🔢 Overview: two different scores
There are two distinct scores you'll encounter in Segmentation, and it's important not to mix them up:
Company score & Tier badge (T1–T5): how well a specific company matches your model, and where it ranks within the list you're currently viewing
Trait weight (-100 to 100): how important a specific value or range is within a single trait, when you're fine-tuning the model yourself in Edit model
🏆 Company score & Tier badges, what they actually mean
Each company gets a score reflecting its fit strength, and a Tier badge (T1–T5).
Important: the tier badge is relative to the current list you're viewing, not a fixed quality band. The system takes the highest and lowest scores in that specific result set (e.g. the page or segment you're currently looking at) and splits that range into five equal slices, the top slice is T1, the bottom is T5.
This means two companies with very close scores can still land in completely different tiers, if they happen to sit near opposite ends of that particular list's range. For example, scores that are only ~1.4% apart could show as T1 and T5 side by side. This is expected, intentional behavior, not a bug or a scoring error.
In practice, this is especially noticeable in segment searches, where most companies are already decent fits to begin with, scores cluster tightly together, which makes the tier labels spread out more dramatically than the raw score difference would suggest.
How to actually use this:
Trust the score number for how strong a company's fit genuinely is
Use the tier badge as a quick "best to worst within this list" guide, not as a standalone quality grade on its own
A lower tier (e.g. T3 or T4) doesn't mean a weak fit, it may just mean that company isn't at the very top of this specific batch, while still being a strong match overall
⚖️ Trait weights: fine-tuning what matters
When you edit a segment's model manually (see Using Segmentation → Edit model), you're setting trait weights, how much a specific value or range should count toward a company's score.
Trait weights run on a scale of -100 to 100:
100 = a very strong positive signal, a company matching this exactly is a great fit
0 = neutral, this value has no real effect either way
-100 = a strong negative signal, a company matching this is a poor fit
Example: if you're weighting company size, and your ICP targets 5,000+ employees:
A range like "1,000–5,000 employees" being excluded from your ICP might be weighted around -82, a strong negative signal
A range matching your ideal target closely might be weighted +63, a strong positive signal
The closer a company aligns with your high-weighted (positive) values, and the further it stays from your negative-weighted ones, the higher its overall company score will be, and the more likely it is to land in a top tier within its list.
Trait weights ≠ company score, and neither is the same as the tier badge. Trait weights are the inputs you control, they define what your ICP model considers important. The company score is the output, the result of a company being evaluated against all your weighted traits combined. The tier badge is simply where that score ranks within the current list.
🧠 How the score is calculated
The score is based on how closely a company matches patterns learned from your input, whether that's a CRM sync, CSV upload, or pasted domain list. When you create a segment, the system builds a model of your ICP using signals including:
Industry and business model
Company size and growth stage
Keywords and positioning
Technology usage
Hiring and growth indicators (when available)
Geographic and market signals
Each company is then evaluated against this model and assigned a score, ranked into a tier based on its position within the current list.
🧠 How CRM data is used in segmentation
When segmentation is connected to your CRM, it uses all available data from the connected CRM dataset to build and refine your segmentation model. This includes all objects and fields that are part of the integration scope at the time of syncing.
At the moment, the system does not support selective field or dataset scoping, the full connected CRM dataset is used as input, ensuring the model has a complete view of your customer base when building your ICP.
📊 Score distribution view
Segmentation provides a score distribution panel showing how companies in your results are spread across tiers. This helps you understand:
How concentrated your ICP is
Whether your input list defines a narrow or broad profile
How selective your segment results are
Since tiers are relative to the current list, this distribution reflects the spread within that specific view rather than a fixed, universal benchmark.
🔄 Why scores may change over time
Scores may shift when:
You update your input customer list
You adjust trait weights or signals (via Edit model)
New data is added to the underlying company database
The model improves with better signal extraction
Because tier badges are relative to the current list, a company's tier can also shift simply based on what other companies appear alongside it in that list, even if its underlying score hasn't changed at all.
🧩 Key takeaway
The company score measures actual ICP fit strength, that number is what to trust. The tier badge (T1–T5) is a relative "best to worst" indicator within whatever list you're currently viewing, not a fixed, universal quality grade, so two companies in different tiers can still both be strong fits. Trait weights are the tunable inputs behind the model, letting you define what a good fit actually looks like. Together, these let you rank companies by relevance, compare leads consistently, and identify strong ICP matches at scale.