The concept of an ideal customer profile (ICP) has been around for a long time, but like its cousin, the minimum viable product (MVP), it is often misunderstood. In this post, I’ll offer some background commentary on the ICP concept and then build into one of my favorite sayings: your ICP starts out as an aspiration and becomes a regression.
There are four common questions around ICPs. Here they are, along with my answers:
- Is the ICP about a person or a firm? Both. It should include firmographic as well as role (or persona) information. Example: VPs of sales at technology companies between $500M and $2B in revenues. Here, we included the size and industry of the company along with the target buyer’s title.
- Should an ICP include a problem to be solved? Yes. VPs of sales have lots of different problems from recruiting to training to pipeline management to forecasting, just to name a few. Thus, your ICP should include the ideal person at the ideal company and the problem you’re looking to solve for them.
- Should the ICP include adjacent systems? Yes. Deciding at the outset if you want to focus on customers using specific, adjacent systems is often critical (e.g., NetSuite vs. Oracle vs. Xero for core financials, Salesforce vs. HubSpot for CRM). The alternative is drowning in integration work while never having the time to support the idiosyncrasies of a given package which, when you do it, is usually adored by customers.
- Should an ICP include sales qualification criteria? No. The ICP is about the buyer: this is the person we’re looking for. They have this job at this kind of company. Whether they’re out shopping right now, whether they have budget, whether they have a buying timeframe and purchasing authority are all important qualification questions, but they are not part of the ICP itself. People differ on this, I know [1].
Because the world is imperfect and it’s difficult to find “Mr. or Ms. Right” every time, it’s useful to think of the ICP as a bullseye. The absolute perfect customer is in ring 0, the next level off in ring 1, after that ring 2, et cetera. Note that I have no religion about the things you vary across the rings, but the usual candidates are: job title, industry, size, adjacent systems, and problem (aka use-case). And you might do them in unusual combinations. For example, if you think a director of finance with a budgeting problem is about as good as a manager of finance at a bigger company with an operational reporting problem, then you can put them both in ring 2.
The idea is to give you a simple and flexible model to agree on who to target and who to prioritize across sales, marketing, and product.
For a zero-to-one startup, you might focus exclusively on ring 0. As you grow you will typically get more use-cases, more industries, more adjacent systems, and thus more rings. That’s fine as long as you’re defining the rings clearly and triaging them into: hot pursue, pursue, and slow-roll or some similar encoding system.
With a few clearly established tiers you are now ready to report on ARR and pipeline by ICP tier to see if “you’re walking the talk” when it comes to your ICP. At many companies, you will find the majority of the ARR and pipeline [2] outside ring 2 or 3. In these cases, you simply aren’t living your ICP and instead suffering from a faux focus. The usual cause is an inability to control the sales force and prevent their default “chase anything” behavior [3].
The ICP is typically born in the founder’s head as an hypothesis. Think: I bet if we can build something like this, it will solve a problem like that. By the time a company has been founded and a product built, it becomes an aspiration. Think: I want to sell to people like this to solve a problem like that. So you sharpen your definitions of this and that, and add some additional targeting criteria like company size, industry, or adjacent systems. And then you go off and sell.
Let’s say it works. One day you look up and you’re now $50M or $100M in ARR. Congratulations. Should your ICP still be an intuition-driven aspiration? No. It should be a regression. Reality happened. Let’s find out what reality is telling us.
Are the people in our ICP ring 0 really our best customers?
Well, what do we mean by best? Do they have higher ASPs? Do they have shorter sales cycles? Do they renew at higher rates? Do they expand at higher rates (e.g., NRR)? Do we win new deals at higher rates? Do they give us higher CSAT scores?
At the first order, these are all just simple segmented metrics calculations that you can and should do. Your QBR and board decks should show these key metrics segmented by ICP tier [4]. And — since not all these metrics can be important — your e-team also needs to have the conversation about “what do we mean by best” so you can have a common, precise definition of the “best” customers that you are trying to target [5].
But the best answers to these questions are not performed using segment analysis [6]. Segment analysis is great for finding anomalies — e.g., why do we have a higher win rate in ICP tier 3 than tier 1? But it’s not a great technique for actually finding the impact of different variables on the success criteria.
For that, we need regression analysis. Regression analysis will tell us which variables most strongly correlate with the outcomes we want (e.g., that the strongest predictor of renewal is company size, not CSAT) [7]. A good regression analysis will tell you not only which factors most correlate with the outcome, but it can also be the best way to bucket those variables (e.g., the real breakpoint is at 250 employees, but your initial segment went from 0 to 500).
Odds are, when we do this kind of analysis we’ll find lots of surprises. Some of your intuition will be proven correct, but some won’t. And you’ll likely find entirely new variables (e.g., number of data scientists) that you didn’t even consider in your initial ICP exercises.
So this is why I like to say that your ICP starts as an aspiration — about who you want to sell to — and over time becomes a regression. Because one day you will have lots of data to analyze to determine who your best customers are — subject to your definition of best, of course — as opposed to who you thought they would be.
# # #
Notes
[1] Regardless of where you land at least be aware there are two types of criteria: those that change slowly or not at all (e.g., company size, adjacent systems, industry) and those that can change overnight (e.g., out shopping, budget, authority). My analogy here is dating: you can meet the right person at the wrong time. It doesn’t change the fact that they’re the right person. (And that’s why God made nurture tracks.)
[2] Think of pipeline as a potential leading indicator of ARR. Well, it should be, at least.
[3] Using the ICP in territory and compensation plan definitions can help with that. Think: you only earn commissions on customers in ICP rings 1 through 3 within your geographic territory. That will get your sellers’ attention.
[4] Note that I’m kind of using ICP tier and ring synonomously here and that’s generally OK. However, in cases where you have lots of rings, I would then sort those rings into tiers, so ring is the more specific and tier the more general term. For me, because I like simplicity, I want to see ICP segmentation in at most 3-4 buckets, so if there are N rings underlying those, I’d prefer to hide those by using 3-4 tiers.
[5] You probably don’t want marketing targeting high LTV prospects when sales wants to target high win rate ones. We should all be on the same targeting page.
[6] One of the key problems being that the segments themselves were somewhat arbitrarily chosen. Sure, we did our best to guess who’d be our best customers. But who are they actually? We may have used not only the wrong bucket boundaries (e.g., 100 emps vs. 500 emps) but even the wrong dimensions (e.g., maybe company size is a poor predictor and industry or use-case a powerful one).
[7] I cheated here on purpose to see if you were paying attention. Thus far, we’ve largely said the ICP is about a company (firmographics) and a role/persona. But here I’ve said that company size is a better predictor or renewal than CSAT — and CSAT isn’t a ICP-style criteria. The reality is these tools can do precisely that, looking across a wide range of input variables to see which most influence the output. Obviously, for marketing targeting purposes we don’t want CSAT to be an input variable to the model, but for renewals analysis we sure would.
