Improve your B2B segmentation with employee count data
It would be highly unusual to hear a company proclaim the fact that they do not make any attempt at filtering companies or segmenting accounts, and rightly so. By separating the good from the not-so-good, businesses are able to allocate their limited resources efficiently. One of the “Big Three” firmographic data points often used when segmenting accounts is company size, and one of the most telling indicators of company size is employee count. Employee count range as an indicator of company size will typically look something like this:
- Micro/No touch: < 10 employees
- Small: 11-50 employees
- Mid-size: 51-200
- Large: 200-1,000 employees
- Enterprise: > 1,000 employees
Now, there are many potential sources of employee count data, and we'll be discussing several of them here.
Business registries’ employee count data
One of the most common sources is local business registries, which will often provide a number based on companies’ official financial statements. It’s important to note that this number typically indicates the average number of full-time employees in the previous fiscal year, and this number has, by own estimation, at least two shortcomings.
First of all, the number provided by local business registries is not very “real time”. Companies typically file their financial statements several months after the end of their fiscal year, and, since they are reporting the average number of full-time employees in the previous year, not the number of full-time employees at the end of the period, that number might end up being a year old, which could end up giving you an inaccurate estimation of a company’s size, especially when it comes to growth companies.
Another potential issue with this official number is that it’s often specific to the legal entity that files the papers. This becomes an issue if the company has several business entities (holding companies, subsidiaries, etc.), as the number will only include the employees of that specific entity. It’s not uncommon to see huge global giants with employee count less than 10, and it’s because most employees aren’t employed by the entity you’re looking at.
LinkedIn’s employee count data
Many companies today have and maintain a Linkedin company profile, and it’s another popular source of employee count data. As you may well be aware, the administrator of the company profile page has the ability to choose an employee count range for companies, e.g. 51-200 employees. Additionally, LinkedIn shows the number of people who have selected that company as their current workplace. In this way, LinkedIn provides two different employee count estimations.
In our review of LinkedIn as a source of employee counts, we’ve found that companies will sometimes, for reasons unknown to us, end up picking a completely incorrect employee count range. The self-reported nature of this number makes it difficult to know the degree to which you can trust it.
Comparatively, we’ve found that the number of LinkedIn individual profiles connected to a company is a good and reliable indicator of actual company size. However, it’s only a good estimate for those industries where people have a LinkedIn profile. In some traditional industries, e.g. manufacturing, we’ve found that the number of employees connected to a company on LinkedIn can be a lot lower than the real number. Some companies also have regional Linkedin company profiles, e.g. Acme Benelux, and, in those cases, Linkedin’s number is not a total employee count but more like an in-country or in-region employee count.
CRMs’ employee count data
Some modern CRMs, such as HubSpot, provide their own CRM insights engine. This means that they’ll automatically populate data fields containing basic company information, such as employee count, when a new company or account is created. The actual source of this data varies, but, no matter the source, it’s a big step towards improving CRM data coverage since it’ll decrease the number of companies without basic firmographic data dramatically. Based on our experience, these data engines will often perform well for large, well-known companies, but make more mistakes when it comes to small and local companies.
Vainu’s employee count data
You might have seen it coming, but we have our own predictor of employee count. Our global company database relies on our proprietary AI model to predict employee count ranges. The model is based on a training dataset of more than one million companies, and it predicts a company’s number of employees by using the company’s website as input data. Its prediction is one of the following five employee count ranges that our model believes to have the highest likelihood of being correct:
- Micro company: 1-10 employees
- Small company: 11-50 employees
- Mid-sized company: 51-200 employees
- Large company: 201 - 1,000 employees
- Enterprise: 1,000+ employees
We use an ordinal regression and classification approach in our models. Our model is multifactorial, with it taking into account factors such as: Web traffic, number of office locations, detected web technologies, mentions of certain key phrases, etc.
This might all sound terribly complex, but you’d be surprised to learn how easy it is for a model to differentiate between companies based on their website content. For example, it’s not hard for the model to see that a company that has huge amount of web traffic, uses several enterprise technologies, talks about topics such as “global footprint” and “investor relations” and has 20+ office locations listed is more likely than not going to be larger than the one that has low traffic, only has a few sub pages, built their website on top of a do-it-yourself website builder, and describes themselves as a small family business.
So how do we evaluate the performance of the model? By looking at the precision, of course. But that’s not the only metric we’ve found valuable. We also pay close attention to something called Mean Absolute Error (MAE). Together, these metrics tell us how accurate our predictions are and the degree to which our predictions deviate from the actual values, which makes it easy to make tweaks to the model.
Our own internal testing indicates that our model generally outperforms other employee count models, but it truly shines when it comes to minimizing the large, important mistakes. What that means is that it might be difficult for the model to choose between 51-200 and 201-1,000 if the company has roughly 200 employees, but it has an easy time avoiding big and important mistakes, such as enterprise companies being classified as micro companies. In other words, our model still makes mistakes, but those mistakes are often small in magnitude, i.e. the model might predict the nearest neighbor.
As most readers have probably already figured out, our model has its own significant limitation: It requires companies to have a website to function. If there’s no website for a company, we’re not able to provide an estimate. But, on the other hand, if a website does exist, then we can provide close to 100% coverage for our customers, something which is very difficult to achieve if you use the more traditional data sources for employee count range.
Where can I hear more about Vainu’s data?
Hopefully, you’ve learnt a bit about some of the different sources of employee count data and their unique advantages and disadvantages. If you want to hear more about our employee count data, or our broad selection of other available data fields, and discuss their applications in more detail, then please feel free to contact us via our chat. You might also consider signing up for our free trial, which would allow you to access our entire database for a limited time.