Methodology · published in full
How we recommend
Every rule Find providers follows — deal-breakers, weights, confidence and stability — with the numbers read straight from the code that runs it.
From your answers to your matches
- 1Classify on facts
Every listing is described with the same controlled vocabulary: problems solved, industries, channels, markets, languages, deployment, connectors, pricing. A facet is recorded only when a source confirms it.
- 2Apply deal-breakers
A listing is left out if it does not serve your market, lacks a deployment model you require, solves none of your problems, is built for another industry, or misses something you marked as a must-have.
- 3Score only what you answered
The rest are scored as a weighted share of the questions you actually answered. A blank question never penalises anyone; marking something important doubles its weight.
- 4Report confidence and stability
Alongside each score: how much of the listing has been checked, and whether its place survives different weightings.
- 5Explain every result
Each result carries why it fits, what it does not cover, what to check before signing, and the per-dimension breakdown behind its score.
Weights and deal-breakers
Base weights out of 100 when every question is answered. A dimension you mark as important counts ×2.
- Only answered questions count; leaving one blank never lowers a score.
- A product built for any industry qualifies everywhere, at half the industry credit of one built for yours.
- A ready-made connector to your system counts in full; reaching it through the vendor's API counts half.
- Ties break on how much of the listing has been checked, then by name — never by any relationship.
Try the weights yourself
Pick a mission and move the weights. The real catalogue re-ranks live — so you can see how much any one judgement matters.
- 3A2P Network72
- 4Bird (formerly MessageBird)72
- 5CM.com72
- 1Gupshup94
- 2Infobip94
- 6Sinch72
The details
Fit versus confidence
Fit says how well a listing's recorded facts match your brief. Confidence says how much of that record has been checked. They are kept apart on purpose: a listing we have not finished checking is not pushed down for our backlog, and it does not look as certain as one we have.
- High confidence
- Verified, or at least four of six criteria assessed with two or more sources.
- Medium confidence
- At least two criteria assessed, or two or more sources attached.
- Early data
- Classified, but little has been checked yet. Treat the score as provisional.
Stability
Weights are a judgement, so a recommendation that only holds under exactly our numbers is fragile. For every shortlist the ranking is recomputed under 200 alternative weightings, each weight moved independently by up to ±40%, using a fixed seed so the figure is reproducible. 'Stays top-3 in 92% of weightings' means exactly that.
Profile shape
The radar on each listing shows six breadth measures computed from its recorded facts. Two listings with the same facts draw the same shape. It describes coverage, not quality — a narrow specialist can be the right choice.
- Channels
- Distinct channels carried; eight or more fills the axis.
- Reach
- Widest region served: global 100, a macro-region 70, a sub-region 45, one country 25.
- Connectors
- Confirmed ready-made connectors, plus one for a public API; six or more fills it.
- Deployment
- Deployment models offered; three or more fills it.
- Price clarity
- Published or free to start 100, published subscription 80, quoted 25.
- Criteria shown
- Criteria shown met, out of six. Unassessed counts as not shown — hence 'shown', not 'quality'.
What the engine does not do, and its limits
What the engine does not do
- It does not use reviews, popularity, traffic, revenue or brand size. A large vendor and a small one with the same facts score the same.
- It does not guess. An unclassified listing is never matched; an integration we could not confirm is treated as "possible through the API", not as a connector.
- It does not rank on assessment. Criteria verdicts feed the confidence level shown beside a score, never the score itself, so our own backlog cannot push a vendor down.
Limits worth knowing
- Coverage is not capability. A facet says a vendor offers something, not how well. Use the shortlist to decide whom to talk to, then test.
- Facts age. Each listing shows when it was last reviewed. Prices, channels and coverage change; the community corrections log is how we keep up.
- The catalogue is curated, not exhaustive. If a vendor you need is missing, suggest it on the community page.
Every attribute, measured the same way (43)
Every fact the ecosystem holds is one of these attributes, with one rule that turns a listing's record into a value. The rule never names a listing. A value is either recorded, not recorded yet, or not applicable — and only the ranking engine's own inputs, your answers, ever change an order.
What it does
- Use cases list
- The problems the listing's own documentation shows it solving.
- Channels list
- Channels carried, from the listing's documentation, normalised to one vocabulary.
- Type of solution list
- What kind of product it is; a listing can be several.
- Industries list
- Industries it is built for; "Any industry" for general-purpose products.
- Suits list
- Derived, never tagged: developers if it offers an API; small businesses if it serves them ready-to-use; marketing, support or product teams from its use cases; larger organisations if it serves enterprise or government.
- Features list
- Derived from the record by one rule per feature: e.g. delivery status when the delivery check is met, a shared inbox when the listing solves scattered channels.
- Number of channels number
- Distinct channels carried.
Where, and for whom
- Markets list
- Where it says it operates. A region covers its countries.
- Reach number
- The widest area served: worldwide 100, a region 70, a sub-region 45, one country 25.
- Languages list
- Languages the product or its support is offered in, where documented.
- Company size list
- The customer sizes it targets.
Setup and connections
- Deployment list
- How it can be used: ready-to-use, by API, self-hosted, white-label or inside your environment.
- Has an API yes / no
- Yes when a programmable API is one of its deployment options.
- Usable without coding yes / no
- Yes when it is offered ready-to-use or white-label.
- Can run on your own servers yes / no
- Yes when self-hosted or in-your-environment deployment is offered.
- Connects to list
- Ready-made connectors we could confirm; an unconfirmed one is left off.
- Number of connectors number
- Confirmed ready-made connectors.
- Availability one of
- Generally available, invite-only or coming soon.
Pricing
- Pricing model one of
- How prices are offered: published pay-as-you-go, free to start, published subscription, or quoted on request.
- Prices you can read yes / no
- Yes when prices are published or it is free to start.
- Free to start yes / no
- Yes when a free plan or free tier is stated.
The company
- Founded number
- Year founded, only when a source confirms it.
- Headquarters one of
- Country of the head office, only when sources agree.
- Ownership one of
- Publicly listed, privately held, or part of a larger group.
- Parent company text
- Applies only to subsidiaries.
- Stock ticker text
- Applies only to publicly listed companies.
- Same owner as cpaas.co yes / no
- A disclosure, shown wherever the listing appears. Never a ranking input.
- Added by one of
- The cpaas.co team, or a community submission the team reviewed.
Our checks
- Operational transparency check
- Met, not met, or not recorded until checked against public evidence.
- Documented interfaces check
- Met, not met, or not recorded until checked against public evidence.
- Delivery accountability check
- Met, not met, or not recorded until checked against public evidence.
- Consent and compliance check
- Met, not met, or not recorded until checked against public evidence.
- Data handling check
- Met, not met, or not recorded until checked against public evidence.
- Commercial clarity check
- Met, not met, or not recorded until checked against public evidence.
- Security and privacy attestations list
- Recorded only when the listing's own trust, security or legal page (or an auditor's registry) states it, with that page as the source. Absent means not recorded, never "does not have".
- Checks met number
- How many of the six checks are met. Unchecked counts as not shown, not as failed.
- How much we've checked one of
- High with four or more checks and two or more sources; medium with two of either; otherwise low.
- Sources number
- Cited sources behind the listing.
- Analyst recognitions number
- Placements in analyst evaluations, confirmed on the publisher's or vendor's own page.
- Last reviewed text
- The month the listing was last checked against its sources.
The community
- "Looks right" confirmations number
- One-tap confirmations from visitors. Shown, never ranked.
- Community reviews number
- Moderated reviews from people who used it.
- Community rating number
- Average of moderated reviews, out of 5. Shown beside the facts; never a ranking input.
The whole matrix is open data: /api/v1/matrix · CSV
The six criteria
Every listing is classified on the same three lenses — business (industries, problems solved, company size, pricing), operational (markets, languages, deployment, availability) and technical (channels, connectors) — and assessed against the same six criteria.
- Operational transparency. Is there a public status page that reflects real incidents, including partial degradation?
- Documented interfaces. Is there a published specification a developer can integrate against without a sales call?
- Delivery accountability. Are failures reported with a cause, and is there a defined retry and escalation path?
- Consent and compliance posture. Is opt-out handled at the platform level, and is consent tracked per channel?
- Data handling. Are storage locations, retention periods and sub-processors documented?
- Commercial clarity. Is pricing published, or at least explained without obligation?
A criterion is met, not met or not assessed — three distinct states. Most listings show several "not assessed", including large and reputable companies: marking a criterion met on reputation would make the scorecard decorative. Not assessed is not a failure; it says we have not checked yet.
Listing states. Verified: assessed against all six criteria, with evidence recorded. Listed: included because it is materially relevant, not yet fully assessed — a listing is not an endorsement. Under review: an assessment is in progress or a question is open.
139 of 432 possible criterion verdicts recorded across the catalogue so far.
Independence
The engine never reads ownership or any relationship field. Products that share an owner with cpaas.co are listed on the same terms and disclosed; any other commercial relationship is excluded from the public catalogue. The automated test suite relabels every listing with every possible relationship and fails if a single ranking moves.