Outbound has a bad reputation and mostly deserves it. A list bought from a vendor, a template with a merge field in the first line, ten thousand sends in a week, and a reply rate somewhere near zero that gets blamed on the market.
Then the same team tries again next quarter with a different template, which is the part that gives the game away. Changing the copy is the one variable that was probably fine. B2B lead generation behaves like a pipeline: the output is decided by the inputs and the plumbing, and if you can't see which stage is leaking, you're guessing at which stage to fix.
The list is the product
Who you send to is the single biggest lever, and it isn't close. A tightly defined ideal customer profile with verified contact data beats volume every time, because every downstream number is calculated on a base of people who could plausibly buy.
Define the profile in terms a database can filter on. Not "ambitious mid-market companies", but headcount between 50 and 400, in these four sectors, using this category of software, with an operations or finance lead in post. If a criterion can't be checked, it isn't a criterion, it's a hope.
Then verify. Bounce rate is the number to watch, and it belongs under 2%: above 3% you are teaching every filter in the chain that mail from your domain is untrustworthy, and that reputation follows you into the next campaign and the one after. Verification costs a fraction of a penny per record. A burnt sending domain costs weeks.
Eight hundred relevant, deliverable, correctly researched contacts will outperform eight thousand scraped ones. That is not a moral position about outbound. It is the arithmetic of what happens when a mailbox provider decides what you are.
Sending infrastructure is not admin
Most teams treat the technical setup as a box to tick before the interesting work starts. It is the stage where the largest single failure happens, because when it goes wrong the emails simply never arrive and every metric downstream reads as a copy problem.
What that means in practice:
- Send from a separate domain. A lookalike of your main one, so a deliverability problem never touches invoices, contracts and support mail.
- SPF, DKIM and DMARC, aligned and actually passing. Gmail and Yahoo made authentication a requirement for bulk senders in 2024. Their sender guidelines set the thresholds at 5,000 messages a day to personal accounts, which most B2B outbound sits well under, and the same signals get read on every message regardless of volume. Treat the requirements as the minimum rather than the target.
- One-click unsubscribe, honoured within two days. Cheaper than the complaint.
- Spam complaint rate under 0.1%, where 0.3% is the published ceiling. Complaints are the fastest way to lose a domain.
- Warm up new domains and mailboxes over a few weeks before real volume, and cap daily sends per mailbox rather than pushing everything through one address.
None of this improves your message. All of it decides whether the message is read.
Instrument the whole funnel
Track each stage separately, because a single conversion number tells you nothing about where the loss happened:
- Delivered, not sent. The gap between those two is the infrastructure story.
- Replied, split into positive, neutral and negative. A 6% reply rate made of "not interested" is a targeting failure wearing a good costume.
- Meetings booked.
- Meetings that turned into a real opportunity, which is the only number the business actually cares about.
Two notes on measurement. Open rates have been unreliable since Apple's Mail Privacy Protection started preloading images, which registers as an open whether or not a human looked. Use opens as a rough directional signal at best, and never optimise a subject line on them alone.
And segment everything. An aggregate reply rate of 3% can easily be one sector at 9% and three at under 1%. The aggregate tells you to rewrite the email. The segments tell you to stop mailing three of the four groups, which is a much better instruction.
Feeding this back into the CRM matters more than the dashboard does. If the sales team's notes on why deals died never reach the person choosing the next list, the loop stays open and every campaign starts from the same standing position. Clean inputs and a closed feedback loop are the same habits that keep data pipelines from breaking at 3am, pointed at a different system.
Change one variable at a time
The standard mistake is rewriting the subject line, the opener, the call to action and the target list at once, seeing the numbers improve, and having no idea why.
Run it as a sequence. Hold the list steady and test the message. Hold the message steady and test the segment. Change the sequence timing on its own. Slower than it feels like it should be, and it is the only version that produces knowledge rather than anecdotes.
Watch the sample size while you do it. Four hundred sends producing eight replies is a 2% rate, and a variant producing twelve replies looks 50% better while being well inside what random variation will hand you. At those volumes you need a large gap or a lot more sends before you believe anything. Most outbound "learnings" are noise that got promoted to strategy because somebody made a slide.
Keep a written log of what changed, when, and what happened. Six months in, that document is worth more than any tool in the stack.
Where to start
Take your last campaign and work out three numbers: what percentage of sends were delivered, what percentage of replies were positive, and what the reply rate looks like split by segment rather than in total.
Whichever of those three is worst tells you what to fix, and it is rarely the copy. If delivery is the problem, fix the sending setup before you write another word. If positive replies are the problem, the list needs tightening. If one segment carries everything, you have found your next quarter's targeting for free.
That diagnostic ordering is how we run outbound and lead generation: find the leaking stage, fix that one, measure again. It is slower than a big send and considerably more repeatable.
