How we validate an outbound system before spending a client’s budget on it
Hubex case study
Oriol SerraFounder & CEO
“We don’t need outbound, we get a lot of referrals.”
Hubex said it, and they were right for several years. A strong referral engine is a genuine advantage. It is also a single point of failure, and it fails quietly: referrals do not stop, they just arrive less often, and by the time the trend is obvious you are a quarter behind.
That is where Hubex was when we started. Referrals still coming, coming slower, and no second channel that had ever been proven to work.
The company
Hubex helps B2B technology companies run sales, marketing and customer service on HubSpot.
They are good at building systems for other people. They had never built one for their own pipeline, because they had never needed to.
The situation
Two constraints shaped the engagement.
No outbound history. No previous campaigns, no reply data, no evidence about which angle or which segment would respond. Nothing to optimize against.
No internal time. Building the list manually was not an option for a team already delivering client work.
Building one large campaign and launching it is the expensive way to find out you were wrong about the segment.
So we did not do that.
The method
When there is no historical data, the first campaign is not a campaign. It is an experiment, and it should be sized like one.
We launched multiple narrow tests in parallel, each with its own targeting logic and its own angle, on small volumes. Enough sends to produce a readable signal. Not enough to burn a segment if the signal came back negative.
Underneath the tests sat the infrastructure that makes any of it valid: domains bought, DNS configured, inboxes warmed. None of that is interesting and all of it determines whether a negative result means the angle failed or the email never arrived.
The results
Three tests. Three different outcomes.
One at 6.7% reply on 222 sends. One at 5.3% on 107. One at 0.95%, which is a failure and was treated as one.
Those are small volumes producing high reply rates, which is what a correctly targeted campaign looks like at this stage. A campaign sending ten times the volume for the same number of replies is not ten times better. It is a segment being burned in exchange for the same outcome.
The failed test is the useful one. It cost roughly 200 sends to learn that an angle did not work, instead of several thousand. That angle was killed and the budget moved to variants of what had worked.
The most useful part for me has been the Clay tables, finding leads, and how that all works.
Christian Fredrik Stene, Co-founder, Hubex
Why we build this way
A campaign that works is not the deliverable. A method for finding out what works is.
Launching one large campaign gives you a single data point, and if it underperforms you cannot tell whether the problem was the list, the angle, the offer, the timing or the infrastructure. Everything failed together, so nothing is learnable.
Running narrow tests in parallel gives you a comparison. When one angle returns 6.7% and another returns under 1% on the same infrastructure and the same week, the variable that differed is the answer.
Hubex ended with a system that keeps producing that answer. Their team tests variants of what worked and deletes what does not, which is the only maintenance an outbound system genuinely needs once the architecture underneath it is sound.
Who this is for
Companies with one channel that is working and no proven second one.
Especially if that channel is referrals, because referral businesses tend to discover the dependency late. The referrals do not stop. They just thin out, and the quarter where you notice is not the quarter you can fix it in.
The time to build the second channel is while the first one is still producing.
Want this built for you?
30 minutes. We map what a validation phase would look like for your market and what it would cost you to find out.
Request a Strategy Call