> ## Documentation Index
> Fetch the complete documentation index at: https://docs.thedatacity.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Similar companies

> How Industry Engine finds the companies most like the one you're looking at, and how to get similar companies from the API.

On a company's page in Industry Engine, **Similar companies** lists the companies most like it. Use it to build a competitor shortlist, find a peer set for benchmarking or make a first cut of a market map.

Industry Engine builds the list with composite similarity. It compares what companies say they do on their websites. It also compares what kind of business they are: their sector, their size and where they operate.

## Why we compare more than website text

Website text is one of the strongest clues to what a company does, and it carries the most weight in composite similarity. But two companies can write about the same things without competing. A global pharmaceutical company and a local pharmacy both write about medicines, patients and prescriptions.

Composite similarity also checks whether two companies are in the same sector, of a similar size and with a similar geographical reach. So for the pharmaceutical company, other large pharmaceutical companies rank above the local pharmacy.

## What we compare

* **What a company says it does**: the meaning of its website text.
* **Its sector**: its [RTICs](https://docs.thedatacity.com/our-data/rtics/what-are-rtics.md), and its [RSICs](https://docs.thedatacity.com/our-data/proprietary-data/what-are-rsics.md) where we have them. We don't use SIC codes directly.
* **Where it operates**: the local authorities its locations are in.
* **Its size**: its employee count, a size band based on that count, and its turnover.

Sector codes that often appear on the same kinds of company count as close to each other, even when the codes differ. Places work the same way. Companies of a similar size count as closer than ones that are much larger or much smaller.

## How the list is built

**1. Choose companies we can compare**

We include active companies that have a website and at least one usable signal: a sector or location code, an employee count, turnover or website text. Companies without a website aren't included.

**2. Score each pair**

We combine a company's signals into one score for how close it is to each other company. If a company is missing a signal, such as turnover, we leave that signal out and give the others more weight. The company is still scored.

**3. Rank and tidy the list**

We keep the closest companies, ranked by score. Each website appears once: we drop a neighbour with the same website as the company, or as a closer neighbour. We also leave out pairs we've already marked as poor matches.

## How to read a result

A high score means two companies are close across all these signals together. They can score highly when they share a sector and a similar size, even if they're in different towns. Two companies with similar websites can score lower when their size or sector differs.

If a list looks surprising, one signal is usually thin:

* The company has little website text, so the match leans on sector, place and size.
* It has no RTIC, so the sector match rests on its RSICs.
* It has several sites, so sharing one location is only a partial match on place.

## Similar companies in the API

The Industry Engine API can return two lists of similar companies. To include each one in a `POST /v2/companies/batch` response, set its option to `true`. The [batching guide](https://docs.thedatacity.com/api-reference/guides/batching.md) explains that endpoint.

| Field | Option | Method |
| - | - | - |
| `SimilarCompositeCompanies` | `includeSimilarCompositeCompanies` | Composite similarity, as described on this page. It's the list you see in Industry Engine. |
| `SimilarCompanies` | `includeSimilarCompanies` | Semantic similarity, which compares website text only. |

Semantic similarity finds companies that describe themselves in similar terms, so it returns more matches across sectors. It suits broad exploration, such as finding companies to start a Smart List. Our [blog post on semantic similarity](https://thedatacity.com/blog/building-the-industry-engine-similarity-score-update/) explains how it works.

> [!NOTE]
>
> Similar companies isn't how we build RTICs or Smart Lists. Those start from a training set and cover a whole sector. Similar companies only returns the companies closest to the one you started from.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.