Short answer: every cell on our /compare comparison pages grades one named, categorical question against a published source a reader can check — the vendor's own product pages, documentation, or pricing page, retrieved and dated. The vocabulary is deliberately narrow ("Yes / No / Partial / Yes — parity"), numeric competitor figures are never asserted without a public page behind them, and a vendor's carve-out gets named in prose beside the table rather than flattened away. This post is the published version of the method, so a vendor, a buyer, or our own editors can hold every page to the same rules.
A comparison page that cannot show its work is marketing copy with a table attached. The pages we publish are only as trustworthy as the discipline behind them, so this is the discipline, stated once, for every /compare page we ship.
1. The claim shape: categorical, checkable, current
A cell earns its grade only when it answers a categorical question — "Does the vendor publish pricing?", "Does the vendor ship a first-class API for this channel?", "Is self-serve signup available without a sales call?" — against a public source. That shape is deliberate:
- Categorical, not numeric. Where a competitor's price figure or private metric would be needed to grade a row, the row reads as a posture state ("published pricing exists — check the vendor page") rather than a number we cannot verify. The Orange Business head-to-head states this verbatim, and it is the rule on every page.
- Source-cited, dated. Each page's footer names what fed the table — the vendor's catalogue, docs, and pricing surfaces — with a retrieval month. When a post mirrors a table, it re-states that retrieval stamp in the prose, as the Aircall head-to-head does. A date-less claim about another company's product is a claim nobody can audit.
- Disclosed where we are in frame. On pages where Orbit is one column of the table, the same page that hosts the comparison hosts the methodology disclosure. Nothing about that layout is hidden.
2. The vocabulary: four grades, no free-form prose
A cell can only be one of four values, and incidentally that constraint is what keeps claims honest:
- Yes — a published capability, checkable on the vendor's own pages tonight.
- Partial — a published capability with a real caveat: a sales gate on the signup path, a scoped product, an add-on purchase behind the base plan, or a developer surface thinner than a first-class API. "Partial" credits a caveat; it never disguises a gap as parity. When a row credits the vendor fully, the table says so, because Aircall's purchasable add-ons genuinely earn more credit than most incumbents can claim.
- No — an honest gap, in either direction. A "Partial" on the competitor side means exists with a caveat, not failed at a thing it never claimed; and sometimes the "No" row belongs on our side of the table, which the page then says plainly.
- Yes — parity — a tie vote. Where the competitor publicly ships the same named capability we do (a carrier group validating SIM-swap on its own network is the current canonical case), the cell names it as parity rather than leaving the favourable-every-row shape a one-sided marketing read.
That vocabulary is the entire grammar of the tables. Anything a row needs to say beyond it lands in prose beside the table — the carve-out paragraph in the Orange Business head-to-head is the pattern, and it is the pattern precisely because flattening a real carve-out into a cell would be a false claim.
3. Orbit-side cells: shipped capability, never aspiration
The same discipline runs on our own column. An Orbit "yes" cell maps to a shipped capability — voice, SMS/MMS, WhatsApp, RCS, email, embeddable video, AI voice agents, a built-in contact center, a native customer data platform — and a published commitment like a latency budget gets footnoted on-page with the page that substantiates it. Where a capability is a genuine platform commitment but the on-page surface is deliberately conservative (a healthcare Business Associate Agreement granted case-by-case, for instance), the footnote points at the conservative page, not at a stronger claim the page never made.
That direction matters: the table never grades our own column with a claim the substantiating page would refuse to carry.
4. Retire and refresh: how stale cells die
Sources age. The registry that backs the pages treats a stale reference as a broken cell — when a vendor's pricing page reshapes, when a catalogue re-scopes, when a capability moves from "published" to "sunset," the retrieval stamp on the page footer re-runs and the cells re-grade. Two failure modes get named explicitly so they cannot hide:
- A retrieval stamp that no longer matches the vendor page means the page stops asserting until the table re-verifies; dated sources are the detection mechanism, not decoration.
- A capability that sunset since the retrieval date downgrades to an honest "No" in the next refresh; it never stays a "Yes" carried on a dated source.
5. The correction path
If a vendor or a reader finds a cell mis-graded — a published source the row missed, a caveat the row flattened, a retrieval stamp gone stale — the fix path is the same one the page itself uses: point us to the public source, and the cell re-grades. There is no appeals process, no editorial discretion layer between a public source and a cell. The whole method reduces to that one commitment: a cell answers a categorical question from a dated public source, and the vocabulary above keeps it from pretending otherwise.
6. Why this matters for a buyer
A buyer comparing vendors on opaque tables inherits the vendor's marketing posture as their own research. A buyer on a sourced table inherits the vendor's published pages instead. The whole point of this discipline is to make the second posture the only one we ship — and to leave a rebuttable audit trail on every cell, so when the claim is wrong, finding it takes minutes, not a sales cycle.
Frequently asked questions
Where do the numeric claims come from?
Numeric claims about a competitor's pricing are never asserted unless the vendor's own pricing page publishes them. Where the page would need a private figure to grade a row, the row reads as a posture state ("published pricing exists — check the vendor page") instead of a number we cannot verify.
What does a "Partial" grade mean?
"Partial" credits a published capability with a real caveat: a sales gate, a scoped product, an add-on purchase, or a thinner developer surface. It never disguises a gap as parity; where a competitor genuinely matches us, the cell reads "Yes — parity" instead.
How do the pages stay honest over time?
Each page's sources are dated at retrieval, and a stale source stops the table from asserting until it re-verifies. A dated source is the detection mechanism for staleness — without it, a claim about another vendor's product is a claim nobody can audit.
Can a vendor dispute a grade?
Yes, and the path is the same one the page itself uses: a public source. If the source shows the cell mis-graded, the cell re-grades. There is no editorial discretion layer between a published page and a cell.
Why does Orbit score itself on these tables at all?
Because a comparison that excludes the author is a comparison the reader cannot use for a decision. Where Orbit is in frame, the method is disclosed on the same page, and the Orbit column grades shipped capability only — a footnote links any commitment claim to the page that substantiates it.