I work across a lot of B2B SaaS companies running outbound, and I see the same thing almost everywhere. The outbound isn’t landing, so the founder reaches for the most visible lever (the messaging) and starts rewriting emails. New subject lines, a different hook, a punchier opener.
It rarely fixes anything, because copy is the wrong end of the problem.
There are three things you have to get right for outbound to work, and they go in a specific order: who you target, which of them to talk to right now, and what you say. Most teams work that list backwards. They fiddle with the third thing on a list they never questioned, sent to accounts they never prioritised.
Here is the order, why getting it wrong is more expensive than it looks, and what you should actually do at each step.
One: who you’re actually targeting
Not the version you’d like to be true. The version your closed deals already prove.
Most founders, when I ask who their ICP is, give me a list of industries, a revenue band, and a couple of countries. That isn’t an ICP. It’s a mailing list with a profile attached. It tells a rep who to email but not a single reason why anyone would reply.
The work is to go the other way round. Look at who actually bought (and who didn’t), find the patterns in the deals that closed, and break the broad industry into sub-segments that have genuinely different problems. “Fintech” isn’t one market, it’s payments and lending and wealth and insurtech and regtech, and they don’t lie awake worrying about the same thing. The problems a payments company has with fraud and settlement look nothing like the problems a lending business has with credit risk and funding lines. If you’re selling to “fintech,” you’re really selling to five or six different buyers who each need a different conversation.
Then, for each persona, name the problem. Not the profile, the problem. What breaks for them, what it costs them, and what happens if they do nothing about it. If you can’t answer those three, that persona isn’t really in your ICP however neatly they fit the firmographics. They’re just a row in a spreadsheet.
What you should do.
- Pull your last ten to twenty closed deals and look hard for the patterns nobody wrote down: company size, sub-segment, who signed, what they had tried before, and what finally made them move. Do the exact same for the deals you lost.
- Mine the conversations you already have before you go and speak to anyone new. Your call recordings hold the exact language buyers used to describe their problem, the objections that came up, and the moment something clicked. Most founders are sitting on a goldmine of buyer insight they have never listened back to. Go through them.
- Speak to five recent customers and ask each of them the same five questions: what problem were you solving, what were you using before, what made you switch, what nearly stopped you, and who else was in the decision.
Between the recordings and the five conversations, you will have more honest signal about your real ICP than any amount of guessing in a strategy doc.
Two: which of them to talk to right now
Knowing who could buy is not the same as knowing who’s ready. This step has two phases, and they happen in order: first you prioritise the accounts worth going after, then you watch for the signal that one of them is ready this month.
Phase one: prioritise the accounts worth going after. Your ICP tells you who could buy. It doesn’t tell you which of those accounts is worth your time first. Not all fit accounts are equal. Some are worth more to you (a bigger contract, more seats, a faster close, more room to expand later), and those are where your limited hours should go first. This is where you build on the ICP work rather than treat every account in it as the same.
I worked with a company that was treating every account in their ICP as equal. Their pricing scaled with seats, so we segmented their target accounts by likely seat volume and had them go after the larger ones first. Same effort, same number of conversations, but the average deal size went up, because the accounts they were reaching out to were simply worth more the moment they closed. The prioritisation didn’t come from the buyer’s attributes alone. It came from understanding their own economics and ranking accounts against that.
Phase two: watch for the signal that says now. Once you know which accounts matter most, the question becomes timing. At any given moment only a small slice of your market (call it five percent) is actively looking. The rest could be a perfect fit and still not pick up the phone, because nothing has changed for them this month. So the job isn’t to email everyone who fits. It’s to find the accounts where a fit you trust meets a fresh signal that something has shifted: a new hire in a relevant seat, a funding round, a product launch, a competitor they’ve just outgrown.
Here’s the step nearly everyone skips. You have to cross-reference the signal against the fit. A buying signal at an account that’s a poor fit is a distraction, not a lead, and chasing it feels like progress while getting you nowhere. Signal plus fit is the lowest-hanging fruit in all of outbound, and most teams never join the two up. They either spray the whole list (ignoring timing) or chase every signal that moves (ignoring fit).
And one more thing on signals, because it’s where most teams trip up now. Signal data has never been easier to get or to see. There are tools that will show you hiring, funding, job changes, technology shifts and website visits, all of it, in real time. But a signal you can see and do not act on is worse than no signal at all, because it gives you the feeling of being data-driven while nothing actually changes. A signal without a play is trivia. A signal with a play is pipeline.
What you should do.
- Rank your fit accounts by what they are worth to you, not just by fit. Decide the one or two attributes that make an account worth more (seats, contract size, expansion potential) and sort by those first. Go after the most valuable fits before the rest.
- Pick your top three signals. The ones that genuinely suggest something has changed for a fit account (a relevant senior hire, a funding round, a competitor switch).
- Write the play for each one. Not “we track senior leadership hires” but “when a fintech we are a fit for hires a new Head of Risk, here is exactly what we say, here is the angle, here is who reaches out and on which channel.”
- Do not switch a signal on until it has a play attached, and never action a signal without checking the account is a fit first.
Three: what you say
Now, and only now, the copy.
I’ve put it last on purpose. Not because it doesn’t matter, but because it’s the cheap part once the first two are right, and a waste of effort while they’re wrong. Better copy on a bad list just means more persuasive emails to people who were never going to buy.
When you do get here, the whole game is relevancy. Everything in the message should earn its place by being relevant to this specific buyer and what’s going on for them right now. The “I saw your LinkedIn post” style of personalisation is finished as a differentiator, because every rep now has the same AI doing the same trick, so it cancels out into noise. Worse, when personalisation is forced (the school they went to, the city they live in, the team they support) it actively reads as a tactic, and the buyer clocks it immediately.
The shift that matters is from personalisation to situation. Swap every forced piece of personalisation for something situational: what just happened to them, and the specific problem it creates, in their own language, with a real number attached. A new regulation landing on a lending business is relevant. A funding round that’s about to put a payments team under pressure to scale is relevant. Where they went to university is not. The test for a good message is simple: would they have been glad to receive it even if they never buy from you? If yes, you’ve written something worth sending. If it’s a thinly dressed “book a demo,” you haven’t.
What you should do.
- Take your best-performing email and run it through one test: could any rep at any company have sent this? If the answer is yes, it is too generic. Rewrite it until the answer is no.
- Go through it line by line and strip out anything that is not relevant to this exact buyer in this exact situation, including any personalisation that feels forced.
- Replace what you stripped out with the real, current situation you found in layer two.
- Read the final version back and ask: would they be glad they got this, even if they never reply? If not, it is not relevant enough yet.
Why the order is the whole point
This is the bit that’s counterintuitive, so it’s worth being blunt about it.
If your targeting is wrong, better copy makes things worse, not better. You scale up persuasive messages to the wrong people, burn your domain reputation, and grind down your reps in the process. If your prioritisation is wrong, you spray perfectly good messages across a market where most aren’t ready, and then conclude outbound doesn’t work, when what doesn’t work is the timing. Copy is the layer everyone reaches for first because it’s the most visible and the easiest to change. It’s also the one with the least leverage when the foundations are off.
And here’s the part that should worry a founder watching the burn. Get the foundations wrong and you pay for the same problem twice. First when you execute the wrong thing at scale, then again when you try to unpick execution that was misaligned from the start. I’ve watched a founder spend the best part of a year, and a six-figure sum, scaling reps and tools on top of a motion he’d never actually proved. The reps weren’t the problem. The order was.
AI hasn’t made this less important. It’s made it more
You’d think the obvious move is to throw AI at all of this and let it sort itself out. It’s the opposite.
When everyone has the same AI doing the same research and writing the same “personalised” emails, the personalisation cancels out and becomes background noise. The thing AI can’t commoditise is the judgement underneath it: a genuinely evidenced ICP, a problem named in the buyer’s own words, the instinct to know which signal actually predicts a deal at your accounts rather than everyone else’s.
AI is brilliant as an extra pair of hands. Building the list, pulling the signals, drafting the first version of a message. It’s weak as a substitute brain. Deciding who matters, working out the real pain, knowing when to move. The teams winning right now put AI on the eighty percent that’s mechanical and keep humans on the twenty percent that’s judgement. The teams losing are automating a process they never proved in the first place, which is just a way of doing the wrong thing faster.
So if your outbound isn’t working, resist the urge to rewrite the emails. Go back to the start. Get clear on who you’re targeting and why they’d care, work out which of them is worth talking to this month, and then (only then) worry about what you say. Get those three right, in that order, and the copy almost writes itself.