State of SMB Outbound 2026

Methodology

The data in this report comes from a single live outbound system operated by Gray Reserve over a thirty-day window ending 27 April 2026. Emails were sent to first-party-enriched prospect records sourced from Apollo (people search) and an internal warchest of Google Places businesses across the Greater Houston metropolitan area. All sends went through the Resend transactional infrastructure across nine rotating sender domains. Every email used the same four-step NEPQ (Neuro-Emotional Persuasion Questioning) sequence template with per-prospect personalization derived from Apollo enrichment fields, Google PageSpeed scores, and Google Places ratings.

Aggregate counts were extracted directly from the production database. No prospect-level personally identifying information is published. Reply categorization was performed by an Anthropic Claude Sonnet classifier reading the body text of each inbound reply and assigning one of four categories: positive, negative, neutral, or out-of-office.

Finding 1 — The breakup email is the most-opened email by far

Industry conventional wisdom treats the fourth and final email in a cold sequence as a procedural breakup — a polite acknowledgement that the prospect did not engage, framed as a courtesy. The data from this window suggests that framing is wrong.

The step-4 cohort is smaller because not every prospect reaches it — only those who survived three prior touches without disqualification or reply. But on a per-send basis, the breakup is opened more than six times as often as the cold introduction. Two operational implications follow. First, the breakup email is not a courtesy; it is a closing mechanism worthy of as much creative attention as the opening. Second, shortening sequences to two or three steps in pursuit of "respect for the prospect's inbox" forfeits the highest-engagement touch in the arc.

Finding 2 — Sender domain matters more than copy in this dataset

Across nine sender domains in active rotation, reply rates varied by a factor of three for prospects receiving identical sequence templates on identical schedules. Copy was held constant; the only meaningful variable was the from-address.

The .us domain underperformed by a factor of three relative to the .agency domain. Forensic review of bounce rates on the underperformer showed a five-day climb from 14% to 23% before the domain was paused — the sender reputation was visibly degrading inside Resend's deliverability dashboard. Domain reputation is not a copy problem dressed up as a deliverability problem; it is upstream of every other variable an outbound operator can tune. Operators investing in a single sender domain are not running an A/B test against themselves.

Finding 3 — Reply taxonomy is dominated by out-of-office

Of the 52 replies received in the window, 30 (58%) were classified as out-of-office. Thirteen (25%) were classified as neutral ("interesting, send more later"; "not the right person, try X"; etc.), six (12%) were negative ("remove me"; "not interested"), and three (6%) were positive — direct expressions of interest.

The out-of-office majority is, counterintuitively, the most encouraging finding in the report. An out-of-office reply requires the sender's message to have reached the recipient's primary inbox, triggered the mail client's automatic reply rule, and returned via the same routing path. Bulk-flagged messages do not produce out-of-office replies; they sit in spam folders that the mail client does not auto-respond from. The 58% out-of-office share is a positive-deliverability signal masquerading as a non-event.

The 25% neutral category warrants closer attention. These are prospects who responded but did not commit. Many include a referral to a colleague, a request for a later quarter's budget, or a request for additional context. In our internal classification, neutral replies convert to booked meetings at a non-trivial rate when followed up inside the same business day with a specific, contextual question rather than a generic re-pitch.

Finding 4 — The four-step sequence carries 16,247 prospects through full progression

A common objection to multi-step cold sequences is that prospects drop off so steeply between touches that the latter steps are mathematically irrelevant. The funnel attrition in this data window tells a different story.

The progression is roughly even — once a prospect makes it to step 2, they typically reach step 4. The drop from 5,117 (step 2) to 3,530 (step 4) reflects natural progression timing rather than mass attrition. This means the per-prospect cost of running a four-step sequence is not significantly higher than running a two-step one. The marginal email costs fractions of a cent. The marginal AI generation cost is roughly four cents. The marginal opportunity cost of not sending steps 3 and 4 is forfeiting the most-engaged email of the arc (see Finding 1).

Operational implications for SMB outbound operators

1. Treat the breakup as a closing tool, not a courtesy. Allocate creative attention to the step-4 email proportional to its engagement weight. The 19.3% open rate is not an artifact; it is the highest-leverage touch in the sequence.
2. Operate a sender-domain rotation, not a single domain. The 3× variance in reply rate across domains in this dataset represents a strictly free uplift available to any operator running multiple .com / .pro / .agency / .site domains in parallel. Single domain operations are absorbing reputation risk that a rotation distributes.
3. Read out-of-office as a deliverability win, not a non-reply. A 50%+ OOO share among replies is consistent with messages reaching primary inboxes. If your OOO share is below 30%, your domain reputation is likely the binding constraint, not your copy.
4. Resist the urge to shorten sequences. The argument that shorter sequences "respect the inbox" assumes negative engagement on later touches. The data does not support that assumption — late touches are more engaged-with, not less.
5. Treat the neutral reply as the highest-leverage follow-up opportunity. Same-day, contextual, non-pitch responses to neutral replies convert at materially higher rates than treating neutral as a soft no.

Limitations and caveats

• Single-operator dataset. Conclusions reflect Gray Reserve's deliverability infrastructure, NEPQ sequence template, and personalization model. They do not generalize to all SMB outbound programs.
• Greater Houston geographic concentration. The local-market composition of the prospect pool may differ from national or international outbound programs.
• 30-day observation window. Sender-domain reputation effects compound over longer windows; a quarterly analysis would shift the relative domain rankings.
• Open-rate measurement is subject to the mail-client tracking-pixel blocking environment. Apple Mail Privacy Protection inflates reported opens by triggering pixels at delivery. Click rate and reply rate are not subject to this distortion.
• Reply categorization is automated (Claude Sonnet classifier) and not human-audited at scale in this report. Sample audits suggest classification accuracy in the 90-95% range; the four reported categories should be read as approximations, not exact counts.

Data availability and citation

Aggregate counts in this report are released under a Creative Commons Attribution 4.0 International license. The underlying prospect-level data is not released. To cite this report:

Gray, J. (2026). State of SMB Outbound 2026 — Gray Reserve Field Report. Gray Reserve Research, 28 April 2026. Retrieved from https://grayreserve.com/research/state-of-smb-outbound-2026

The structured Schema.org Dataset for this report is embedded in the page head and indexed by Google Dataset Search.