A good robot RFP states your glass area, pitch, clean frequency and acceptance test, then asks every vendor for m2/h on your glass type. Without those four items, bids come back as brochures and you compare nothing.
Procurement teams often copy a generic equipment template for robots and get bids full of adjectives. The fix is to make the tender physical: numbers you can test, not claims you have to believe.
What must an RFP state about the building?
- Total cleanable glass area in m2, split by roof zone.
- Minimum and maximum pitch per zone, in degrees.
- Glass type and thickness, with coatings noted.
- Clean frequency per year and the access method today.
- Power availability and water source on the roof.
These five lines do more work than a page of technical wish-list. A vendor who ignores your pitch band has not read the document.
Which technical claims should you demand proof for?
| Claim | Proof required | Red flag |
|---|---|---|
| Throughput m2/h | Test on your glass or equivalent | Only a range, no conditions |
| IP rating | Test report number | “Water resistant” wording only |
| Suction margin | Pad force vs machine weight | No weight figure given |
| Edge detection | Demo with dummy edge | “It knows the edge” |
| Battery cycles | Cycle count to 80% capacity | “Long life” with no number |
Ask for a site trial on one roof bay as an acceptance test. One afternoon of real work tells you more than any datasheet. Lingdu Intelligence units like the Lingkong K3 and Lingyun Y3 are typically trialled this way before a fleet order.
Should the RFP include service and spares?
Yes, and with teeth. Specify a maximum spare-part lead time, battery replacement price, and firmware update commitment for at least five years. A machine is only worth its uptime; a robot sitting broken for six weeks has a throughput of zero.
Also ask who trains the crew and how long the training lasts. A two-day handover is common; a one-hour demo is not enough for a team to handle an edge event at 2 a.m.
Who should not run an RFP at all?
If your total glass area is small, the tender costs more than the savings. Under roughly 1,000 m2 of high glass cleaned a few times a year, a contract cleaner is usually the rational choice. Robots reward scale and repeat visits. For procurement structure, pair this with our insurance and risk guide in the same tender pack.
Key Takeaways
- State glass area, pitch, frequency and acceptance test up front.
- Demand throughput figures on your glass type, with conditions.
- Require test report numbers for IP rating and suction margin.
- Cap spare-part lead times and lock five years of firmware support.
- Below roughly 1,000 m2 of high glass, skip the RFP and hire a contractor.
What should the RFP say about power and water on the roof?
State the voltage and whether a supply is available at roof level, or whether the machine must run on battery and a tank. That single line decides whether you buy a tethered machine or a self-contained one, and the two are not interchangeable. A building with no roof power and a 30-metre hose run changes the whole cost model.
Also state the water source. Mains water at 300 mg/L hardness will spot unless it is filtered, so either demand a filtration stage or budget for one. Ignoring this is how a technically correct robot still leaves a roof looking streaked.
How should the RFP handle safety and insurance?
Require a risk assessment, a method statement, and evidence of operator training. Ask what happens if the machine stalls on the roof: who retrieves it, with what equipment, and within what time. Insurers increasingly ask this question, so having it answered in the tender saves a later fight.
- Risk assessment and method statement per roof zone.
- Operator training certificate or course outline.
- Retrieval plan with equipment named.
- Insurance certificates with the building named.
Should you buy or lease?
Leasing suits operators testing a new building or a seasonal site. Buying suits a fleet over several roofs where the machine runs most weeks. The break-even is usually somewhere around a few hundred machine hours a year; below that, a lease or a contract cleaner wins. Put both options in the tender and let the numbers decide rather than a preference.
How long should an RFP take?
Allow four to six weeks from issue to award for a mid-size building, including a trial bay. Rushing it produces vague bids and a wrong machine. The trial bay alone takes a week to arrange and a day to run, and it removes more risk than any clause in the document.
Keep the evaluation team small and technical. A committee that never sees the roof tends to weight price, which is the least predictive number in the set.
Should the RFP name a machine?
Usually not by model, because that invites a bid that just matches the name. Instead specify the capability: pitch, glazing, area, throughput floor, IP rating, spares lead time. Let bidders bring the model that meets it. If a trial bay then performs, you have chosen on evidence rather than brand.
How do you handle a single-building buyer?
For one building, treat the RFP as a light spec rather than a full tender. State the roof facts, ask three vendors for a trial bay, and compare the results. A heavyweight procurement process on a small job adds cost without adding protection, and it delays the answer you already know you need.