Why published benchmarks mislead
The industry averages you can look up describe a blend of businesses that don’t work like yours. Here is how badly, and what MonoBridge uses instead.
Three businesses wearing one code
Government industry codes are filing categories, not business models. The code the sample business files under is a catch-all: it holds hosted phone companies, satellite ground stations and the backbone services of large carriers. Its official examples still open with dial-up Internet.
Hosted voice, UCaaS and cloud phone systems that ride the customer’s own broadband, sold per seat each month. They own a phone platform, carrier connections and a customer book, but no network. This is the sample business’s economy.
Satellite earth stations, tracking, telemetry and radar run for others under contract. Heavy equipment, double-digit capital spending, and often one customer. Its numbers are averaged into every published statistic for the code.
Carrier and backbone services of large firms coded here: 52.9% of the code’s 2022 receipts, reported by only 302 establishments. It is why the code looks concentrated, and it is benchmarked nowhere in the dossier on purpose.
Average those together and you get a number that describes no real business. Then measure a small phone company against it, and you can be told your spending is fine when it’s high, or your margin is weak when it’s normal.
What the published numbers get wrong
How to read it. Grey is what a generic benchmark file says; orange is what a phone company that owns no network actually spends. The published middle is almost five times the real one.
The published capital-spending figures come from businesses that own antennas and fiber. A hosted provider spends almost nothing on equipment: one public provider spent about $18,000 on $68M of revenue in 2025. Benchmark a small provider against the published figure and its free cash flow looks better than it should, or its spending looks like neglect.
How to read it. Owners of their own platform and white-label resellers sit about 18 points apart. The blended median falls in the empty space between them.
What MonoBridge does instead
- A screening test places the business in the right economy and the right kind of business before any benchmark is read.
- Each kind of business gets its own low, middle and high values, so a reseller is never measured against platform owners.
- Every benchmark carries its basis, observed or constructed, and a grade; constructed ones are labelled as scaffolding until a second source exists.
- The report prints the dossier’s weakest evidence in its own words, so nobody mistakes a judgment for a measurement.
The weakest evidence, as the dossier records it
- No private ratio panel exists for hosted voice; every figure for that economy is scaled from five public platform owners plus federal counts.
- The link between company size and churn rests on a 2018 study: the direction is firm, the level is dated.
- Customer acquisition cost and its payback are derived, not observed.
- Price drift of −1% to −3% a year and long-run growth of 2–3% are judgment, with no measured series.
- The 3% prior for removal from the Robocall Mitigation Database is judgment against an unknown denominator.
- Inherited data. 24 metrics from sector 51 and 8 from subsector 517; none sits at the 4-, 5- or 6-digit node
- Capital spending. Published 3 / 7 / 14% of revenue against 0.5 / 1.5 / 3.5% for an over-the-top provider
- Margins. Own-platform service gross margin about 68%, reseller about 50%, blended median 62%
- Carrier and backbone. 52.9% of 2022 receipts ($11.24B) from 302 establishments; valued on facilities-carrier economics and kept out of the other economies’ numbers
See it on a real report, or see it on yours.Download a sample reportSchedule demo