Strategy · 2026-01-09 · 9 min
Defining an ICP a Dev Agency Can Actually Sell Into
Most dev agency ICPs are written to sound broad and end up selling to nobody. Here is how to define one narrow enough to message and wide enough to keep a pipeline full.
Ask most dev agency founders who their ideal client is and the answer is some version of "growing companies that need technical help." That is not an ICP. It is a description of most of the internet, and it produces an outbound list that could reach for months without ever narrowing down to a person who actually needs to hire an agency this quarter.
An ICP that works for outbound has to do one specific job: it has to tell a researcher which named companies to put on a list and which to leave off. If it cannot do that, it is a mission statement, not a targeting tool.
The test for a usable ICP
A workable ICP answers three questions specifically enough that two different people would build the same list from it:
- What situation is this company in right now that makes hiring an agency likely?
- What have they tried or built already that tells us where the gap is?
- Who inside the company actually owns the decision to bring in outside help?
If your ICP definition cannot answer these with names of real companies you could point to, it needs more work before it goes anywhere near a Sales Navigator search.
Situation beats industry
Dev agencies tend to define ICP by industry: fintech, healthtech, e-commerce. Industry is a weak filter on its own, because it says nothing about whether a company currently needs an agency. A better filter is situation:
- A company that just raised a round and needs to ship faster than its current headcount allows.
- A company whose product team is shipping features but has no dedicated mobile or platform engineering, so that work gets outsourced by default.
- A company migrating off a legacy stack, visible through job postings mentioning the old and new technology together.
- A non-technical founder building a first product who needs a build partner, not a hire.
Each of these describes a moment, not a sector. A fintech company and a logistics company in the same moment are both good prospects. Two fintech companies in different moments are not equally good prospects, even though an industry-based ICP would treat them the same.
A worked example
Take an agency that builds custom internal tools for operations-heavy businesses. An industry-based ICP might say "we work with logistics and manufacturing companies." That is broad enough to include thousands of companies and specific enough to tell a researcher almost nothing.
A situation-based version of the same ICP: "Companies with 50 to 300 staff running critical processes on spreadsheets or a patchwork of no-code tools, where a recent operations hire or COO has been in the role less than a year." That version tells a researcher exactly what to look for: recent leadership changes in operations roles, job postings mentioning manual processes, or public complaints about tooling gaps. It also tells the copywriter exactly what to write about, because the message can name the situation directly.
Buyer, not just title
Dev agencies often default to targeting a CTO or Head of Engineering. That is right in some companies and wrong in others. In a company without an in-house technical leader, the buyer for an agency is often the founder, a COO, or a Head of Product who has been handed the problem by default. Treating title as fixed across every company on the list means missing the accounts where the real buyer sits somewhere else entirely.
Build two or three plausible buyer profiles per ICP and let the specific company's structure decide which one gets messaged, the same way you would let the account's actual situation decide the message content.
When the ICP is too narrow
There is an opposite failure mode worth naming. An ICP defined so tightly that it only matches a handful of companies in the world will starve a campaign of volume no matter how good the targeting is. If your situation-based filter only turns up 20 qualifying companies globally, either widen the geography, widen the headcount band, or add a second adjacent situation that produces a similar need. The goal is a list that can sustain months of steady, researched outreach, not a list you exhaust in three weeks.
Testing the ICP before committing to it
Before building a full campaign around an ICP definition, pull together a sample list of 20 to 30 companies by hand. If a researcher struggles to find real, specific reasons for more than half of them to be on the list, the ICP definition is still too vague and needs another pass. This small check, done before any outbound goes out, saves weeks of messaging the wrong accounts and is the same diligence step we run through with agencies before a campaign starts, as described on our how it works page.
Next: write down your current ICP in one sentence, then try to name five real companies it points to. If you cannot, the definition needs a rewrite before the list-building starts.