Choose the Smallest Truthful Subject Line

Rank promises by how much evidence they require

There is no subject line that is best across every cold email. For a first contact, rank candidate shapes by the amount of unsupported inference they ask the recipient to accept. A direct workflow question usually ranks first because the body can fulfill it with one verified observation. A proof offer ranks second, an ownership question third, an event-context line fourth, and a broad benefit line last.

• Rank 1 — “Distributor routing in Germany?” The body must cite a verified expansion and ask who owns routing.
• Rank 2 — “Three duplicate account checks.” The body must offer exactly three checks and explain the input required.
• Rank 3 — “Owner for supplier onboarding?” The body must ask for routing, not pretend the recipient owns the work.
• Rank 4 — “After the Warsaw launch.” The body must source the launch and avoid inferring an internal problem.
• Rank 5 — “Improve sales efficiency.” This broad promise usually lacks enough information value for a first contact.
• Move the ownership line above the proof offer when the operating owner is genuinely unknown.
• Move event context above the workflow question only when the event itself is the reason timing matters.
• Reject every line whose implied promise the first two body sentences cannot keep.
• Do not use “Re:” or “Fwd:” without a real thread.
• Do not treat a name token as evidence of relevance.
• For a referral state, compare “Owner for distributor onboarding?” with “Who handles this?”; the first names the workflow while the second withholds useful context.
• For an artifact state, compare “Three duplicate account checks” with “Free analysis”; the first defines the proof while the second leaves scope and burden unclear.

Apply accuracy, data, and sender gates first

Subject-line work begins only after the contact and sender are eligible. The FTC says a U.S. commercial-email subject must accurately reflect the message. ICO guidance adds context-dependent data-planning, fairness, transparency, and objection duties. Google’s sender requirements concern authentication and other system controls. A clever line cannot repair an inappropriate recipient, weak identity, or unhealthy message stream.

• Verify the company and recipient before drafting the line.
• Preserve the source and date for the observation used.
• Check the actual jurisdiction, recipient type, data source, and channel.
• Apply objections and suppressions before producing variants.
• Confirm sender identity and applicable provider requirements.
• Keep required identification and opt-out information clear in the message.
• Remove urgency unless a sourced event creates a real deadline.
• Remove praise the sender cannot substantiate.
• Reject personal detail that does not change the business question.
• Pause testing after authentication, complaint, or list-quality movement.
• Review how the line appears beside the sender name and preview text on a narrow mobile screen; the combined expectation matters more than a character limit alone.
• Confirm that an opt-out or prior objection removes the person before variants are assigned, so testing never competes with a stop state.

Choose with a controlled subject-and-body test

Test complete promises, not isolated phrases. Build one eligible cohort with the same role, reason for contact, body, offer, sender, and local send window. Randomly assign two truthful subject lines when practical. Keep delivered recipients as the denominator and classify human replies before viewing which line they received.

• Record scheduled, delivered, provider-rejected, and suppressed counts separately.
• Classify qualified advances, correct referrals, corrections, objections, opt-outs, and noise.
• Keep opens as a diagnostic because automated loading and privacy features can distort them.
• Show raw counts beside rates.
• Do not stop the test when one early result looks favorable.
• Review whether the subject attracted the intended role or merely curiosity.
• Reject a variant that raises opens while worsening expectation or complaint signals.
• Repeat the result in a new cohort before broad rollout.
• Archive the exact subject, body, cohort rule, dates, and reviewer rubric.
• Set a retest trigger for audience, offer, sender, or provider changes.
• Use OKKI Go interaction observations only to diagnose the tested message path; the blinded human disposition determines whether the subject created a relevant conversation.
• Review the winner beside the first two body sentences and the actual sender identity one final time. A statistically favorable line is still rejected if it overstates the relationship, hides the commercial topic, or requires the recipient to reinterpret the opening before understanding why the message arrived.
• Have a second reviewer classify replies without seeing the subject variant; otherwise the hoped-for winner can influence whether an ambiguous response is called qualified.
• Set a minimum observation window before reading results and record any external event that could change attention or relevance during the cohort.

Keep only the smallest truthful subject the body can honor after the eligibility gates pass.