Cold Email Infrastructure for Agencies

Most lead generation agencies start the same way. A founder takes on three or four clients, manually sets up Google Workspace mailboxes for each, and bills the client at $10-15 per mailbox per month. Margin looks healthy at four clients. Then the agency lands its tenth client, then its twentieth, then its fortieth, and the math quietly inverts.

At 30-50 client setups per month, the bottleneck stops being lead generation and starts being mailbox provisioning. A single cohort of 10 mailboxes at the standard sequencer-bundled vendor costs $89-99/month - so a 40-client agency running an average of 10 mailboxes per client is paying $3,560-$3,960/month in raw infrastructure cost. If the client retainer is $1,500/month and you carry $400/month of mailbox cost per client, you are running on $1,100 of gross margin before headcount, sequencer fees, lead data, or sales tooling. That is not a business; that is a hobby with admin overhead.

The fix is not heroic. The fix is volume pricing on the underlying mailbox - because the underlying mailbox is a Google Workspace seat or a Microsoft 365 seat that costs Google or Microsoft about six dollars wholesale. Everything above six dollars is markup for automation, dashboard, and warmup. Vendors that charge agencies $7-10/mailbox at scale are charging four to six dollars in pure markup. The unit economics of running a 300-mailbox agency on those vendors mean the agency works for the vendor, not the other way around.

StackMail's volume pricing collapses to $2.50/mailbox at 100 mailboxes on annual billing and $2.37/mailbox at 600 mailboxes on annual billing. At 50-mailbox scale, the per-mailbox price is $2.58. That is a 60-75% reduction in infrastructure cost versus sequencer-bundled vendors, and a 5-25% reduction versus the closest specialist competitors. The savings flow directly into agency margin or, if the agency wants to pass them along, into a price advantage at the client RFP table.

The other dimension agencies forget is operational cost. If a vendor charges per-tenant rather than per-mailbox, every new client triggers a fixed setup fee. If a vendor does not support bulk CSV onboarding, every new client costs four to six hours of account-manager time at $75-100/hour - another $400-$600 of soft cost per client onboarded. Volume pricing on the mailbox is necessary but not sufficient; the agency also needs bulk operations, per-workspace isolation, and one-click sequencer export to make the per-client onboarding cost stay below $50.

The most expensive mistake an agency can make is co-mingling clients inside a single mailbox tenant. The reasoning is simple but worth spelling out: deliverability reputation lives at the level of the IP, the sender domain, and the upstream Google Workspace or Microsoft 365 tenant. If two clients share a tenant, they share a reputation. If client A sends a campaign that gets a wave of spam complaints, client B's mailboxes inherit the reputation hit even though client B did nothing wrong. Open rates drop, bounces spike, and the agency spends two weeks remediating an issue that should have been contained to one client.

The blast radius problem compounds at 30+ clients. A single regulated-vertical client (debt relief, lending, supplements, nutraceuticals) sending a borderline-compliant campaign can take down a tenant shared with 15 other clients. Industry data from the major email-monitoring vendors shows that mixed-vertical tenants experience 2-3x more reputation incidents than vertical-isolated tenants. Agencies that do not isolate by client (or at minimum by vertical) end up running a permanent firefighting operation.

StackMail's workspace primitive maps one-to-one with one client. Every workspace gets its own tenant binding, its own sender IP cohort, its own warmup peer graph, and its own audit log. A complaint storm on client A's workspace stays inside client A's workspace. The agency's other 49 clients see no reputation drift. This is the single most important architectural feature for an agency operating at 100+ mailbox scale, and it is the feature most under-emphasized by sequencer-bundled vendors because their architecture cannot offer it cleanly.

Three layers of isolation exist in the StackMail multi-tenant model. The outermost is the agency account - a single billing entity, a single payment method, a single consolidated invoice. Inside the agency account live workspaces, one per client. Inside each workspace live domains, mailboxes, sender configurations, and the audit log. The architecture is intentionally hierarchical so that an agency operator can navigate from agency-wide cost reporting down to a single mailbox bounce log in three clicks.

At the infrastructure layer, every workspace binds to a Google Workspace tenant or a Microsoft 365 tenant - and crucially, those tenants are not shared. When client A's workspace is provisioned, we create a new Workspace organization for client A under the StackMail reseller agreement with Google. Client B gets a different Workspace organization. The separation is enforced upstream by Google itself; we are not relying on dashboard-level filtering to keep them apart. This is the same model InboxKit and the better Mailforge tier use; it is not the model that sequencer-bundled vendors like Instantly or Smartlead-Forge use, which is why those vendors cannot offer true per-client isolation at the infrastructure level even when their dashboards present a workspace abstraction.

At the warmup layer, peer selection is scoped to the workspace. When mailbox X in client A's workspace warms up, its peer pool is drawn from client A's other mailboxes plus a curated cross-agency pool of healthy senders we maintain. Mailbox X never warms up against client B's mailboxes, which means a poisoned warmup signal in one workspace cannot leak into another. This sounds like a small thing; at 600 mailbox scale it is the difference between a healthy fleet and a slow-bleed reputation issue across the entire agency book.

At the access layer, role-based permissions scope agency users to specific workspaces. An account manager assigned to clients A, B, and C cannot see client D's mailboxes, audit log, or cost line. The agency owner sees everything. Optional client-side read-only access (Pro tier and above) gives the client a view of their own workspace and nothing else. The permission model is enforced server-side, not just hidden in the UI.

Sequencer choice at the agency level is rarely uniform. Client A wants Smartlead because their previous agency used it. Client B wants Apollo because their sales ops team standardized on it. Client C wants Instantly because their founder watched a YouTube tutorial. Client D wants Lemlist because they need the personalization features. An agency that picks a single sequencer at the infrastructure layer is forcing a stack onto the client, which costs deals at the proposal stage.

StackMail is sequencer-agnostic by design. The export pipe at the workspace level lets you push mailbox credentials into any of 20+ supported destinations: Instantly, Smartlead, Saleshandy, Apollo, Lemlist, Woodpecker, Reply, Mailshake, Outreach, Quickmail, and others. The export is one-click bulk for the most-used destinations and OAuth-based for the rest. Different workspaces can use different sequencers - workspace A can run Smartlead, workspace B can run Apollo - without infrastructure conflicts.

The full integration surface is documented on our integrations page. If a sequencer you use is not on the list, ping support and we will add a connector inside two weeks - this is a routine request from new agency customers.

White-label cold email infrastructure means different things to different agencies. The honest version of what we offer today, and what is on the roadmap, is below.

Available today (Pro and Custom tiers): rebranded client-facing reporting view with the agency's logo and primary color; custom CNAME for the client portal subdomain (clients.youragency.com pointing at our infrastructure); StackMail branding removed from CSV exports and PDF deliverability reports; agency-name attribution on the warmup audit log; agency-controlled email-template branding for client-facing notifications. The dashboard backbone, the support channel, and the billing entity remain StackMail; clients sign in to a co-branded experience but the underlying platform identity is visible at the footer level.

On the roadmap (H2 2026): a fully white-label reseller program where the entire dashboard runs under your domain (app.youragency.com), the support emails come from your team with our infrastructure underneath, the billing entity flips to the agency for client-facing invoices, and StackMail becomes invisible to the end client. This is a higher-touch tier with a minimum commitment of 500 mailboxes and a separate contract structure - if you want to be on the early-access list, mention it during the call.

Not on the roadmap: reselling the StackMail brand under a different name as a turnkey product. The reseller program is white-labeling our infrastructure under your brand; it is not packaging our SaaS for resale to other infrastructure agencies. We make this distinction explicit because a few prospective customers have asked and we want the answer in writing.