Orum

Telephony infrastructure with a model layer on top of it, and the model layer is real without being the product. What the customer buys is the ability to place many simultaneous calls, hold the lines open, determine which one a human answered, and bridge a representative into it within a second or two.

Most of that is carrier integration, concurrency management and signal processing, and answer detection in particular is a problem the telephone industry has worked on since long before the current vocabulary. Above it sit capabilities that are genuinely model driven: call summarisation, scorecards, objection detection, and roleplay agents a representative practises against. The acquired phone verification technology adds automated agents that check and refresh numbers.

Strip all of that and a customer still has a parallel dialer, a sales floor, call logging into their record system, and the connect rate that is the reason to buy. What they lose is the coaching layer and the automatic note taking, which are the accessories rather than the engine. The vendor's own positioning supports the reading, since it competes on lines dialled simultaneously and on connect rate rather than on model quality. Ask which capabilities stop working without the model layer, and what powers answer detection.

Strong human oversight of representatives, and nothing published about the control that regulators actually care about. The oversight of people is genuinely good: the virtual sales floor gives managers live listening, spotlighting and room based supervision of a distributed team, scorecards evaluate calls afterwards, and roleplay agents let a representative practise before speaking to anyone real.

Line concurrency is configurable and tiered, at up to five simultaneously on the entry plan and up to ten above it, so a customer chooses their own aggressiveness rather than inheriting a default. The gap is on the dialing system itself.

Simultaneous dialing means the platform routinely places more calls than it has representatives to answer, so some answered calls will find nobody available, and the resulting abandoned or dead air call is the specific behaviour that telephone regulators cap and measure. Nothing published describes an abandonment rate limit, what the system does when every representative is busy, whether an abandoned call plays a required message, or how any of it is reported to the customer.

For a product whose entire value is dialing more numbers than you have people, that is the control a buyer most needs to see. Ask what abandonment rate the system enforces, what happens when no representative is available, and how it is reported.

Five model driven capabilities named and no measurement published for the one that decides whether the product works. Live answer detection, call summarisation, scorecards, objection detection and roleplay agents are all promoted, and the acquired verification technology adds automated phone number checking. No model provider, family or version is named for any of them, no model card exists, and no accuracy figure is published anywhere.

Answer detection is the one that matters and it has a measurable error rate in both directions. A false positive bridges a representative onto a voicemail greeting, wasting the moment and often the recording. A false negative drops a live human who has just picked up the telephone, which from the recipient's side is an unexplained silent call, and repeated silent calls are precisely what regulators and carrier reputation systems treat as abuse.

The vendor holds a call dataset numbering in the billions, so it can measure this better than anyone, and publishes nothing. Ask what the answer detection accuracy is in both directions, measured across what volume, and what happens to a call classified wrongly.

A substantial proprietary dataset, published research drawn from it, and one finding that appears to cut against the product's own premise. The vendor states more than twelve hundred customer teams and a call dataset numbering in the billions, which is a genuine asset and unusual in this category, and it publishes research derived from that dataset rather than only marketing claims.

A competitor selling directly against this product cites that research as showing that meetings per connection fall materially on parallel dialing compared with single line dialing, attributing the drop to the one to two second pause while a representative is bridged onto an answered call. If that characterisation is accurate it is a striking piece of disclosure, since the vendor would be publishing evidence that its central capability trades meeting quality for connection volume.

The underlying document was not retrieved on this pass and the characterisation comes from an interested party, so it is recorded as a question rather than a finding. Beyond that no outcome study with a stated customer population, period and method was located, and the customer figures that circulate come from third parties rather than the vendor. Ask for the vendor's own comparison of meetings per connection between parallel and single line dialing, and for connect rate measured across a stated customer population.

High volume outbound telephone calling is the most heavily regulated activity in this index and almost nothing about that regulation was located. The obligations are specific and well known: suppression against national and internal do not call registers, permitted calling hours enforced by the recipient's own time zone rather than the caller's, caps on the proportion of answered calls abandoned when no representative is free, state level registration and bonding requirements in several jurisdictions, identification of the caller at the start of a call, and consent to record where the recipient's jurisdiction requires all parties to agree.

Every one of those falls on the customer operating the platform, and the platform is what makes compliance either automatic or impossible. Nothing published describes register scrubbing, time zone aware dialing windows, an abandonment cap, recording consent handling or jurisdictional controls. The vendor's own security announcement from 2022 referred to compliance with international privacy and telephony laws as part of a roadmap rather than as a delivered capability.

A competitor selling against the product names regulatory exposure as a cost that reviews overlook. Ask what register scrubbing, calling window enforcement and recording consent controls the platform provides, and which are the customer's responsibility.

Security is described in reasonable technical detail and privacy is barely addressed at all, which for this product are different questions. The vendor's security page covers encryption of sensitive data, encrypted transport, multi factor authentication depending on login method, minimum password complexity, and backup replicated across multiple availability zones. Those protect data once held. What is not addressed is the material this platform holds and on what basis it holds it.

Every conversation is recorded and transcribed, which means audio of prospects who were called without warning and did not choose to participate, retained alongside their telephone numbers. Several jurisdictions require every party to a call to consent before it is recorded, and nothing published describes whether an announcement is played, whether it can be configured by jurisdiction, or how consent is evidenced afterwards.

No processing agreement, subprocessor list, retention period for recordings or transcripts, transfer mechanism or data protection contact was located. A claim to flexible region specific control appears on the security page without any description of what it covers. Ask how recording consent is obtained and evidenced, what retention applies to recordings and transcripts, and what region specific control actually means.

An acquisition has just moved this vendor into the data business and the provenance questions arrived with it. Until recently the product consumed telephone numbers the customer supplied and produced calls. The acquired verification technology changes that: automated agents now check and enrich prospect numbers, and the stated purpose of combining it with the existing call dataset is to give representatives phone data they can trust, which makes this vendor a supplier of contact data rather than only a conduit for it.

Nothing published describes where verified or enriched numbers originate, what licences support them, or what a customer is warranted when a number is supplied. The dataset itself raises the sharper question. A corpus of billions of calls is built from customers dialing their own prospect lists, and using it to improve number quality means one customer's calling activity informs the data another customer receives.

That may be entirely reasonable and disclosed in the terms, and nothing published addresses it in either direction. Ask where verified numbers originate and under what licence, and whether one customer's call outcomes inform data supplied to others.

Official connections throughout, and a dependency that is not a software platform at all. Named integrations cover two customer record platforms, two sequencing tools, a prospecting database, a conversation intelligence product and a contact data provider, each through that product's documented route, and several of those are separately recorded in this index so a buyer can evaluate the pieces independently.

Nothing scrapes, no browser extension harvests a professional network, and no social automation exists. The vendor operates its own dialing infrastructure rather than reselling somebody else's application layer. The exposure that actually threatens continuity here runs through the telephone network.

Carriers and the analytics companies they contract with decide whether a number is labelled as likely spam, that labelling is applied without notice or appeal, and it can render a customer's numbers commercially useless overnight regardless of how the platform behaves. Held below the top band because nothing published describes that relationship: no carrier registration position, no statement on reputation services, and no account of what recourse exists when numbers are labelled. Ask what relationships exist with carriers and reputation services, and what happens when a customer's numbers are flagged.

Recorded conversations with people who did not agree to be recorded, processed by models, with no stewardship position published. Summarisation, scorecards and objection detection all operate on audio and transcripts of live calls, which contain not only the representative's side but the prospect's, including whatever a surprised person says when a stranger telephones them.

Nothing published states whether recordings or transcripts train or tune any model, which provider processes them, what retention applies to derived summaries and scores as distinct from the audio, or whether anything crosses a customer boundary. The roleplay agents raise a smaller version of the same question, since a convincing practice opponent has to have been built from something.

No governance document, evaluation record or independently audited management standard for artificial intelligence was located, and the security page addresses data protection rather than model governance. The acquired verification agents add a further unaddressed flow, since automated checking of telephone numbers implies contacting or querying something about identifiable individuals at scale. Ask whether call recordings or transcripts train any model, which providers process them, and what retention applies to derived summaries separately from the audio.

A real person speaks and the mechanics of getting them there are audible to the recipient. There is no synthetic voice, no agent conducting the conversation and no persona: when a prospect answers, a human representative is bridged onto the line and speaks in their own voice, which places this well above the practices this axis usually records. What the recipient experiences is the machinery around that.

Simultaneous dialing produces a pause of a second or two while the representative is connected, which is the delay most people recognise as marking an automated call, so the method announces itself at the moment of answering. Nothing published describes what a recipient is told at the start of a call, whether the representative or the platform identifies the calling organisation, or whether any announcement is made about recording, which several jurisdictions require before the conversation may proceed.

Recipients who answer when no representative is free experience silence, and nothing published states what they hear. Ask what a recipient hears at the start of a call, whether recording is announced, and what plays when no representative is available.

The right connections for a dialer and no route to build anything else. Named integrations cover two customer record platforms, two sequencing tools, a prospecting database, a conversation intelligence product and a contact data provider, which is a well chosen set: a dialer needs to know who to call, log what happened, and hand the conversation to whatever analyses it. Activity logging and context surfacing back into the record system are stated.

A public status page reports availability with subscribable notifications, which is a real operational commitment. What is absent is the developer surface. No public interface documentation, no webhook description, no marketplace and no partner directory was located on this pass, so a customer whose stack is not on the published list has no visible route, and the product's deliberate narrowness makes that matter more than it would elsewhere.

Because this is a dialer only tool with no messaging, no electronic mail and no inbound handling, it is always one component of a larger assembly, and a component that cannot be programmed against constrains the whole. Ask whether a documented interface exists, and for the full connector list.

One concrete architectural statement and no geography behind it. The security page states that data is backed up to multiple durable data stores and replicated across multiple availability zones, which is a specific resilience claim rather than an adjective and is worth crediting, and a public status page reports platform availability with subscribable notifications. Together those give a buyer more operational visibility than most products in this category offer.

What is missing is everything a location obligation turns on. No hosting provider is named, no region is stated, no tenancy or isolation model is described, and no recovery time or recovery point objective appears, so the replication claim establishes that failure is survivable without establishing where anything lives or how quickly service returns. A reference to flexible region specific control appears without explanation of what it controls or which regions are available.

The question has weight because call recordings are voice data about identifiable individuals and international dialing is sold as an upper tier feature, so calls to European recipients are an expected use rather than an edge case. Ask which provider and regions host recordings, what region specific control covers, and what the recovery objectives are.

Two certifications, a publicly named auditor, and a scope broader than this index usually sees. The vendor states on its own security page that it holds an audited control report of the continuous type and a certified information security management standard.

The detail that lifts this above an assertion comes from the company's own announcement of that audit, which names the assurance firm that performed it and states the scope as covering security, availability, processing integrity and confidentiality. Four of the five trust criteria is materially wider than the two most vendors carry, and processing integrity is a meaningful inclusion for a platform whose function is placing and connecting telephone calls correctly.

The same announcement named the executive responsible and stated that the information security certification was then being pursued, which the current security page confirms was achieved, so a reader can trace a progression rather than a claim. Technical measures are described specifically, covering encryption at rest and in transit, multi factor authentication, password complexity and replicated backup, and a public status page operates.

Held below the top band because the auditor and scope come from an announcement four years old, no current audit period, certificate number or report request route appears, and no trust portal was located. Ask for the current report with its period and auditor.

Pricing was published and now is not, and the estimates that filled the gap do not agree. Every tier routes through a demonstration, there is no self serve purchase path, and independent sources state plainly that public pricing has been withdrawn.

What third parties report does not converge: the entry tier appears at two hundred and fifty dollars per user per month in several accounts and at three hundred in one that cites the vendor's own pricing page as of April 2026, while the upper tier is estimated between four and five hundred dollars per user per month with no source claiming to have seen a figure. A buyer cannot establish an order of magnitude from that spread.

Three structural terms are consistently reported and matter as much as the rate. Contracts are annual with no monthly option, so the commitment begins at twelve months. A three seat minimum applies, putting the floor near nine thousand dollars a year before any discussion of the upper tier.

And a reduced seat type for part time callers is reported at a thirty to fifty percent discount, which is a sensible accommodation for a product priced per seat and used intermittently, published nowhere by the vendor. The project brief recorded partial pricing as published; that is no longer accurate. Ask for the rate at both tiers, the seat minimum, the term, and what the reduced seat option covers.

Nothing published on leaving, and what accumulates here is unusually sensitive. No export scope or format, post termination access window, retention period, deletion timeline or notice term was located on any surface reached. The asset at stake is the call archive: recordings and transcripts of every conversation a team has had with a prospect, accumulated over an annual contract at the volume this product exists to enable, alongside scorecards, coaching history and the connect rate analytics that justify the spend.

Recordings are also records a business may be obliged to retain or to delete on request from the person recorded, so the question is not only commercial. Nothing states whether audio leaves in a playable form, whether transcripts and derived scores accompany it, or what happens to any of it once a subscription ends.

The commercial terms compound the practical exposure, since contracts are reported as annual only with a seat minimum, so a customer deciding the product is not working waits out the term. Ask what exports at termination and in what format, whether recordings and transcripts are included, and what is deleted and when.

The telephone equivalent of this axis is number reputation, it is the category's defining problem, and this vendor addresses it from an unusual direction. When carriers and their analytics partners label a number as likely spam, connect rates collapse regardless of how good the calling is, and remediation ordinarily means monitoring number health, rotating numbers, registering with reputation services and displaying a verified business name on the recipient's handset.

Nothing published by this vendor describes any of those, and a competitor states flatly that branded calling is not offered at all. What the vendor does instead is attack the input rather than the identity: the recently acquired verification technology checks and refreshes prospect numbers, and the stated rationale is trustworthy phone data.

That is a real strategy, because dialing disconnected and wrong numbers is one of the behaviours that gets a caller flagged in the first place, and improving list quality reduces the flagging pressure upstream of any reputation management. It is also incomplete on its own, since a number can acquire a poor reputation from volume alone. Ask what number health monitoring, rotation and branded caller identity the platform provides, and what happens to connect rates when a customer's numbers are flagged.

A narrow market chosen deliberately, with the boundaries drawn by the commercial terms rather than by a statement. The product suits organisations whose outbound motion is substantially the telephone, staffed by dedicated calling teams large enough to keep several simultaneous lines usefully busy, and independent accounts place it with well funded teams of ten representatives and above in software and technology.

The terms enforce that boundary: a three seat minimum, annual contracts with no monthly option, and per seat rates in the hundreds mean a small or seasonal team is priced out rather than served badly, which is a cleaner outcome than selling them something that will not pay back. Two design decisions extend coverage thoughtfully.

International dialing appears at the upper tier for teams calling outside their home market, and a reduced seat type for part time callers at a substantial discount recognises that many organisations have people who dial occasionally rather than daily. Held below the top band on two counts. The vendor states none of this itself, so a buyer learns where they fit from third parties.

And the product is dialer only, with no messaging, electronic mail, inbound handling or contact data, which competitors raise repeatedly and which a buyer assembling a stack should weigh. Ask what team size and call volume the seat minimum assumes.