Every deal you've ever won is trying to tell you something.
Not in some abstract, philosophical way. Literally. Sitting in your CRM right now are dozens, maybe hundreds, of closed-won deals, and hidden inside them is a pattern. A very specific, very useful pattern that describes exactly what a winning deal looks like at your company. Not in general. Not based on benchmarks or industry reports. At your company, with your product, in your market.
The problem is that almost nobody extracts it.
Not because they don't want to. Because it's genuinely hard to do properly, and the tools most sales teams rely on were never designed for it. But understanding why it matters (and what happens when you actually do it) is the difference between a sales team that targets with precision and one that's essentially throwing darts with a blindfold on.
What your closed-won data actually contains
When a deal closes, your CRM captures a constellation of signals. Some are obvious: the company's industry, their size, the deal value. Others are less obvious but equally important: how long the sales cycle took, which lead source brought them in, what region they're based in, who the decision-maker was, what stage they entered the pipeline at.
Individually, these fields are boring. They're the stuff of CRM hygiene reminders and quarterly reporting decks. But collectively, they contain the fingerprint of your winning customer.
Here's what you can learn from a proper analysis of your closed-won data:
Which characteristics actually predict a win. Not which ones you think matter, but which ones the data says matter. Most teams are surprised. Company size, the thing everyone qualifies on first, often accounts for a relatively small portion of the win signal. Industry alignment and deal value frequently matter far more.
How much each characteristic matters relative to the others. This is the crucial bit that a CRM pie chart can never show you. Knowing that "industry matters" is useful. Knowing that "industry alignment accounts for 40% of your win signal while company size accounts for 12%" is actionable. That's the difference between a description and a scoring system.
Whether you have one winning profile or several. Many teams discover they actually win in two, three, or even four distinct patterns. A high-velocity mid-market motion and a slower enterprise motion, for example. Each has completely different characteristics. An ICP that tries to describe both at once describes neither accurately.
What your anti-ICP looks like. Equally valuable: which characteristics predict a loss. The deals that linger, stall, consume pipeline review bandwidth, and then die quietly. They have a pattern too, and it's usually the inverse of your winning pattern.
Why nobody does this
If it's so valuable, why isn't every sales team doing it? There are four reasons, and they compound each other.
Nobody has time. The VP Sales is in pipeline reviews four days a week, not sitting in Excel running statistical analysis on historical deals. RevOps is building dashboards, configuring workflows, and firefighting data quality issues. The analysis falls between the cracks because it's important but never urgent.
Nobody has the skills. A proper win analysis requires more than sorting a spreadsheet by industry and eyeballing the results. You need to identify natural clusters in multi-dimensional data, weight characteristics by predictive importance, and account for interactions between variables. That's data science work, and most sales teams don't have a data scientist. Nor should they need one for something this fundamental.
Nobody has the right tools. Your CRM can show you averages. It can give you a pie chart of won deals by industry or a bar chart by company size. What it cannot do is tell you that the combination of industry alignment and deal value predicts wins with three times more accuracy than either signal alone. CRM reporting is designed for backward-looking summaries, not forward-looking pattern recognition.
Get pipeline intelligence insights weekly
Join sales leaders learning to turn CRM data into pipeline confidence. No spam. Unsubscribe anytime.
Nobody has the mandate. Who owns this? Is it sales strategy? Marketing analytics? RevOps? Product? It falls between departments, and everyone assumes someone else is doing it. Meanwhile, leadership is making strategic decisions about hiring, territory design, and market expansion based on assumptions from an offsite two years ago. The actual answer sits untouched in the database.
How to actually do it
If you want to run this analysis yourself, here's what's involved. Fair warning: it's not a quick exercise, but the output is worth more than almost anything else you could spend the time on.
Step 1: Export your closed-won deals. Pull every deal that closed in the last 12 to 24 months. You want enough volume for the patterns to be statistically meaningful: at least 30 deals, ideally 50 or more. Include every field you have: industry, employee count, deal value, sales cycle length, lead source, region, owner, stage history, any custom properties.
Step 2: Clean the data. This is where most attempts die. CRM data is messy. Industries are spelled differently, company sizes are inconsistent, fields are blank. You'll spend more time cleaning than analysing. Accept that now.
Step 3: Look for clusters, not averages. The biggest mistake is averaging everything. "Our average deal size is £45K" tells you nothing useful if half your deals are £15K and half are £80K. Those are two different customer segments with two different sales motions. Look for natural groupings. Do your wins cluster around certain industries? Do deals from certain lead sources win at dramatically different rates?
Step 4: Weight the characteristics. This is the hard part. For each attribute, you need to determine how strongly it correlates with winning. A simple approach: take each characteristic, split your deals into groups (e.g. deals where industry matched your top segment vs. those that didn't), and compare win rates. The characteristics where the gap is widest are the ones that matter most. If you want the full method with the maths shown, we walk through it step by step in how to build a weighted ICP from your closed-won data.
Step 5: Build a rubric. Combine your weighted characteristics into a scoring system. "Industry alignment: 40%. Deal value fit: 25%. Company size: 12%. Lead source: 10%. Region: 8%. Decision-maker seniority: 5%." Now you have something a rep can actually use to evaluate their pipeline on a Tuesday morning.
Step 6: Score your open pipeline. Apply the rubric to every open deal. Suddenly you can see which deals match your winning pattern and which don't. The ones that score low aren't necessarily going to lose, but they're statistically less likely to close, and your team should know that before investing another month of effort.
Step 7: Revisit quarterly. Your market changes. Your product evolves. The ICP drifts. What won deals six months ago may not be what wins deals today. The analysis isn't a one-off project. It's a recurring discipline.
What does a closed-won analysis look like in practice?
Here is the whole method run end to end on a worked example. The numbers below are illustrative, not from a real customer, but the method is exactly what you would run on your own data.
Meet Harborview Analytics, a fictional B2B data platform. Over 18 months they closed 280 deals: 62 won, 218 lost. That gives a baseline win rate of 22%, and, crucially, it gives them both halves of the comparison. You cannot weight signals from wins alone. The pattern only appears when you can ask "how did deals with this characteristic perform against deals without it?"
The averages lie first. Harborview's average won deal is £38K. But nobody actually buys at £38K. Plotting the deals shows two clear clusters: 41 wins between £12K and £18K (mid-market logistics software firms, mostly inbound, 45-day cycles) and 21 wins between £60K and £85K (enterprise retail groups, mostly referral, 160-day cycles). Two segments, two motions. Any analysis that averaged them would describe a customer that does not exist.
Then the win-rate gaps reveal the weights. For each characteristic, split all 280 closed deals into "matches the winning pattern" and "does not", and compare the win rates:
Industry (logistics software): 34% matched, 11% not. A 23-point gap.
Deal value (in-band for segment): 29% matched, 16% not. A 13-point gap.
Lead source (inbound): 27% matched, 18% not. A 9-point gap.
Company size (in-band): 25% matched, 20% not. A 5-point gap.
Region (UK and Ireland): 24% matched, 20% not. A 4-point gap.
Normalise the gaps and you have your weights: industry 42%, deal value 24%, lead source 17%, company size 10%, region 7%.
Notice what just happened. Company size, the characteristic almost every team qualifies on first, carries 10% of the signal. Industry carries four times as much. Harborview's reps have been sorting their pipeline by the fourth most predictive signal on the board.
Now score an open deal. A logistics software company, £15K opportunity, came in through inbound, 700 employees, based in the Netherlands. Industry matches (42), value matches (24), source matches (17), size and region do not (0). Score: 83 out of 100. Another open deal, a facilities management firm at £30K from a cold list, 140 employees in Leeds: company size and region match, nothing else does. Score: 17 out of 100. Same pipeline, same reps, and now a defensible reason to spend Tuesday morning on the first deal rather than the second.
Two honest caveats before you run this on your own data. First, you need the losses, not just the wins; without closed-lost records there is no comparison group and the weights cannot be computed. Second, 280 deals is a comfortable sample. Below roughly 50 wins the gaps get noisy, and you should treat the weights as directional rather than settled.
Or let the data do it for you
Everything I just described is exactly what pipeline intelligence automates.
Telepath Pro connects to your CRM, reads your closed-won data, and runs this entire analysis in minutes, not weeks. It identifies your winning segments, weights each characteristic, produces a scoring rubric, and scores every open deal against it. The output is a T-Score from 0 to 100 for every deal in your pipeline, telling you how closely it matches your actual winning pattern.
No spreadsheets. No data science skills required. No quarterly manual refresh. The system updates as new deals close, so your ICP evolves as your business does.
Your deals already know what wins. The question is whether you're listening.
See what your closed-won data has been trying to tell you. Three minutes, free: telepath.pro
Related reading: