Walk into any mid-market credit union, community bank, or fintech and ask for the ICP. You will get a document. The document will name a segment, list a few attributes, sketch a buyer persona, and include a couple of representative use cases. The document will be polished. It will be roughly two years old. And it will have been written more from aspiration than from evidence.
That document is not useless. It captures something about the leadership team's intent. But it almost never captures the actual pattern of where the institution wins. The gap between the two is where most strategic clarity hides, and it is the part of the competitive intelligence engagement that surprises leadership teams most often.
The honest exercise starts with the closed deals from the last twenty four months. Not the open pipeline. Not the marketing-qualified leads. Not the segments the institution is targeting next year. The closed deals. The ones that won. The ones that paid. The ones that stayed.
What the win data tells you that the strategy deck does not
The closed-deal set has a structure. Cluster it by firmographic and behavioral attributes and the structure shows up. Some clusters win at above-average rates. Some clusters bring above-average economics. Some clusters retain at above-average rates. The intersection of those three is the real ICP. The clusters that are in the official document but not in the win data are the marketing team's projection of where it wishes the institution won.
That gap is the source of most ICP work that actually changes outcomes. The leadership team that sees the gap has three options. Lean into the segments the win data confirms, which is usually under-served by the current marketing program. Quietly de-emphasize the segments the official document names but the win data does not support, which is usually where marketing spend is leaking. Run a deliberate test to see whether the un-supported segments could be made to work, which is honest in a way the official document was not.
None of those options requires the marketing team to admit it was wrong. The official ICP captured intent, and intent matters. The real ICP captures evidence, and evidence is what the strategy needs to compound. Both are useful. The combination is what most teams have never built.
What ICP looks like for a credit union
The relevant variables for a credit union ICP cluster around field of membership, asset tier, digital maturity, and product mix. A credit union with a multi-county geographic field of membership and a tech-forward digital banking platform wins different members than one with a single-employer SEG and a legacy core. The leadership team that knows which version of itself it actually is, and which members that version actually wins, runs a sharper marketing program. The team operating from a generic "everyone in our geography" ICP runs a fuzzier one.
The win data usually surfaces three or four specific patterns. The young household opening a first deposit account because the digital onboarding was painless. The mid-career borrower refinancing because the local credit union came up in their AI-mediated search for the lowest reasonable rate. The small business owner moving the operating account because the local banker called twice. Each of those patterns implies a different marketing posture. The institution that recognizes its actual mix, instead of marketing to all three equally, wins more of the segment it is built to win.
What ICP looks like for a community or regional bank
Community and regional banks layer in commercial concentration and deposit relationship depth on top of the consumer ICP variables. A bank that wins at above-average rates in commercial real estate inside a specific MSA has a structurally different ICP than one that wins at above-average rates in small-business operating accounts across a multi-state footprint. The two banks may have similar balance sheets and similar marketing programs. They are running different businesses, and the leadership teams that name the difference operate with more clarity than the ones that do not.
The harder ICP question for banks is usually the negative one. Which segments has the bank been chasing that the win data does not support? The answer often surprises the leadership team, because the segments the bank has been chasing are the segments it talked about chasing five years ago and never re-examined.
What ICP looks like for a fintech selling into financial services
For a fintech, the ICP variables are not about the end consumer. They are about the buying institution. Budget owner. Technical environment. Regulatory posture. Procurement path. Pace of decision. Most fintechs selling into financial services have an official ICP that names a segment by asset tier or institution type, and a real ICP that is narrower than the official document admits.
The narrower version usually looks something like this. The institution has a specific role that owns the budget for the category, and the role is the same across the institutions that have closed. The institution has a technical environment with a specific characteristic, sometimes a specific core, sometimes a specific middleware layer, that made integration tractable. The institution has a decision pace that matched the fintech's sales cycle, which often correlates more with the executive sponsor's personal pace than with the institution's asset tier. The closed deals come from the intersection of those three. The pipeline outside that intersection looks promising and rarely closes.
The fintech that names the intersection, builds the marketing and sales motion around it, and is willing to be selective about which institutions it pursues, compounds faster than the one chasing every name in the asset tier band.
The step away
The hardest part of ICP work is not naming the real ICP. It is the deliberate decision to step away from segments the data does not support. Marketing programs accumulate segments over time. Each one had a reason when it was added. Few are ever pruned. The cumulative effect is that the marketing program tries to win in too many places at once and wins clearly in none of them.
The leadership team that says, in writing, "we are not pursuing segment X for the next twelve months because the win data does not support it," is making a more sophisticated decision than the team that quietly adds another segment to the target list. The step away is the move most leadership teams hesitate on, because it feels like giving something up. It is the move that produces the most growth, because it frees the marketing team to win clearly where the win data says it can.
What changes when AI is in the mix
The work of clustering closed deals, comparing economics, and recommending segments is not new. What is new is the cost. Pre-AI, this was a senior analyst sprint that mid-market institutions struggled to justify. AI compresses the synthesis to days. The output is the same. The economics changed.
For a mid-market institution, that change means an ICP refresh is no longer a once-every-three-years exercise. It can be a quarterly check against the moving win data. The leadership team that runs the check sees the segments shifting in close to real time. The team that does not is operating from a snapshot that was current the last time the strategy deck was rebuilt.
Working on a decision the real ICP would inform?
Atlas Instinct runs ICP development as part of competitive intelligence engagements for credit unions, community and regional banks, and fintechs selling into financial services. The work surfaces the gap between the official ICP and the real one and produces a recommendation the leadership team can act on. Every engagement is led directly by a senior operator and scoped to the decision in front of you. Start a conversation.