How Do You Define an ICP From Closed-Lost Deals?

Can closed-lost deals really tell me who my ideal customer is?

Yes, but mostly by telling you who your ideal customer is not. Closed-lost deals are the clearest record you have of buyers who looked, engaged, and still said no. Sorted properly, they show which company types, sizes, and situations never convert, so you can draw a sharper edge around the ideal customer profile you already have.

Most teams build an ICP by looking only at closed-won deals. That is natural, because wins feel like proof. The trouble is that a short list of wins can make almost any segment look promising. Two good customers in healthcare and the founder starts saying "we sell to healthcare."

Lost deals correct that bias. They are usually more numerous than wins in an early B2B company, and they carry the details nobody wants to talk about in a pipeline review. That makes them the most honest data set in your CRM, if you are willing to read it.

Why do most ICP definitions ignore lost deals?

Lost deals get ignored because the data on them is messy and the story is uncomfortable. Reps close them out quickly, pick a vague reason like "no decision," and move on. Nobody goes back. So the CRM holds hundreds of losses that look identical, and the ICP gets written from memory and wins instead.

There is also a quieter reason. A lost deal invites the question "was it them or was it us?" Looking at a loss honestly can point at the pitch, the pricing, the demo, or the follow-up. It is easier to write an aspirational ICP than to admit that some losses were self-inflicted.

I think that discomfort is exactly why lost deals are worth the effort. The ICP is supposed to be a decision tool for where to spend sales time. If it never says "do not pursue this," it is not doing its job. Closed-lost is where the "do not pursue" rules come from.

What data do I need from my CRM before I start?

You need an export of closed-lost deals with the company record attached: industry, employee count, region, deal source, deal size, stage reached before the loss, days in pipeline, and the lost reason. Pull at least the last twelve months, then pull closed-won for the same period so you can compare like with like.

In HubSpot, Closed lost reason is one of the default deal properties, described in HubSpot's own documentation as the reason the deal was lost. There is a matching Closed won reason property too. If your team never filled these in, the export will show it fast, and that gap is your first finding.

Do not worry about perfect data before you begin. I would rather work from a rough export and a few hours of reading deal notes than wait a quarter for clean fields. The reading is where the insight lives. The fields only tell you where to look.

How should I sort the reasons a deal was lost?

Sort every lost deal into one of three buckets: wrong fit, wrong time, or wrong execution. Wrong fit means the company could never have been a good customer. Wrong time means they could be, just not now. Wrong execution means you had a fit buyer and lost them through your own process.

This is the framework I use, and the reason is simple. Only the first bucket should change your ICP. Wrong-time losses belong in a nurture sequence with a reason to come back. Wrong-execution losses belong in a sales process review. Mixing them together is how teams end up excluding good segments because a demo went badly.

The sorting is manual, and it should be. Read the notes, the emails, and the call summaries for each loss. "Too expensive" can mean the company is too small to afford you, which is fit, or that you failed to show value, which is execution. Only the context tells you which one it is.

How do I turn wrong-fit losses into ICP rules?

Look for attributes that repeat across wrong-fit losses and rarely appear in wins. Turn each repeating attribute into an exclusion rule with a clear threshold, such as a minimum employee count, a missing tool in their stack, or a region you cannot serve. Write each rule so a rep can apply it in seconds.

A good exclusion rule is specific and testable. "Companies under a certain headcount with no dedicated sales ops person" is a rule. "Companies that are not serious" is not. If a rep cannot check the rule from a LinkedIn page, a website, or an enrichment tool like Clay or Apollo, it will not get used.

Then compare each rule against your wins. If a rule would have excluded a meaningful share of your best customers, it is too broad. Tighten it or drop it. The goal is a fence that keeps out the deals that waste time, not a fence that keeps out revenue.

What patterns usually show up in a closed-lost review?

The common patterns are a company size floor below which deals stall, a missing prerequisite such as no CRM or no ops owner, a buying committee you never reached, and a use case your product only half solves. Each of these points to a fit rule, a qualification question, or both.

I want to be careful here. These are patterns I look for, not results I can promise you will find. Every company's losses are different, and the value of the exercise is that it is specific to your pipeline. A generic list of "common lost reasons" is the opposite of what you are trying to build.

What I can say is that the patterns are usually boring and obvious in hindsight. That is fine. An ICP that encodes a boring, obvious truth your reps keep ignoring is more useful than a clever one nobody trusts. If you want a sense of how this pairs with the win side, my notes on turning closed-won notes into an ICP update cover the other half.

How do I make sure the new rules get used by sales?

Put the rules where reps already work. Turn the strongest exclusion rule into a required qualification property at the first deal stage, and keep the full ICP in a one-page doc linked from the CRM. A rule that lives only in a slide deck will be forgotten within a month.

HubSpot lets you set conditional stage properties so certain properties appear when a deal is moved to a specific stage. Its documentation says that if a property is set as required, users cannot create or update the record until they set a value for it. I use that to make the qualification question unavoidable at the right moment.

The same move helps the next review. If you make the closed-lost reason required on your lost stage, and give it options that map to fit, time, and execution, your next export sorts itself. Six months from now you will have a data set you can trust instead of one you have to rebuild by hand.

How often should I repeat a closed-lost ICP review?

Run it once a quarter in an early-stage company and twice a year once the motion is stable. The point is not to rewrite the ICP every time. It is to check whether the fence is in the right place as your product, pricing, and market change underneath it.

A quarterly review fits naturally next to pipeline planning. Pull the losses from the last quarter, sort them into the three buckets, and ask one question: did any new wrong-fit pattern appear? If yes, write a rule. If not, leave the ICP alone. Stability is a feature here.

If you are working with a small number of deals, the same approach still works, just with more reading and less counting. I wrote about running a win-loss review with only five deals for exactly that situation, and the bucket method carries straight over.

What should you do next?

Export last year's closed-lost deals with company data attached. Read every loss and tag it fit, time, or execution. Write exclusion rules only from the fit bucket, test them against your wins, and put the strongest one into your CRM as a required question. Then schedule the next review.

If your losses point to two segments that both seem to buy, that is a different decision, and I covered it in picking an ICP when two segments both buy. Either way, the lost deals are already sitting in your CRM. The cost of reading them is a few hours. The cost of ignoring them is a sales team chasing deals that were never going to close.

If you want help running this review or wiring the rules into HubSpot so they stick, reach out at pravinkumar.co. I build the systems behind go-to-market for B2B teams, and this is one of my favorite places to start.