The agency bulk problem
Agencies running newsletter or announcement campaigns for five, ten, or twenty clients hit the same wall every month: how do you push a million-recipient send in a tight window, per client, without manually chunking and without reputation contamination across clients? Postmark is the wrong tool for that shape.
What a bulk email API is
A batch-submission endpoint that accepts N recipients in one HTTP call, fans the send out across configured providers in parallel, enforces per-recipient suppression, and emits progress events. Idempotency keys let you retry without double sending.
How it works
- Client workspace with configured providers (SendGrid, Amazon SES, Mailgun, Postmark, etc.).
- Agency submits a batch via
POST /batcheswith recipients, template, and idempotency key. - Policy selects providers based on volume headroom, region, and brand.
- Batch runs in parallel chunks; events stream to webhooks.
- Progress endpoint gives real-time state; resumable on failure.
Benefits
- Batch semantics: one API call, thousands-millions of sends.
- Idempotency keys make retries safe.
- Parallel fan-out across providers gives throughput a single provider cannot.
- Per-client workspace isolation (reputation, suppression, audit).
- Canonical progress events regardless of provider.
Drawbacks
- Not a cadence sequencer; agencies still use external tooling for drip campaigns.
- Provider accounts remain the client’s; agency operates the batches.
- Template storage and rendering are expected on the agency’s side (or via a template system of choice).
Who should use which
- Postmark: transactional traffic for client apps (receipts, notifications). Stays best-in-class.
- Bulk email API: monthly/weekly broadcasts per client, announcements, newsletters, reactivation campaigns.
Comparison table
| Dimension | Postmark | Bulk email API |
|---|---|---|
| Optimised for | Transactional | Broadcast |
| Batch submission | Limited | Native (thousands/call) |
| Idempotency keys | Per-request | Per-batch |
| Provider fan-out | Postmark only | Multi-provider |
| Client isolation | Account per client | Workspace per client |
| Suppression | Account | Per workspace |
| Progress webhooks | Per message | Per batch + message |
| Compliance | Postmark DPA | Mailers.io DPA + GDPR |
Pricing at /pricing; agency surface at /platform/agency-email-platform; bulk email API at /api/bulk-email-api.
Testing discipline for guide-style problems usually improves when you separate “content experiments” from “infrastructure changes.” If you must change both, sequence them: stabilize the path, then test creative, or you will not know which variable moved the signal you care about. If you are comparing providers, do it with the same list ethics and the same segment definitions; otherwise the comparison is a story, not a measurement.
Runbooks are underrated. A good runbook is not a PDF nobody opens; it is a checklist that includes who is allowed to do what, what “pause sending” does in your configuration, and how to verify suppression state after an incident. Mailers.io is built as orchestration and policy on infrastructure you connect—useful when you have multiple paths, shared templates, and need consistent governance across teams. It is the wrong product if the primary pain is a missing CRM surface or a guarantee that mail will “land in primary.”
Related depth for “Article”: operators often underestimate how much time is spent on credential lifecycle (API keys, SMTP passwords, domain delegation) and how little time is left for improving message quality. Rebalance that intentionally if revenue depends on reliable outbound. Multi-provider routing can reduce provider-specific lock-in and separate blast radius, but it does not remove your obligation to own consent, suppression, and record-keeping. Not legal advice. Where GDPR, CCPA/CPRA, or similar apply, align with counsel. We do not use generic marketing copy to assert SOC 2 or ISO 27001.
Cross-functional alignment fails quietly: Marketing ships a new domain, Data updates a list export, and Engineering rotates an API key—each change reasonable alone, but together they break assumptions about identity and suppression. A useful discipline is a lightweight change log for anything that touches a live sending identity, even if the change is “small.” The goal is not paperwork theatre; the goal is that the next on-call can reconstruct state without heroics.
Procurement and security questions often ask for certifications as shorthand. The better question is: what logs exist, for how long, and who can access them? A control plane can unify routing, but you still need your own data map for personal data, subprocessors, and incident response. This article is educational; align final commitments with your counsel and your customer contracts. We do not claim outcomes we cannot own (placement, read rates, or a unified sales inbox) because that would mis-sell the product’s shape.
Related depth for “Article”: operators often underestimate how much time is spent on credential lifecycle (API keys, SMTP passwords, domain delegation) and how little time is left for improving message quality. Rebalance that intentionally if revenue depends on reliable outbound. Multi-provider routing can reduce provider-specific lock-in and separate blast radius, but it does not remove your obligation to own consent, suppression, and record-keeping. Not legal advice. Where GDPR, CCPA/CPRA, or similar apply, align with counsel. We do not use generic marketing copy to assert SOC 2 or ISO 27001.
Finally, treat deliverability talk as a constraint problem, not a battle of slogans. Recipients, mailbox providers, and local IT policies are not under your vendor’s control. What you can control is list provenance, authentication, throttles, content hygiene, and how fast you stop repeating mistakes. The organizations that do well here look boring: fewer surprises, fewer “unknown unknowns” in audits, and operators who can show receipts.
When you operationalize Article at scale, the durable win is a repeatable review loop: weekly metrics that surface drift before leadership notices. That usually means bounces and complaints as first-class series—not vanity engagement charts—paired with a written rule for when a program pauses. This matters whether your stack is a single console or a multi-provider layer; the work is the same even when “Article” is the public label on the project.
Related depth for “Article”: operators often underestimate how much time is spent on credential lifecycle (API keys, SMTP passwords, domain delegation) and how little time is left for improving message quality. Rebalance that intentionally if revenue depends on reliable outbound. Multi-provider routing can reduce provider-specific lock-in and separate blast radius, but it does not remove your obligation to own consent, suppression, and record-keeping. Not legal advice. Where GDPR, CCPA/CPRA, or similar apply, align with counsel. We do not use generic marketing copy to assert SOC 2 or ISO 27001.
Testing discipline for guide-style problems usually improves when you separate “content experiments” from “infrastructure changes.” If you must change both, sequence them: stabilize the path, then test creative, or you will not know which variable moved the signal you care about. If you are comparing providers, do it with the same list ethics and the same segment definitions; otherwise the comparison is a story, not a measurement.
Runbooks are underrated. A good runbook is not a PDF nobody opens; it is a checklist that includes who is allowed to do what, what “pause sending” does in your configuration, and how to verify suppression state after an incident. Mailers.io is built as orchestration and policy on infrastructure you connect—useful when you have multiple paths, shared templates, and need consistent governance across teams. It is the wrong product if the primary pain is a missing CRM surface or a guarantee that mail will “land in primary.”
Related depth for “Article”: operators often underestimate how much time is spent on credential lifecycle (API keys, SMTP passwords, domain delegation) and how little time is left for improving message quality. Rebalance that intentionally if revenue depends on reliable outbound. Multi-provider routing can reduce provider-specific lock-in and separate blast radius, but it does not remove your obligation to own consent, suppression, and record-keeping. Not legal advice. Where GDPR, CCPA/CPRA, or similar apply, align with counsel. We do not use generic marketing copy to assert SOC 2 or ISO 27001.
Cross-functional alignment fails quietly: Marketing ships a new domain, Data updates a list export, and Engineering rotates an API key—each change reasonable alone, but together they break assumptions about identity and suppression. A useful discipline is a lightweight change log for anything that touches a live sending identity, even if the change is “small.” The goal is not paperwork theatre; the goal is that the next on-call can reconstruct state without heroics.
Procurement and security questions often ask for certifications as shorthand. The better question is: what logs exist, for how long, and who can access them? A control plane can unify routing, but you still need your own data map for personal data, subprocessors, and incident response. This article is educational; align final commitments with your counsel and your customer contracts. We do not claim outcomes we cannot own (placement, read rates, or a unified sales inbox) because that would mis-sell the product’s shape.
Related depth for “Article”: operators often underestimate how much time is spent on credential lifecycle (API keys, SMTP passwords, domain delegation) and how little time is left for improving message quality. Rebalance that intentionally if revenue depends on reliable outbound. Multi-provider routing can reduce provider-specific lock-in and separate blast radius, but it does not remove your obligation to own consent, suppression, and record-keeping. Not legal advice. Where GDPR, CCPA/CPRA, or similar apply, align with counsel. We do not use generic marketing copy to assert SOC 2 or ISO 27001.
Finally, treat deliverability talk as a constraint problem, not a battle of slogans. Recipients, mailbox providers, and local IT policies are not under your vendor’s control. What you can control is list provenance, authentication, throttles, content hygiene, and how fast you stop repeating mistakes. The organizations that do well here look boring: fewer surprises, fewer “unknown unknowns” in audits, and operators who can show receipts.
When you operationalize Article at scale, the durable win is a repeatable review loop: weekly metrics that surface drift before leadership notices. That usually means bounces and complaints as first-class series—not vanity engagement charts—paired with a written rule for when a program pauses. This matters whether your stack is a single console or a multi-provider layer; the work is the same even when “Article” is the public label on the project.
Related depth for “Article”: operators often underestimate how much time is spent on credential lifecycle (API keys, SMTP passwords, domain delegation) and how little time is left for improving message quality. Rebalance that intentionally if revenue depends on reliable outbound. Multi-provider routing can reduce provider-specific lock-in and separate blast radius, but it does not remove your obligation to own consent, suppression, and record-keeping. Not legal advice. Where GDPR, CCPA/CPRA, or similar apply, align with counsel. We do not use generic marketing copy to assert SOC 2 or ISO 27001.
Testing discipline for guide-style problems usually improves when you separate “content experiments” from “infrastructure changes.” If you must change both, sequence them: stabilize the path, then test creative, or you will not know which variable moved the signal you care about. If you are comparing providers, do it with the same list ethics and the same segment definitions; otherwise the comparison is a story, not a measurement.
Runbooks are underrated. A good runbook is not a PDF nobody opens; it is a checklist that includes who is allowed to do what, what “pause sending” does in your configuration, and how to verify suppression state after an incident. Mailers.io is built as orchestration and policy on infrastructure you connect—useful when you have multiple paths, shared templates, and need consistent governance across teams. It is the wrong product if the primary pain is a missing CRM surface or a guarantee that mail will “land in primary.”
Related depth for “Article”: operators often underestimate how much time is spent on credential lifecycle (API keys, SMTP passwords, domain delegation) and how little time is left for improving message quality. Rebalance that intentionally if revenue depends on reliable outbound. Multi-provider routing can reduce provider-specific lock-in and separate blast radius, but it does not remove your obligation to own consent, suppression, and record-keeping. Not legal advice. Where GDPR, CCPA/CPRA, or similar apply, align with counsel. We do not use generic marketing copy to assert SOC 2 or ISO 27001.
Cross-functional alignment fails quietly: Marketing ships a new domain, Data updates a list export, and Engineering rotates an API key—each change reasonable alone, but together they break assumptions about identity and suppression. A useful discipline is a lightweight change log for anything that touches a live sending identity, even if the change is “small.” The goal is not paperwork theatre; the goal is that the next on-call can reconstruct state without heroics.
Procurement and security questions often ask for certifications as shorthand. The better question is: what logs exist, for how long, and who can access them? A control plane can unify routing, but you still need your own data map for personal data, subprocessors, and incident response. This article is educational; align final commitments with your counsel and your customer contracts. We do not claim outcomes we cannot own (placement, read rates, or a unified sales inbox) because that would mis-sell the product’s shape.