Best Practices
Real patterns from real campaigns. What actually works, what wastes time, and the mistakes to avoid.
Cold email works when you do a handful of things right — and fails when you get the basics wrong, even if the rest is perfect. Here's what consistently works across the campaigns running on this platform.
Picking your angle
The method you pick shapes the whole campaign — what the AI writes, how prospects react, what the booked call looks like. Get this wrong and nothing else matters.
When to use which method
- Podcast — almost always the best default. Highest reply rates across most audiences because you're offering the prospect something (exposure, airtime, recognition) rather than asking for something.
- Interview — similar to Podcast but framed as written or video interview. Works for audiences who'd rather not commit to a podcast recording.
- AI Business Audit — you offer a free AI-powered review of their business. Good for service businesses that want to show value upfront.
- Direct — pitching your service or product directly. Only use this when you have strong proof. At least one real case study with specific numbers. Without proof, Direct sounds like every other sales email.
See Email methods for the full breakdown including sample copy.
Writing better inputs
The AI produces what it's given. Bad inputs → bad emails. But don't confuse a useful niche input with an over-specific search brief.
Target industry
| Quality | What you typed | What happens |
|---|---|---|
| Bad | "Tech" | A generic email that could be about any technology company. |
| Good | "Software" or "B2B SaaS" | A campaign aimed at a clear market category. From there, use industries, keywords, headcount, locations, and excluded keywords to narrow the lead search. |
Think of Target Industry as the starting point for finding companies, not a full ICP definition. AI Lead Search can search against company industries, database keywords, headcount, locations, and exclusions. It does not reliably know every company's exact product line, who their ICP is, or whether they sell one very specific niche product to one very specific buyer.
Do not put the whole perfect customer profile into the niche box, such as "B2B SaaS companies with 50-200 employees selling to enterprise IT." Start with the market, then refine from the preview. If you need companies that sell one exact product to one exact ICP, use the preview to sanity-check and prune, or upload a CSV from a source that already has that research.
Core offer (Direct method)
| Quality | What you typed | What the AI writes |
|---|---|---|
| Bad | "We help with marketing" | "We help companies with their marketing efforts..." — vague and forgettable. |
| Good | "We help B2B SaaS companies double demo requests using AI-powered outbound in 90 days" | "We help companies like {{companyName}} double their demo requests in 90 days using AI-powered outbound..." — specific and compelling. |
Case studies (Direct method)
| Quality | What you typed | What the AI writes |
|---|---|---|
| Bad | Company: "A client" / Result: "More leads" / Timeframe: "Recently" | "We recently helped a client get more leads." — no credibility. |
| Good | Company: "Acme Corp" / Result: "+147% qualified demos" / Timeframe: "In 6 months" | "We helped Acme Corp achieve +147% qualified demos in just 6 months." — specific and believable. |
Subject lines
The AI generates subject lines for you. You can edit them — often should.
| Quality | Example | Why |
|---|---|---|
| Bad | "Partnership Opportunity for {{companyName}}" | Too formal. Sounds like spam. Everyone writes this. |
| Bad | "URGENT: Don't miss this, {{firstName}}!!" | All caps and exclamation marks trigger spam filters. |
| Good | "Quick question about {{companyName}}" | Short, casual, curiosity-driven. |
| Good | "{{firstName}}, thought of you for this" | Personal, low-pressure, reads like a real person wrote it. |
Setting up the campaign
Start with enough leads to learn
A 50-100 lead test is not enough to learn from. For a quick campaign test, start with at least 1,000-1,500 leads. If you already know the niche you want to target, 2,000-2,500 leads is the better starting point.
Do not usually go above 5,000 leads in one campaign. Once you have enough data, it is better to pivot, create new campaigns, and learn from each niche than to keep piling more leads into a campaign that is not working.
Use multiple inboxes
Use enough inboxes to get meaningful sending volume. Campaign sending is split evenly across your available inboxes, including when you have multiple campaigns running at the same time. We recommend at least 18 inboxes for enough volume to test properly and book a decent number of calls.
More inboxes means more emails sent, which usually means more positive replies and more booked calls. There are diminishing returns if the niche is tiny, so scale inboxes with the size of the market you are targeting. See the FAQ on distributing volume across inboxes for the rule of thumb.
Review every email before launching
The AI writes a draft, not a final. Every email in the sequence should be read by you before the first send. Pay special attention to:
- The opening line — does it feel natural?
- The call to action — is it clear what you want them to do?
- Any claims or numbers — are they accurate?
Read it like you are the ICP. Ask: what would I say yes to, what would resonate with me, and what would I actually find interesting?
Set realistic expectations
| Method | Typical positive reply rate | A really good rate |
|---|---|---|
| Podcast | 1-2% | 2%+ |
| Interview | 0.5-1.5% | 1.5%+ |
| AI Business Audit | 0.5-1% | 1%+ |
| Direct | 0.3-0.8% | 0.8%+ |
One Direct campaign below 1% is not automatically broken. If your positive reply rate is below 1% across multiple campaigns or campaign types, review deliverability, links in the copy, spammy wording, and sender reputation. If you run lots of Direct campaigns with little or no interest, your inboxes may also be taking reputation damage from spam reports.
After launching
Check the AI SDR inbox daily
When prospects reply, the AI SDR handles many responses automatically. But you should review "Needs Attention" threads regularly to catch edge cases the AI wasn't sure about.
Look at reply rates by method
If Podcast campaigns get 2.1% and Direct campaigns get 0.4%, guess which method you should run more of? Don't emotionally cling to a method that isn't working.
Iterate on your scripts
Use the Cold Email Writer's Variations tool to generate fresh versions of your best-performing emails. Run the variations alongside the original — don't replace a winner just because you have a new candidate.
Don't fiddle with what's working
If an email is getting good replies, keep running it. Add variations to test alongside it, not instead of it.
The five mistakes that kill campaigns
Related
- Email methods — deep dive on each angle
- Personalization — variables and how to use them well
- Deliverability — the technical foundations
- AI SDR — how replies get handled