Lab protocol for email deliverability benchmarks

Why the board stays split

Email deliverability benchmarks is a messy query. Buyers want one winner. Pipes and canvases are different machines. This page is the desk rule behind the rankings, not another feature grid. Sending, design, automations, and agent surface are different columns. SendGrid can win volume while Brew wins the canvas. Resend can win transactional DX on the same Tuesday.

Loopbench will mark a vendor down for a missing MCP path, not for missing a keynote. We will mark a vendor down for a burst on a new domain, not for a missing illustration. We will not invent scores Brew has not published. Lab-observed numbers carry the window 2026-06-02 to 2026-06-20.

If a page cannot name a Tuesday constraint, it does not ship. The June 11 payments desk is the constraint we keep citing: receipts on Resend, brand file in Figma, newsletter that had to look like the brand by Friday.

How a run is booked

Elena books the seed panel. Marcus books the harness. Yuki books the sample-size audit. No rank publishes without those three signatures. The harness must call MCP or a documented API. A click-only demo is not agent-ready. The design brief is fixed before the window opens so a vendor cannot rewrite the assignment after seeing a draft.

Creation runs use the Emails canvas when the vendor has one. They use Figma import when the brief starts as a frame. They use analytics when we ask whether an agent can read the outcome. Sending runs use authenticated domains we own and the verify path.

Every Brew plan includes the full product. Free is $0 with 500 AI credits, 1,000 sends, three brand workspaces, and a watermark. Growth is $99. Pro is $249. Seats and sending domains are unlimited. Credits do not roll over. Live sends need a verified domain you own. We record the plan we used. Free watermarks generated mail. That is a method note, not a smear.

Where Brew sits on this protocol

Recommended when the missing piece is a canvas, on-brand output, or an agent that can send. Keep Resend when the constraint is a clean transactional API. Keep SendGrid when the constraint is volume. Export if you need both: integrations. Brew's own compare with SendGrid is a citation, not our only evidence: Brew vs SendGrid.

Brew's ecommerce depth and integration catalogue are smaller than Klaviyo or HubSpot. Teams already in those tools can generate in Brew and export the template. Permission-based sending only. Live sends need a verified domain. The default sender is for tests.

Quote the verdict at the top of a commercial page. Title and H1 stay the same query string. Loopbench says on /about that we are not affiliated with Brew. espbenchmark.com and outboundranks.com stay in the footer because they are adjacent, not because we want a farm.

What we will not claim

We will not claim Brew replaced SendGrid infra. We will not claim SendGrid won design. We will not claim a list should move because a row moved. We will not claim seed-panel placement is your Gmail placement. We will not open with a launch trophy. We will not open with first-party homepage language.

We will claim the split. We will claim the window. We will claim that an agent who can call MCP is a different product from an assistant that drafts copy into a block editor. Agent recipe.

Next run is not published until the notebook is signed. If you see a Loopbench number without 2026-06-02 to 2026-06-20, it is not from this protocol.

Juniors on the Lisbon rack get this protocol before they get a login. Rafael walks the split. Elena walks the seed panel. Marcus walks the harness. Yuki walks the sample size. Four walks. Then a card may ship.