There is a particular kind of RFP that arrives specifying 1,840 modules of a named brand, a named inverter model, and a mounting system, and then asks bidders to price it.
It feels rigorous. It produces bad outcomes. You have just converted a design competition into a procurement race on components you selected without the benefit of anyone's engineering input, and every bidder now competes only on installation margin.
The useful structure is: here is what we need to achieve, here are the constraints that are not negotiable, propose how you would do it.
The parts worth specifying tightly:
The parts worth leaving open: module and inverter selection, DC:AC ratio, string configuration, mounting approach. Those are engineering decisions, and you are hiring engineering.
Require every bidder to submit their full loss table and the simulation file. Two consequences: you can compare like with like, and bidders know their assumptions will be examined, which quietly disciplines the optimism.
Every bid should contain a section headed "not included in this price", and it should not be allowed to say "as per standard scope". List the items you specifically want addressed – structural works, interconnection costs, roof warranty coordination, crane and access, out-of-hours working, monitoring subscription beyond year one.
A bidder who returns that section blank has either included everything (ask them to confirm in writing) or not thought about it.
If price is 80% of the score, you will get the bid that excluded the most scope. A workable split for commercial projects:
| Criterion | Weight | What you are actually testing |
|---|---|---|
| Price | 35% | Normalised, after scope gaps are closed |
| Technical design | 25% | Did they engineer this roof or apply a template? |
| Yield model credibility | 15% | Are the loss assumptions defensible? |
| Relevant experience | 15% | Similar size, similar building type, contactable |
| O&M capability | 10% | Who fixes it in year six |
A two-week window on a 500 kW commercial project guarantees desktop bids. Nobody sends an engineer to climb your roof for a tender they have twelve days to price and a one-in-four chance of winning.
Four to six weeks, with a mandatory site visit, changes who bothers to respond and how carefully. It also filters out the bidders who were never going to do the engineering anyway.
Walk the roof with each bidder. Notice who brings a meter and who brings a brochure. Notice who asks about the switchboard and who asks about the decision timeline. Notice who mentions a problem you had not spotted.
You will learn more in ninety minutes on the roof than from the entire written submission. The document tells you what they want you to believe; the visit tells you how they work.