Why we don't resell carrier minutes — and why you shouldn't pay anyone who does

Reselling carrier minutes is a pricing model with an ownership problem attached. The margin is the visible half; the invisible half is that your numbers, your recordings, your call detail records and your relationship with the carrier that signs your calls all live inside somebody else's account. Autocloz charges a platform fee and passes carrier rates through unchanged, which is a commercial choice — but the argument for it is not mainly about price. It is about which assets survive the day you cancel.

The question is what you own when you leave

Every calling-tool comparison is a features table. That table is measuring the wrong thing, because the features converge and the ownership does not.

Run this test instead. Imagine cancelling the vendor at the end of next month, then list what you still have. Not what you can export — what you still *have*, working, in an account with your name on it. For most bundled or resold calling products the honest answer is a CSV of call metadata and nothing else: the numbers are registered to the vendor, the recordings are in the vendor's object storage behind URLs that stop resolving, the call detail records are rows in the vendor's database, and the rate you were paying was never a rate you negotiated.

That is not an accusation of bad faith. It is the natural consequence of the architecture. If the vendor buys the carrier relationship and sells you a slice of it, everything downstream of that relationship is theirs by construction.

The reason to care is that the assets are not equally replaceable. Software is replaceable in a week. A phone number that has been on your website for four years, has accumulated a reputation with the analytics engines that label calls, and appears in every signature your team has ever sent, is not.

Six assets, and who holds each one under each model

Take the two architectures and walk the assets one at a time. Under resale, the vendor holds the carrier account and bills you for minutes. Under pass-through, you hold the carrier account and the vendor charges for software.

The telephone numbers. Under resale they are provisioned in the vendor's carrier account, which means porting them out is a request to the vendor before it is a port request to a carrier. Under pass-through they sit in your own carrier account from the first day.

The attestation relationship. Under resale, the originating provider is whichever carrier the vendor bought from, and your identity is one layer removed from the entity making the signing decision. Under pass-through, the carrier that onboarded you is the carrier that signs. This is the asset people understand least and it is covered in its own section below.

The call recordings. Under resale they are stored by the vendor, usually behind time-limited URLs that expire when the account does. Under pass-through they are objects in your carrier's storage, or fetched and stored by you.

The call detail records. Under resale, rows in the vendor's system, exportable as a file. Under pass-through, the carrier keeps its own authoritative CDR and your software keeps a copy, so there are two independent records of what happened.

The rate card. Under resale you pay a published price with an undisclosed margin. Under pass-through you pay the rate you negotiated, and volume discounts land in your account rather than improving somebody else's gross margin.

The compliance obligation. This one does not move. It is yours under both models, and the section on the honest costs of pass-through explains why resale can make it feel otherwise.

Notice that five of the six move and the sixth only appears to. That asymmetry is the argument.

The number is an asset, and porting it is a legal right

People treat phone numbers as configuration. In the United States they are closer to property with a statutory transfer procedure, and knowing the procedure changes how you negotiate.

The FCC's porting rules are specific. Under 47 CFR 52.35(a), carriers required to port numbers "must complete a simple wireline-to-wireline or simple intermodal port request within one business day unless a longer period is requested by the new provider or by the customer", and the rule goes further into the mechanics: a complete Local Service Request received between 8 a.m. and 1 p.m. local time is eligible for activation at midnight the same day, and anything after that is treated as received at 8 a.m. the next business day. Non-simple ports get four business days under the same section. Separately, 47 CFR 52.36 limits the data a carrier may require to accomplish a simple port to a standardised set of fields, which exists precisely to stop a losing carrier from inventing paperwork.

So a one-business-day clock and a fixed field list are the law, not a favour. What the law cannot fix is a number that is not registered to you. A port request comes from the gaining carrier to the losing carrier on behalf of the account holder, and if the account holder is your software vendor, your exit depends on their cooperation rather than on the rule.

Two practical consequences:

• Ask where your numbers are provisioned before you buy, not when you leave. The answer is a single sentence and it predicts your entire exit cost.
• A number's reputation is not portable in the way the number is. Porting moves the digits. The behavioural history that the terminating carriers' analytics engines have built around those digits travels with the number, which is good if the history is clean and is an argument against burning numbers if it is not.

The signing relationship cannot be sublet

This is the ownership point that almost nobody makes, and the rule text is unusually clear about it.

Under the FCC's caller ID authentication framework, the originating voice service provider signs each call and attaches an attestation level reflecting how confident it is about the caller's identity and their right to use the number. A provider may outsource the mechanical authentication work — but 47 CFR 64.6301(b) attaches conditions that keep the judgement with the provider. The provider must require the third party to "sign all calls using the certificate obtained by the voice service provider", it must itself make "all attestation-level decisions regarding the caller identification information of each SIP call it originates", it must memorialise the arrangement in writing specifying the third party's tasks, and it must retain that agreement "for a period of two (2) years from the end or termination of the agreement".

Read that as an architectural statement rather than a compliance chore. The attestation decision belongs to the originating provider by regulation and cannot be delegated away. No layer of software above the carrier changes how your calls are signed, and no vendor can promise you an attestation level it does not control.

The consequence for the resale-versus-pass-through choice is direct. Full attestation depends on the originating provider being able to confirm both who the caller is and that the caller legitimately holds the number being presented. That confirmation is easiest when the carrier onboarded you, holds your know-your-customer documentation and provisioned the number in your account. When you are a customer of a customer, the entity making the decision has a weaker basis for it. That is a structural pressure toward lower attestation under resale, not a certainty — but it is the mechanism, and it is the one worth asking about. What actually happens between dial and ring, including how attestation is presented covers the signing chain end to end.

Recordings and call detail records outlive the subscription

Two record types, two different failure modes.

Recordings are the one people notice late. A recording accessed through a vendor's expiring URL is not a file you hold; it is a permission you rent. The moment to discover this is not the moment you receive a dispute, a regulatory request or a discovery obligation covering a call from eighteen months ago.

Under pass-through the recording is written into your own carrier's storage, and the software's job is to hold a reference and refresh it — Autocloz maintains the recording reference and re-resolves it against the provider rather than caching a URL that dies quietly. That is a smaller promise than "we keep your recordings forever", and it is a more honest one, because the durable copy is in an account you control.

Call detail records are the other half. Under pass-through there are two independent records of every call: the carrier's, which is the billing-authoritative one, and your workspace's, which carries the CRM context — which lead, which campaign, which disposition, which outcome. Being able to reconcile the two is how you catch the failure where calls were placed and the CRM never learned about them, which is a real and specific class of bug rather than a hypothetical.

One honest note on that pairing: two records means two things that can disagree, and reconciling them is work you now own. Under resale there is one record and no reconciliation, which is simpler and also means there is nothing to check it against.

The arithmetic, with every rate stated as an assumption

Any specific per-minute figure in a post like this would be invented, so here is the structure with placeholders you fill in from two real quotes.

Let R be your carrier's per-minute rate, M the resold per-minute rate, N the number of billable minutes per month, P the platform fee, and S the per-seat fee times seats.

• Resale total = (M × N) + S
• Pass-through total = (R × N) + P

The crossover is at N = (P − S) ÷ (M − R). Below that many minutes, the model with the lower fixed component wins. Above it, the per-minute difference dominates and keeps growing, because one term scales with usage and the other does not.

That is the whole analysis, and it produces three observations that hold regardless of the numbers you put in:

A margin on minutes is a tax on activity. It grows precisely when your calling programme is working. A flat platform fee does the opposite.

The seat fee compounds with the margin. Resold products commonly charge both, so adding a rep raises the fixed component and the variable one at the same time. That is also why the per-seat model quietly discourages the coverage — a manager, an operations person, someone who only needs to listen to a call — that makes a calling team better.

You cannot compute the crossover without both quotes. Ask the resale vendor for the per-minute rate to the destinations you actually call, and get a quote from a carrier for the same destinations. If the vendor will not state a per-minute rate for your destinations, that refusal is itself the answer.

Do the arithmetic before deciding, not after — the outbound stack cost calculator will do the fixed-versus-variable split for you if you would rather not build the spreadsheet.

What bring-your-own actually costs you

The pass-through pitch is usually sold as free. It is not, and pretending otherwise is how people end up resenting the choice halfway through onboarding.

Carrier onboarding is a real process. Opening an account means passing that carrier's know-your-customer checks, funding a balance, and provisioning numbers. It is much lighter than it was a decade ago — for Telnyx it is an API key and a connection identifier pasted into voice settings — but it is a day of work and it involves a vendor relationship you did not previously have.

You hold the compliance obligation, visibly. In the United States that means the TCPA and the rules under it, registration and the 31-day refresh discipline around the National Do Not Call Registry administered by the FTC, and the calling-hours window computed in the called party's local time. In India, commercial messaging and voice sit under TRAI's Telecom Commercial Communications Customer Preference Regulations of 2018, which require registered entities, headers and templates on a distributed ledger. None of this is created by choosing pass-through — it applies either way — but resale can create the impression that the vendor's compliance posture is yours, and it is not. What TCPA compliance actually requires before you build a dialling motion is the right place to start on the US half.

Two bills instead of one. Somebody has to reconcile carrier spend against platform spend, and notice when a balance runs low. In practice this is a monthly half-hour and an alert, but it is a job that did not exist before.

Support has a seam. When audio is bad, the question of whether it is the carrier or the software has to be answered before it can be fixed. A single vendor owns that question end to end; two vendors means you own the triage.

Those are genuine costs. They are also mostly one-time or fixed, which is why they favour pass-through at volume and favour a bundle for a team making twenty calls a day.

Which carrier you bring decides which dialler modes you can run

This is the part of pass-through that gets glossed over in every article about it, including, historically, ours. The carrier is not an interchangeable commodity behind a uniform interface. What the carrier exposes decides what the dialler can do.

Four capabilities decide it: whether the provider offers server-side call origination over a REST API, whether it offers provider-native answering-machine detection, whether it can bridge call legs server-side, and whether it permits multiple simultaneous legs per agent. A parallel dialler needs all four. Preview and power dialling need only the ability to place a call, which a browser softphone or a SIP endpoint satisfies.

Autocloz keeps that as an explicit capability matrix rather than discovering it at runtime, and the differences between the supported providers are large:

• Telnyx carries the full set — dial, call control, answering-machine detection, bridging, simultaneous legs, server-side dispatch, and a programmable answer webhook — which is why it is the only provider on which an outbound IVR campaign or a parallel dialler runs today.
• FreJun has no public REST outbound-dial endpoint. Calls originate from the rep's browser through its WebRTC softphone, so it is marked rep-console-only and a server-side power campaign on it is refused rather than silently degraded.
• DIDLogic is SIP and browser-SDK based, likewise rep-console-only, so the supervisor cannot originate calls on it at all.

When a mode is unavailable on the chosen provider, the API rejects it with a 409 and a stated reason rather than quietly falling back to something that looks similar — a deliberate design rule in this codebase, because a power campaign that silently runs as preview is a campaign the operator believes is running and is not.

The practical instruction is short: choose the carrier for the dialler mode you need, then negotiate the rate. Picking on price and discovering afterwards that parallel dial is unavailable is the most expensive ordering of those two decisions. The differences between preview, power and predictive dialling is the right place to work out which mode you actually need, and the Telnyx integration page lists what connecting that account gives you.

Autocloz's free plan covers 5 users and 10 mailboxes across all five channels, and connecting your own carrier does not add a seat charge — start free and put your own rate card in before you scale the dialling.

When paying for resale is the right call

The honest cases, because a post arguing one side and admitting nothing is an advertisement.

Low volume. Below the crossover, the margin on minutes is smaller than the operational overhead of a second vendor relationship. A team placing a few dozen calls a day should buy the simplest thing.

No operations capacity. If nobody owns vendor accounts, top-ups and number provisioning, pass-through will decay: a balance will run out on a Friday, and the discovery will be a silent campaign on Monday.

A genuinely bundled product you need the rest of. If the calling is one module of a suite whose other modules you use daily, splitting the calling out to save margin may cost more in integration than it saves.

Regulated environments with an existing approved vendor. Where procurement has already cleared a supplier and clearing another is a quarter of work, the margin is cheaper than the process.

What is not a good reason is the claim that pass-through is technically difficult. It was, when carrier APIs were raw and every integration meant SIP trunk configuration by hand. Today the connection is credentials in a settings page, and a vendor still using complexity as the argument is describing a product from 2015.

What bring-your-own-carrier does not fix, and what Autocloz does not do

The limits, stated plainly, because the ownership argument above is strong enough not to need help.

Pass-through does not lower your carrier rate on its own. It gives you the relationship in which a rate can be negotiated; if you never negotiate, you pay list price to a carrier instead of a marked-up price to a vendor, and the saving is smaller than the model implies.

It does not improve call quality. Audio path, jitter and codec behaviour are properties of the carrier and the route, and choosing your own carrier means you own that outcome rather than escaping it.

It does not change your compliance exposure. Consent, registry scrubbing, calling hours and registration are obligations of whoever places the call, under both models.

And it does not stop your numbers being labelled. Terminating carriers' analytics engines weigh behaviour — high volumes of short unanswered calls look like a robocaller regardless of who signs them — so a number burnt by a bad pattern is burnt whether you rented it or own it.

Autocloz specifically. It supports Telnyx, DIDLogic and FreJun for voice, with an Exotel capability slot whose adapter is not built, so the choice of carrier is real but not unlimited. Parallel dialling and outbound IVR run on Telnyx only, and on FreJun and DIDLogic the dialler runs through the rep's browser rather than server-side — a limitation of those carriers' interfaces, not something a future release can wish away. It does not negotiate carrier rates on your behalf, does not resell numbers, and does not hold your recordings as the durable copy. Emergency calling obligations attach to your carrier relationship and are yours to satisfy. And the compliance gate enforces the rules you configure in the jurisdictions you tell it about; knowing which rules apply to a given recipient stays with you. The voice platform's number inventory and caller-ID strategies and the dialler modes themselves are where those choices are made concrete.