The safest way to roll out an SMS feature in a SaaS product is in three stages: your own team first, a small group of beta tenants second, then everyone, with a feature flag per tenant, a plan for the carrier registration queue, a trained support team and spending limits at every step. Texting differs from most features because a mistake leaves your servers. A bad send reaches a real phone, can cost real money and can't be recalled.

This guide is for the product and engineering leads planning that rollout. It assumes you send through Ready, where each customer gets a sub-account with its own numbers and carrier registration, but the staging logic applies to any provider.

Why texting needs a slower rollout

Three things make this launch different from a normal feature release.

  • Registration takes time. US business texting runs through 10DLC registration, and each tenant needs its own. On Ready, most approvals take 2 to 7 days. That wait sets the pace of your rollout, so you can't flip it on for everyone on a Monday.
  • Sends are irreversible. A text can't be unsent. Compliance rules such as quiet hours and opt-outs apply from the first message.
  • Costs scale with a bug. A loop that retries can burn money in minutes, because each segment is billed.

Our launch checklist for SMS in a SaaS lists the compliance items. This post is the order of work around them.

The three stages at a glance

StageWho has itGoalExit when
0. InternalYour team, on test tenantsProve the plumbingSends, webhooks and opt-outs behave
1. BetaA handful of chosen tenantsLearn from real use casesRegistrations approve, support can handle questions
2. GeneralAll tenants, opt-inScale upCost and failure rates are stable

Don't skip stage 0 because it feels like testing. It's where you find that your webhook handler drops duplicates wrongly or that a test number isn't registered.

Stage 0: internal

Create a few sub-accounts that belong to your own company and use them like customers would. Register one for real, so your team sees the registration flow end to end, including a rejection if you can provoke one. Walk through these checks:

  1. Send a text with the Idempotency-Key header, retry the request and confirm only one text goes out.
  2. Reply STOP from a test phone and confirm that number can't be messaged again.
  3. Send during the recipient's quiet hours and confirm your UI handles the answer. An API text outside the window gets a 409 QUIET_HOURS_BLOCK unless the window opens within 6 hours.
  4. Trigger a failure and confirm the status shows in your UI. Our guide to showing delivery status honestly covers the states.

Every text is charged from the first one, and sending needs a card and a number on the account. A number is $2 a month, so plan test numbers deliberately. Accounts opened before November 12, 2026 keep their signup prices, shown under Billing.

Stage 1: beta tenants

Pick beta tenants on purpose. Good ones have a clear, simple use case (appointment reminders or order updates), an owner who answers email, and consent practices you've checked. Avoid tenants whose use case is vague or sensitive.

Flags. Gate the feature per tenant, not per user. A tenant-level flag lets you turn texting off for one customer without a deploy. Keep a second flag for sending itself, separate from the flag that shows the settings page, so a tenant can fill in registration details before any send is possible.

The registration queue. Treat registration status as a visible state in your product: not started, in review, approved, needs changes. Ready exposes status and submission through its registration endpoints, described in the registration API docs. Don't let beta tenants send at scale before approval. An account can send up to 500 texts a month before it registers, at its normal per-text rate, and after that it has to register. Our 10DLC API and developer flow post walks through building that screen.

Feedback. Ask each beta tenant for three things: what they sent, what confused them, and what they'd have wanted before sending. Keep the sample small enough that you read every message thread.

Support readiness

Texting creates support questions you haven't had before. Before stage 2, make sure your support team can answer:

  • Why a message shows as not delivered, or has no delivery report at all
  • Why a contact can't be texted (opted out, on a Ban list, a landline)
  • Why a registration was rejected and how to resubmit
  • Why a text was held or refused in quiet hours

Write short macros for each and link your internal runbook. The post on triaging delivery failures for support teams is a ready-made starting point. Also tell support what they must not do: Ready isn't HIPAA compliant and doesn't sign BAAs, so support should tell tenants to keep protected health details out of texts.

Cost guardrails

Set limits before anyone can send, not after the first surprise. Ready bills per segment: $0.0075 plus a $0.0045 carrier fee, with the number and registration billed monthly. Billing is prepaid credits. Useful guardrails:

  • A per-tenant daily send cap in your own code, starting low during beta.
  • A maximum batch size on bulk sends. A bulk call to Ready takes at most 100 recipients, so split bigger lists.
  • A segment estimate shown before a campaign launches. An emoji drops a segment to 70 characters, so a 140-character text with one emoji is 3 segments, not 1.
  • An alert when a tenant's volume jumps well above its own average.

Next, decide how you bill your customers. Our post on metering SMS usage per customer covers it. For worked math, label your assumptions: for example, assume 1,000 beta messages of 2 segments each, which is 2,000 segments, then multiply by the Standard rate plus the carrier fee from the pricing page.

Stage 2: general release

Open it gradually even now. Release to 10 percent of tenants, then half, then all, and watch these numbers at each step (assumed thresholds, set your own):

  1. Registration approval share and time in the queue
  2. Failure rate and opt-out rate per tenant
  3. Support tickets about texting per 100 tenants
  4. Spend per tenant against your caps

Keep the kill switch. A flag that disables sending for one tenant, or for everyone, is the cheapest insurance you'll buy.

Frequently asked questions

How long should a beta last?

Long enough for your beta tenants to finish registration and send a real campaign (most approvals take 2 to 7 days). Registration approvals alone can take most of a week, so a one-week beta often ends before anyone has sent. Set the exit criteria first, then the date.

Should I turn texting on for everyone at once?

No. Each tenant needs its own registration, and the review wait means most tenants can't send on day one anyway. A staged release also protects your support team and your budget.

What's the first guardrail to build?

A per-tenant send cap you control in your own code. It's quick to build and covers bugs, runaway loops and one tenant's surprise volume.

Where to go next

Write down the three stages, the exit criteria and the kill switch before you write any UI. Then create an account, set up a test sub-account and run the stage 0 checks. The API quickstart gets you to a first send and a webhook, and the sub-account docs show how each of your customers is isolated.