Go-to-market

Your Outbound Isn't Broken. Your ICP Is.

How to choose an ICP, build the offer around it, and prospect it properly. An example from a B2B SaaS scaleup that was spending thousands on ineffective outbound with little success.

Book a GTM audit
James Ker-Reid

A B2B SaaS scaleup I worked with was spending $9,060 a month on outbound. $4,000 of that was software. The rest went to an agency. Across seven months, it produced about 7 meetings, which is $9,060 a meeting before anyone's time is counted. Fully loaded, the vendor modelling it put the number at $13,070. The average deal was $1,760 in monthly recurring revenue.

One pilot inside that spend ran 200 cold calls. 5% connected. It produced 2 meetings.

The natural instinct in that situation is just to do more. Change agency, change sequence, change the copy. That's the wrong first move almost every time. The outbound wasn't broken. The target list was, and everything downstream of a bad list is expensive noise.

What follows is the sequence I run instead. Choose the ICP. Build the offer around it. Then prospect it, inbound and outbound, at the same time.

Choosing: start from the customers you already won

Build the list backwards

Most target lists get built the wrong way round. Someone picks an industry, sets a headcount range, exports 4,000 rows, and calls it a market.

We did the opposite. We took the existing customer base, ran each account through a data enhancement process through Claude Code, and extracted the product and service tags those companies had in common. Not their industry classification. What they actually sell and who they sell to.

That mattered because the end customer of their customer didn't have a correlation with deal value. Especially as this was an API offering. Under the old industry-code approach, the list had filled with services businesses sharing a label with the software vendors we wanted to prospect. Same code, completely different business, completely different buyer.

The tag-based pull returned about 22,000 companies. The majority had between one and ten employees. Cutting that band left 6,400 at eleven or more. Layer on geography and sub-segment and you land at 1,023 net-new accounts. Merge in disparate CRM records and a spreadsheet or two, and the working list is 1,200.

The validation step almost nobody runs

Take your finished target list. Now check your existing customers against it. Not to add them. To see whether your own list would have found them.

Roughly half weren't on it.

That's the most useful number in the whole exercise. The ICP definition and the reality had drifted, and nobody had noticed because nobody had checked or upheld the criteria.

While we were in there, the CRM said 93 customers, a second list said 126, and a third file held 46 accounts nobody could match. Three sources, three different truths, and a sales team quoting whichever one supported their point.
Part of what a target list does is settle that argument. It's a data cleanse and a piece of governance at the same time. This is our market. This is how many customers we have. This is the classification. This is the rep who owns it.

Tier on revenue, then be honest about tier three

We banded three tiers on company revenue. The middle band was the historic ICP, kept deliberately so we could see how much of the new list overlapped the old assumption.
Tier
Company revenue
Target accounts already in the CRM
Tier 1
Above $150m
39
Tier 2
$10m to $150m
157
Tier 3
$1m to $10m
246
The existing customers didn't cluster in tier one. They landed across all three.
That's the shape you'd expect, and it's also the shape that quietly eats a rep's week. Loosen your criteria and the count goes up while the fit goes down. Say that out loud when you hand the list over, rather than letting a rep discover it in month two.

Exclusions are decisions, so put a number on them

Every cut we made had an account count attached to it before we made it.

Services businesses came out entirely, which was the largest single cut and the reason the list dropped to 1,200. Central and South America went into a long tail, because revenue was already weighted elsewhere. The one-to-ten employee band was held back for self-serve, inbound, rather than proactive outbound, on the basis that it's a needle in a haystack for the sales team and perfectly good traffic for marketing.

That's the test for any exclusion. Not "is this segment hard" but "what does avoiding it cost me, in accounts, and is that a trade I'd make in the open". A cut you can't size is a preference wearing a strategy costume.
There's a second half to that, and it's the part I'd now insist on. An exclusion isn't finished until you've written down what would bring it back. Otherwise you've handed someone a smaller market and called it a decision.

Why you pick one segment and not three

The strongest argument for narrowing came out of a conversation about a second market we were considering.
Going into a new segment you already have unknowns. Company size is one. The stakeholder map is another, and not just the buyer, but the champion, the sponsor and the blocker. Messaging and call to action is a third, because you don't have any yet. Three variables.
Add a second segment and you have the same three problems again. Six. Add a third and you're at nine. Add a second geography and you're at twelve. You will not get a clean read on any of them.
Concentrate your effort and fail in one area in several ways, quickly. Market selection gets blamed for a lot of failures that were really failures of concentration.

Delivering: the offer follows the buyer, not your preference

An ICP is not a buyer profile, and the difference is operational

This came up when a founder handed me a document titled ICP that described job titles.
You can build a target list and qualification criteria from an ICP. You can't build either from a buyer profile. One comes after the other.
A product manager at one software company and a product manager at another can carry identical titles inside businesses that operate nothing alike. One sells to twenty-person startups and manages seventy accounts. The other sells to global corporates and manages three. Different motion, different marketing, different acquisition cost, different hire.
Get the company profile right first. The buyer profile sits on top of it, and it's what your messaging gets written for.

Package for the buyer you just chose

Once the ICP was set, the packaging questions answered themselves.
The team wanted to charge on volume of activity. It's the obvious value metric. It's also the one that made the buyer flinch. A product manager who can't predict what a feature will cost next quarter won't roll that feature out widely. They'll use you for the narrow case and build the rest themselves, which is what one of their largest customers had already done.
Seat-based pricing removed the hesitation. Use it in more of the product, no extra charge. That's a packaging decision that follows directly from knowing who signs and what they're afraid of.
Two other things worth taking.
Don't change pricing and ICP in the same quarter. Refocusing a whole company on one segment is already a large change. Repricing at the same time pollutes the experiment, and when the numbers move you won't know which change moved them. We deferred pricing to month four.
Price is also a lagging indicator of a stale ICP. 53% of that customer base hadn't had a pricing change in 24 months. That's what happens when the offer stops tracking who's actually buying.

The floor, and the licence to say no

We set a line. Below $25,000 a year, sales doesn't take the meeting; we pass it to support.

That sits deliberately above the $21,120 the average customer was already worth. A floor that matches your current average isn't a floor, it's a description.

Stating it is easy. Operating it is the hard part, and the founder pushed on exactly that. It sounds great in a document. How does a person do that in a room? Three answers.

  • 1

    Never lead with price

    It's binary, it's exclusive, and it creates no value before it excludes. Qualify on something the buyer will happily tell you instead. How many customers do you have. How many people will use this.

  • 2

    Put the same logic on all the lead forms

    So marketing and sales disqualify on one shared rule rather than two.

  • 3

    Disqualify before the demo, not during it

    A form review, a website, a LinkedIn profile, five minutes of desk research. Sometimes you cancel the demo.

Then give the people you turn away somewhere real to go. Ours went to support, as order-takers with a soft close. Can you do this? Yes. Here's the price. Here's how to buy it. It clears the funnel without burning the relationship, and small deals still close, just without a salesperson attached.

The wider version of this is what makes focus survive contact with the organisation. Everyone gets a licence to say no. If it doesn't grow the target segment, why are we doing it?

Watch the leadership trap while you're at it. Founders think out loud, and what sounds visionary from the top reads as permission at the bottom. Say "we should explore that" in an all-hands and someone will spend three weeks exploring it.

Prospecting: point both engines at the same list

Inbound and outbound, same week, same accounts

Most companies get this structurally wrong. Marketing tests one audience, sales works another, and neither compounds.
Advertise to the customers of a specific platform while you outbound to the customers of that same platform, and the advertising performs better. You draw a reflective wall around one market and everything you throw bounces back into the middle. The alternative is random acts of marketing.

Same list, two motions.

Inbound: one form, two questions

We had two forms. Neither captured size or volume. Neither had a route for people who shouldn't reach a salesperson.

The fix was one global form, deployed everywhere, with triage after submission rather than before. People fill in whatever form they find. Rely on them to self-sort, you get routing chaos, and then you blame the buyer for it.

Two qualifying questions, both answerable in three seconds.
First, a motion split. Are you buying this to build into your own product, or for your own team? That one question separates two completely different sales motions and two different pricing structures.

Second, a size range. How many customers do you have, or how many people will use this. Ranges, not exact figures. Don't ask a question the prospect has to go and look up the answer elsewhere in their first interaction with you. We tested a more precise metric and dropped it, because it made people stop and think and half of them wouldn't know the answer.

Cross-check the answer server-side against firmographic data in the CRM. If a ten-to-fifty person company tells you a thousand people will use it, you've learned something either way.
Routing comes off those two answers. Below the threshold goes to support. Above it and inside the target list goes to the owning rep. Above it and outside the list goes to the sales lead, case by case. Legacy landing pages come out of the main navigation and their traffic goes to self-serve.

Outbound: replace, don't add

When the vendor modelled their proposal, they modelled it as incremental. Existing agency, existing stack, plus them.
I stopped it there. I'm looking to replace, not to add.

That reframing is worth more than any tool choice. The plan became: cut the agency, cut most of the tech stack, build the capability in-house with better tooling, and accept lower output for two quarters while it compounds. Target spend went from $9,060 a month to $1,250 a month, a little over a quarter of what it was. The bar I set for myself was parity of output at that cost within two quarters.

Three things made it defensible

Judge outbound on payback, not on ROI multiples

The pitch you'll get is a deal worth $21,120 in annual contract value against $9,060 of monthly spend. Sounds great. It isn't, if you bill monthly and the deal landed in month five.

My rule is one to one and a half times the first month's recurring revenue to acquire a customer, and I'll spend it happily when retention runs past twelve months. Different question, different answer.

Two campaigns is not two motions

The team was running two segments as two separate operations. When I ran the revenue analysis, 11 of the top 20 accounts sat in the same sub-segment. 55%. Same product, same buyer, mostly the same problem.

So: two campaigns with different content, one motion, one cost base. The other segment moved to organic and search, where it belonged.

Check the signal before you act on it

If you trigger calls off email opens, most of your triggers are firewalls. An open that fires in the same minute as the send isn't a person. Require a gap of at least three minutes, and spacing between subsequent opens.

Inbox privacy protection creates the same false positive across your whole list. Get this wrong and your SDR spends the week calling people who never read the email.

For reference, the best AI-assisted outbound numbers I saw in that period were a vendor reporting open rates moving from 30% to above 50%, and reply rates from 0.1% to 3 or 4%, across six accounts. Useful as a ceiling. Not a plan.

Where this leaves you

Nine times out of ten, the company asking me to fix their outbound doesn't need a new agency. They need to know who they're selling to, at the level of a named list with a number on it.
Build the list backwards from your own customers. Check whether it would have found them. Put a number on every exclusion. Set a floor above your current average deal, and route everyone below it somewhere they can still buy. Then point inbound and outbound at that one list, in the same week, and leave them alone long enough to read the result.
You'll learn more from failing at one segment in five ways in ninety days than from three parallel experiments running all year.

Want this run on your own numbers?

A GTM audit helps you get clarity on your ICP, your target list, sizes every exclusion, and tells you where inbound and outbound are pulling in different directions and how they can work together better.

Book a GTM auditSee how a Fractional CRO could help