What 197 public-sector RFPs reveal about how agencies actually buy

188 of 197 solicitations (95%) required references from comparable public-sector contracts. Not references generally. References from public bodies, usually of similar size, usually within the last three to five years, usually with a named contact and a working phone number.

This is the requirement that quietly decides most competitions, and it is almost never the thing bidders worry about. Firms agonise over pricing and presentation, then submit a proposal whose reference section lists three commercial clients because that is who they have worked for.

An evaluator scoring that submission does not need to reach the pricing envelope. The scoring matrix has already done the work.

The uncomfortable implication is that public-sector work is partly a closed loop: you need public-sector experience to win public-sector experience. It is not entirely closed (smaller contracts, cooperative vehicles and subcontracting all break the cycle), but any strategy that ignores the reference requirement is a strategy for losing slowly. If you have never held a public contract, your first target should be the smallest one you can plausibly service, not the most lucrative one you can find.

We classified every solicitation by what the bidder would actually need to assemble a responsive submission. The results:

109 of 197 (55%) required something the bidding firm could not produce by itself. Rate tables that only a carrier can generate. Network reports. Claims-handling data. Platform integration confirmations. A firm without that partner is not a weak bidder on these; it is a non-responsive one.

A second, broader flag in the same dataset marks an opportunity as not cleanly solo-biddable when the documents leave real doubt: borderline cases where a partner is probably but not certainly required. On that reading the figure is 118 of 197, or 60%. Either way the conclusion is the same: most of what looks attractive in a feed is unbiddable for the company reading it, and nothing in the listing says so. You find out on page sixty-one, after you have spent an afternoon.

The practical move is to establish which category you are in before you look at a single opportunity, and filter on it. Roughly 43% of this market is open to a firm bidding on its own capabilities. The rest requires a relationship you either have or need to go and build.

We sorted each solicitation by how much work a compliant response actually takes:

Only 14% were what most people picture when they imagine bidding: fill in the forms, attach your qualifications, submit. The majority demanded a structured technical response, sometimes eighty-five or more discrete questions covering claims processing, data security, breach notification timelines, network adequacy, provider directory accuracy and subcontractor disclosure.

Industry benchmarks put a single response at around 33 hours of work across roughly seven contributors. Our reading suggests that is a reasonable average for this population and a significant underestimate for the top quartile.

That number is why so many firms bid twice a year. It is not lack of ambition. It is that a compliant response to a serious solicitation is most of a working week, and it competes directly with billable work.

Counting requirements across all 197 solicitations:

Two of these are worth dwelling on.

Certificates of insurance, in 36% of cases, had to be submitted with the proposal, not on award. General liability, auto, workers' compensation and professional liability, often naming the buyer as additional insured. Producing a correctly endorsed certificate takes a broker several days. A firm that starts on the deadline week will miss it, and a missing certificate is frequently an automatic rejection rather than a scoring deduction.

Named platform integration, in 8% of cases, was listed as a mandatory element. Only the platform vendor can confirm it. If you cannot get that confirmation, you cannot bid, regardless of how good the rest of your submission is.

Both are administrative rather than competitive requirements. Both eliminate bidders who were otherwise qualified. Both are entirely avoidable with two weeks' notice, which is the argument for finding solicitations early rather than well.

Estimated annual values across the set, using conservative sector-based bands:

These are estimates, not quotes, and the distribution is heavily skewed. A handful of very large contracts pull the average far above the median. The median is the honest number: a typical opportunity in this set is worth somewhere around $50,000 to $150,000 a year.

The part that does not show up in a single year's figure is renewal. Public contracts are typically awarded for one to three years with option years attached, and incumbents usually keep them. A $120,000 contract that renews four times is worth closer to $600,000, and it is revenue that does not need to be re-sold every quarter. That is the real argument for the channel, and it is why the reference problem is worth solving early: the first public contract is the expensive one to win.

Every one of the 197 opportunities had at least one real document attached. Across the set there were 788 documents in total: a median of two per opportunity, a mean of four, and a maximum of twenty.

That distribution is the whole problem in one line. Half of these are a quick read. A meaningful minority are a full day's work before you know whether you want the job.

There is no way to tell which is which from the listing. The only way to find out is to open them, which is exactly the cost that keeps most firms from looking at more than a handful a year.

Five things follow fairly directly from the numbers.

Solve references first. It is the most common requirement by a wide margin and the slowest to fix. If you have none, target the smallest public contracts you can service and treat the first one as a qualification purchase rather than a profit center.

Know which side of the 60% you are on. Decide whether you can bid on your own capabilities or need a partner, and filter accordingly. Half the frustration in this market is firms reading solicitations they were never eligible for.

Get the certificates in advance. A standing set of correctly endorsed certificates, refreshed annually, removes an entire category of automatic disqualification for the cost of a phone call to your broker.

Budget the real number. If a compliant response is most of a week, then bidding four a year is a real commitment and bidding twenty is a staffing decision. Pretending otherwise produces rushed submissions that lose.

Ask for the debrief. Most public bodies must provide one on request. It is free, almost nobody asks, and it is the only direct evidence you will ever get about how your submission actually scored. See how to request one.

The sample is 197 open public-sector insurance and employee-benefits solicitations collected between June and August 2026, from schools, municipalities, counties, housing authorities, universities and state agencies across roughly 30 states. Every attached document was downloaded and converted to text; classification was done by reading the text, not the listing summary, and each classification records whether a scope document was actually present.

Three honest limitations.

It is one sector. These are insurance and benefits solicitations. The reference and certificate findings almost certainly generalize (they are administrative requirements common to public procurement), but the partner-dependency figure is inflated by the fact that insurance products come from carriers. Construction, janitorial or landscaping would look different.

Values are estimates. Public solicitations rarely publish a budget. Our figures are modeled from sector and scope, deliberately conservative, and should be read as bands rather than quotes.

It is a snapshot. Volumes and requirements shift with fiscal cycles. A set collected in January would not look identical.

We will repeat this as the dataset grows and across more sectors. If you want the underlying breakdown for a particular state or category, ask us. We are happy to share it.