Segmentation isn't a self-serve toggle, it's enabled per organization. If you don't see it in your account, ask a manager to reach out to [email protected].

There's no longer a dedicated Segmentation item in the left-hand menu. The most direct way in today is from Context: select New segmentation from the rail to start one, or open an existing segmentation from its list there.

Starting a new segmentation takes you to the "Let's find your best-fit segments" screen: "Segmentation groups your companies into clusters based on what they actually do. You can see who fits, how many of them exist, and where to focus."

Choose how you want to start:

Paste domains
Paste your company domains directly into the text box, one per line, or comma/space-separated. You'll need a minimum of 50 domains. A requirements checklist confirms this as you type. Click Confirm once met.

Upload a CSV
Drop your CSV file, or click Select file to choose one. Requirements: the file needs headers, and you'll need a minimum of 50 company domains. An example format is shown on-screen (e.g. a "Domain" column listing novonordisk.dk, walmart.com, costco.com).

Connect a CRM
Connect HubSpot, Salesforce, or Pipedrive to segment your already-synced companies. Once connected, confirm to use your synced companies, or pick a different source instead. You'll then be asked: "Which pipelines should we segment? Pick one or more to focus the segmentation, or leave it on all pipelines." Each pipeline is listed with its company count (e.g., "All pipelines, 100 companies," "Sales Pipeline, 56 companies"), select one, several, or leave it on all, then click Continue.

At any stage, click the pencil/edit icon at the top of the screen to go back and choose a different starting method.

Once your data's in, you land in Preview mode, a two-panel view.

Right panel: your segments, each showing its company count, a cohesion percentage, and a few sample domains. At the top, you'll see overall totals (e.g., "4 segments, 637 companies"). Each segment has Split and Drop actions available directly.

⚠️ Important, what "cohesion" actually means: cohesion is not a success or win rate. It reflects how tightly clustered and keyword-confident a segment is, how similar the companies within it look to each other. A high cohesion score (like 93%) just means the companies in that segment are more uniform; it says nothing about whether that segment actually converts or performs well. A segment's real performance only shows once closed deals feed CRM data back into it, until then, win rate shows as 0%, not because the segment is bad, but because there's no signal yet.

The AI conversation (left panel): the assistant proactively reviews your segments and suggests actions, with reasoning attached, for example, suggesting a merge because two segments "describe the same buyer type with near-identical overlap." Each suggestion appears as its own confirmable action.

You can also just ask it questions, for example, "where lies my best success rate and why is there 93% on the second segment?" gets you a plain-language explanation distinguishing cohesion (keyword-match confidence) from actual conversion performance, and tells you honestly when there isn't yet enough closed-deal data to answer that at all.

There's also a free-text box, "Or give a custom instruction...", to type your own request instead of relying on the AI's suggestions.

Doing it manually instead: in the right-hand panel, check the box next to any segment(s) you want to act on, Deselect all and Merge buttons then appear, letting you combine selected segments directly without typing anything.

Once you're happy, click "Approve segments." This brings you to the final step:

Pick a scoring model. Choose an existing lead-scoring model (built in Context) to score this segmentation with, or continue without one for now. New segmentations no longer get a fresh, editable model built from scratch during setup, they're scored by whichever model you pick here.

Once you've made your selection, confirm to run scoring.

Finalizing: Segmentation then runs lead scoring and post-processing on your approved segments. This is the final step and can take a few minutes, feel free to navigate away, you'll be notified when it's ready.

This entire Preview and finalization flow is free to use, refining and scoring doesn't cost credits.

Once finalization completes, you'll get a notification in your Activities tab. From there, head back to Segmentation to see your results.

Switching to the Companies tab (from the main Segmentation overview) shows every individual company across your segmentation, rather than grouped by segment. For each company, you'll see:

You can filter this view by segment using the dropdown at the top (defaults to "All segments"), and search by company name.

Clicking "Show lookalikes" takes you out of Segmentation entirely and into Ocean's general search interface, pre-loaded with a Company lookalikes search based on the company (or companies) you selected. From there, you get the full standard search experience, segment lookalike filters, size, headquarter location, and all other regular filters, letting you refine and explore beyond what Segmentation originally grouped.

At the top, you'll see the segmentation's name (e.g. "HubSpot"), when it was created and last updated, and its source (e.g. "HubSpot").

Below that, a summary bar shows:

Below the summary, you can toggle between two views: Segments and Companies.

In the Segments view, you can search for a specific segment by name, or sort by Pipeline value, Won companies, Lost companies, No deal, Lookalikes, Market breakdown, or Segment name.

Each segment in the list shows a summary card:

All of this data comes directly from your own CRM, the wins, losses, and open deals you've actually had or your data in case you've uploaded a CSV or pasted companies, combined with Ocean's lookalike algorithm, which finds how many similar companies exist in Ocean's database beyond what's already in your CRM.

Actions available per segment:

🔎 Inside a segment: the detail view

Clicking Open on any segment takes you to a dedicated detail page with four tabs (a fifth, Workflows, appears too if your org uses Workflows):

Overview, a full performance snapshot for this specific segment:

Companies: every company in the segment, each showing:

Lookalikes: companies in Ocean's database that resemble this segment but aren't in your CRM yet, shown with the same scoring detail as the Companies tab.

Buying committee: The people most worth talking to at companies in this segment's total addressable market (TAM) - including lookalike companies not yet in your CRM, not just the segment's labeled member companies - based on the actual pattern of who you've successfully closed deals with (seniority and department). Shows matched contacts across that wider pool, ranked by fit to your won-deal pattern, with quick actions to access email/phone and see CRM status (Net new vs. In CRM). You can control how many people are shown per company.
​

(If there isn't yet a won-deal pattern for a segment, e.g. no closed deals yet, this tab will say so plainly rather than guessing.)

Under "Matched contacts," you'll see contacts across the segment, ordered by fit to your won-deal pattern, with:

You can control how many people are shown per company using the "People / company" count selector (e.g. set to 3).

(If there isn't yet a won-deal pattern for a segment, e.g. no closed deals yet, this tab will say so plainly rather than guessing.)

Click View model (top-right of the segment detail view) to see the signals and weights behind the model scoring this segment. This view is read-only, models are built and edited in Context, not from inside a segmentation.

The view shows rule groups like Company Size or Industries (Ocean.io), each listing specific value ranges (e.g. "11-50 employees," "51-200 employees") with a weight per range, positive weights boost that range's importance, negative weights suppress it. Scores are grouped into four tiers (T1 highest, T4 lowest), scoped to this specific segment, not the whole segmentation or any other segment's list.

This view doesn't link out to edit the model directly, models are built and managed from Context. Click Explore all lookalikes to send this segment's expanded lookalike set straight to Ocean's general search tab, where you can refine it further with all standard filters.

The Segmentation overview page lists every segmentation you've created, as a table. From here you can:

You can also create a new one from Context by clicking New segmentation.

Wherever you're viewing companies or contacts, inside a segment, in the Companies view, or in a lookalike search result, exporting is simple: just select what you need and export it. No separate setup or extra steps required.

Can I use Segmentation via API?
Yes, with some limitations, API-created segments are fully accessible, but UI-created segments (via CSV/CRM upload) aren't yet available through the API.

Is my data shared?
No. Your data is private and used only to build your segmentation model. It's not shared, and it's not used to train any global models.

Use the in-app chat, or reach out to us at [email protected].