evidence for individual findings. A guaranteed percentage across the building is marketing, not a success criterion you can audit at day 90.

Cadence across 90 days

A clean 90-day shape looks like this. Adjust the weeks to your access reality; keep the sequence.

Connect (early weeks). Bring the scoped points into the building model. Fix mapping debt that would poison every later finding. Confirm the feed is live end to end.

Baseline and observe. Let the model watch schedules, setpoints and plant behaviour long enough that anomalies have context. This is where history depth helps, and where "we connected yesterday" is not yet a verdict.

Investigate. Work the prioritised queue. Close or defer items with notes. Measure whether ops can use the queue without living in a second BI tool.

Optional permissioned writes. Explore starts read-only by default. Write access starts off. The FrostLogic Edge Agent can write after human review, or automatically inside permissioned scopes only, once you grant those scopes. That story sits on Automation. Do not treat write-back as the definition of a successful pilot; many buildings prove the queue first and open scopes later. Explore is Sensor Intelligence with optional Automation. It is not a CMMS. Work orders and contractor tickets stay where you already keep them.

Decision pack (last stretch). Assemble what the decision owner needs: connection quality, queue examples with evidence, time the team spent, write scopes opened or refused, and the continue-or-stop recommendation.

The day 90 decision

Day 90 is a fork, not a soft extension by default. Four honest outcomes:

• Stop and keep the findings. The window did not earn a subscription. You keep the evidence pack. That is a valid result when the product failed your written criteria, or when the building was the wrong first site.
• Continue on AI Automation. The queue and optional permissioned writes earned a place at the flat AI Automation list price.
• Continue on AI Maintenance. You want the fuller Sensor Intelligence / maintenance-oriented band priced by sensor count.
• Enterprise conversation. The estate is larger than about twenty buildings and needs a portfolio quote rather than stacking single-building list prices.

Write which fork you expect to debate before the pilot starts. Surprises at day 90 are usually governance surprises, not model surprises.

How FrostLogic's public offer maps

Facts only, EUR ex VAT, billed annually, as published on pricing:

• AI Automation: €1,500 per building per year, flat, any building size.
• AI Maintenance: from €5,450 per building per year by sensor band (50–500 → €5,450 / €4,550 on a 3-year deal; 500–2,500 → €6,800 / €5,450; 2,500–10,000 → €11,350 / €10,000; over 10,000 → €14,750 / €12,500).
• Enterprise (more than 20 buildings): quoted per portfolio.
• Free 90-day pilot: one building, connected and running 90 days at no cost. A FrostLogic engineer handles onboarding. If it does not earn its place, you keep the findings.
• No start fee, no hardware. Self-host is available on either plan (one-time setup fee plus server costs if you need them).

Use those numbers after a successful pilot; do not invent parallel commercial language. The public page is the source of truth when figures move.

What one anonymised pilot looked like

In July 2026 an anonymised Dubai office on Tridium Niagara ran as a pilot, not as a named customer reference. The FrostLogic Edge Agent connected about 800 BMS points. Cooling and ventilation anomalies landed in a ranked queue for ops. Later, ops released permissioned write scopes so Explore could write to FCUs and the AHU inside those limits. Write access started off; scopes were granted on purpose.

The full write-up is the Field Note What Explore found in a Dubai office in its first weeks. Use it as a shape check for early weeks: points mapped, queue usable, writes optional and leashed. Do not treat it as a savings case study. It is a pilot narrative with a short allowed fact set.

What to do next

If the checklist above matches how you already buy software for buildings, two concrete next steps:

1. Read the live list prices and plan boundaries on pricing.
2. Ask for a one-building check via Check one of my buildings when you want FrostLogic to map the free 90-day window onto a site you already run.

A pilot is worth the team time only when day 0 writes the proof and day 90 owns the fork. Everything else is calendar noise.

Frequently asked questions

What does the free 90-day building analytics pilot cover? One building, connected and running for 90 days at no cost, with FrostLogic engineering on onboarding. You keep the findings if you stop. It does not include a portfolio rollout, custom integrations outside the normal connect path, or a guaranteed savings percentage.

What data access is needed before week 1? Reachable points for the systems in scope, a path to the BMS/meters/sensors already on site, enough history to baseline when it exists, and a named contact who can approve access and later write scopes. Delay the start if access is not real yet.

What happens at day 90? You decide: stop and keep findings, continue on AI Automation, continue on AI Maintenance, or open an Enterprise conversation if the estate is larger than about twenty buildings. Write the decision owner on day 0 so day 90 is not improvised.

How do list prices work after a pilot? Published EUR ex VAT annual prices on pricing: AI Automation at €1,500 per building per year flat; AI Maintenance from €5,450 per building per year by sensor band (with 3-year band prices listed on the same page); Enterprise quoted above twenty buildings. No start fee and no hardware on the public offer.

What is the difference between a building check and a full pilot? A building check is the conversation and scoping step (Check one of my buildings). The free 90-day pilot is the connected window on one building once access and scope are agreed. Do not treat the form as a substitute for the running pilot.

Is new hardware required? No. Explore connects to the BMS, meters and sensors already on site. Self-host is optional on either plan if you need it later; it is not a hardware kit for the pilot.

Does Explore replace our CMMS? No. Explore ranks findings and can run permissioned Automation against the BMS. Work orders, spares and contractor workflows stay in your maintenance system.

Does write access start on day one? No. Explore starts read-only by default. Write access starts off. The Edge Agent writes after human review or automatically inside permissioned scopes only, when you grant those scopes.