Ocean.io shows several scores. They answer different questions, so it helps to know which one you are looking at. This article defines each one, shows where it appears and explains what moves it.
Company fit, Segment score, Score and tiers appear in Discovery, Context studio and the workflow steps Segment fit and Recommendations. They need Segmentation, which is turned on for your whole organization. If you don't see Discovery or Context studio in the left menu, ask your account manager.
The scores at a glance
Score | Range | The question it answers | Where you see it |
Company fit | 0 to 100 | How well does this company match your scoring model? | Score breakdown, Segment fit step |
Segment score | 0 to 100 | How strong is the segment this company belongs to? | Score breakdown, Segment fit step, exports |
Score | 0 to 100 | Company fit and Segment score together | Discovery, Recommendations list, Segment fit step |
Tier | T1 to T4 | Which fixed band does the Score fall in? | Chips next to a score, tier filters |
Relevance | A, B or C | How similar is this company to your example companies? | Lookalike search results |
Buying committee match | Matches or Close to | How closely does a person fit a committee role? | Discovery, Buying committee column |
To decide where to start, use the Score and its tier in Discovery.
Company fit
Company fit measures how well a company matches the signals in your scoring model. It runs from 0 to 100.
A model is a list of signals. A signal is a value of a company attribute, such as an industry, an employee range or a technology. You mark each signal as positive or negative and give it a strength from 1x to 5x. You build models in Context studio.
Ocean works out Company fit like this:
It adds up the strength of every positive signal the company matches. Values you add to the same chip form a group, and only the best match in a group counts.
Every negative signal the company matches subtracts its full strength.
It compares the total with the most a company could earn from your positive signals. A company that earns the maximum scores 100. The score never goes below 0 or above 100.
A company with no data for an attribute doesn't match it. It isn't penalized on top of that.
For example, a model has three positive signals: Industry at 5x, Employees at 3x and Headquarters at 2x. Each step of strength counts as 2 points, so the most a company can earn is 10 + 6 + 4 = 20. A company that matches Industry and Employees earns 16 and scores 80. A company that matches Industry and Headquarters earns 14, then loses 4 for a matched negative signal at 2x. It earns 10 and scores 50.
In Discovery, each segment gets its own copy of the model when you apply it. A signal counts twice as much when some, but not nearly all, of the segment's companies have it. It counts half as much when none of them have it, and the same when 95% or more already have it. This is why the same company can show a different Company fit on the Models tab and in Recommendations.
Where a model comes from:
In Context studio, select + next to Models. Choose Build from followed by your CRM's name, such as Build from HubSpot, to get suggested signals from the companies in your CRM, or Build from scratch to pick every signal yourself.
A model is optional. If you set up a segmentation with No model, Ocean ranks companies by Segment score alone and the score has no breakdown.
What changes Company fit: the signals and strengths in the model, and which model is active. Score balance, Segment score, CRM win rate and tiers don't change it.
A model that a segmentation uses can't be edited. Choose Duplicate and edit the copy. Then, on the Models tab in Discovery, choose Add a model, pick the copy and choose Make active.
Segment score
Segment score rates a segment as a whole. It runs from 0 to 100, and every company in a segment shares it. By default, Discovery orders your segments by it. The Segments cards show other numbers, so you see the Segment score itself in the Score breakdown and in the workflow Segment fit step.
Ocean works it out once, when the segmentation finishes building. It compares the segments of that segmentation with each other, so a Segment score from one segmentation isn't comparable with one from another. How it is worked out depends on your CRM data:
CRM deal data: average won deal value multiplied by win rate. The best segment gets 100 and the others are scored against it.
CRM wins but no deal values: based on win rate. A segment with very few closed deals is pulled toward your overall win rate, so it doesn't swing to the extremes.
No CRM wins: based on structure. How tightly the segment's companies group together, how different the segment is from the others, and how many companies it has.
The Segment score stays fixed after the build. For your active segmentation, CRM numbers on the cards, such as win rate, refresh every day. The Segment score does not.
Score and Score balance
Score is the number in the Score column on the Recommendations tab of Discovery. It blends Company fit and Segment score:
Score = (Company fit x balance + Segment score x (100 minus balance)) / 100
The balance is the percentage that comes from Company fit. Ocean rounds the Score to a whole number, with .5 rounding up.
Score balance | Company fit | Segment score |
Fit only | 100% | 0% |
Mostly fit | 75% | 25% |
Balanced | 50% | 50% |
Mostly segment | 25% | 75% |
Segment only | 0% | 100% |
Until someone applies a different balance, Recommendations, the Tiers card and the workflow steps use Mostly fit. With Segment only, every company in a segment gets the same Score.
For example, Company fit 72.4 and Segment score 40 with Mostly fit gives 0.75 x 72.4 + 0.25 x 40 = 64.3. It shows as 64, which is T2.
To change the balance, open Discovery and select the Models tab. Each model panel has a dropdown that shows its current balance, such as Mostly fit. This is the Score balance. Choosing a value only previews the effect on that tab. On the active model, select Apply to save it, and Recommendations rerank. On any other model, Make active applies that model with the balance shown on its panel. A model only appears on this tab after you choose Add a model and pick it. Changing the model rescores every segment, so Recommendations update after that finishes.
A company that matches several segments appears once, with the segment that gives it the highest Score.
Tiers
A tier is a fixed band of the Score:
T1: 80 and up
T2: 60 to 79
T3: 40 to 59
T4: below 40
The same Score always gets the same tier. Ocean works it out from the rounded Score, so 79.5 shows as 80 and is T1.
You see tiers as chips next to a score, as the Tier 1 to Tier 4 filters above Recommendations, on the Tiers card of a segment, as bars on the Models tab, and in the workflow Recommendations step.
The tier counts on different screens cover different groups of companies. Compare them with care:
The Tiers card of a segment, and the Tier 1-2 number on a Segments card, count the segment's own companies, at the saved Score balance.
The filters above Recommendations count the recommended companies, which are lookalikes of your segments.
The Models tab counts a sample of up to 2,000 of your own companies, at the balance you selected on that tab.
Score breakdown
Hover over a Score, or select it, to open the Score breakdown. It shows the Score and its tier, the positive signals the company matches and any Negative signals. When the Score balance is below 100, it also shows how the Score is made: Company fit times its share, plus Segment score times its share. At Fit only it says "Segment score isn't weighted in". Companies ranked without a model have no breakdown.
The Matched signals column in Recommendations shows the first signal as a chip and +N more for the rest.
Relevance: A, B and C
In a lookalike search, the Score column shows a letter. This is lookalike relevance, and it is not Company fit. It compares a company with the example companies you gave. It needs no model and ignores CRM outcomes.
A, Very High: 0.66 or higher
B, High: 0.53 or higher
C, Moderate: below that
The badge doesn't show the number behind it. With Precise matching and several example companies, each result keeps its best match and shows which example it came from.
In the lookalike filter, open Advanced settings to change Min. relevance. A higher level returns fewer, more precise results. The default is C. Company matching mode can be Precise or Broad. Precise compares company descriptions. Broad ranks by how close a company's industry profile is to your examples. Min. relevance applies to Precise matching.
Discovery doesn't show a similarity number. It shows counts: Similar on a segment, and Open N similar companies in search. That button opens a company search with the Segment lookalikes filter, and the Score column then shows a tier chip with the Company fit instead of A, B or C. In that search view the Company fit is shown with decimals, and the tier is worked out from the rounded number.
Buying committee match
In the Buying committee column in Discovery, each person shows Matches or Close to followed by a role. Matches means the person is at the role's seniority level with a job title very similar to one of its example titles. Close to is a looser match. People are ranked by a score, with strong matches first, then close ones, then everyone else. See Buying committees.
Numbers next to the scores
With a CRM connected, these appear on the Segments cards in Discovery. They are not scores:
Win rate, all time: won companies divided by won plus lost companies, pulled toward the average of all segments when a segment has few closed deals. It needs at least 5 closed companies, otherwise it shows "Too few closed deals".
Revenue won, all time: the total value of won deals.
New deals, last 12 weeks: deals opened in the last 12 weeks. It appears after the first daily refresh.
Scores in workflows
Step | Output | What it is |
Segment fit | Score | Company fit and Segment score blended by your Score balance, as a whole number |
Segment fit | Tier | 1 (best) to 4 |
Segment fit | Company fit, Segment score | The two parts of the Score |
Recommendations | Recommendation score | The Score from Discovery, from 0 to 100. It is not rounded, so it can have decimals. |
Company search, lookalikes, People search | Match score | 0 to 1. It means similarity when the step has a lookalike example. |
Find buying committee | Buying committee score | Higher ranks first. It runs from about 0 to 3, not 0 to 100. |
The Segment fit output Segment win rate is the segment's plain win rate. It can differ from the smoothed win rate on a Discovery card.
Scores in exports
When you export, turn on Add Context columns to add five columns. Ocean score is the Score, and Lead score is Company fit. The others are Segment score, Segment and Matched signals. All scores are rounded to whole numbers.
Why a score changes
You make another model active, or apply a different Score balance.
You rebuild the segmentation. This also resets the Segment score.
You apply a model to the segmentation again, which resets how each segment's copy of the model is weighted.
Related articles: Segmentation scoring, Discovery, Context, Segment fit and Recommendations.