Brand Model: How Helio Quantifies the Seemingly Unquantifiable

By Arvind Ganesan and William Dowling 

In a world where consumers have access to an ever-growing variety of products, it can be increasingly difficult for products to stand out. A good brand, which has traditionally been one of the ways for a product to win over competitors, has become more important than ever before. In the consumer goods space, we think a lot about branding. When a customer has a couple seconds to browse the aisle of a grocery store and dozens of choices for each product, the product associated with a strong brand can often be the product that makes its way into the shopping cart. We have found that brand is an  attribute that is very predictive of future revenue growth.

But what makes a strong brand? Is it recognition? The feeling that comes from seeing an image or product that you are familiar with? When we see a strong brand, we often think not only of the company’s products but of the way those products make us feel or could make us feel. As an example of this, when you look at the below logos, what comes to mind?

Screen Shot 2018-06-11 at 11.38.00 AM.png

Most people probably think of a functional but stylish coat that you can take outdoors, a phone that can do everything a computer can, or a car that is well-engineered and sophisticated. Recognition certainly plays a part in maintaining a good brand, but what kind of recognition is the right kind? Let’s play the same game again- this time with a different set of logos. What comes to mind now?  

Screen Shot 2018-06-11 at 11.38.09 AM.png

Accounting scandals, nicotine laden cigarettes, and seagulls covered in oil aren’t exactly the hallmarks of great brands even though most people would probably recognize these companies on sight.

A good brand is not just about mass recognition, it’s about recognition in a way that resonates positively with consumers. This can be general or targeted towards select consumer segments. Not everyone knows about guitars, but those who do know about guitars know that Gibson Les Paul is a great brand.

Quantifying what makes a good brand

We built the brand model in our machine learning platform Helio to capture many of the signals that  indicate brand strength at scale. We calculate brand scores not just for one company or a hundred, but for over 500,000 companies operating in the consumer space. We collect billions of data points to build brand scores for companies, and once we have these scores, we think hard about how to best use them. In scoring these brands, we look at things like the size of their followings on social media, how often they share content, consumer engagement with that content, the rate of growth of followers, and a combination of other point in time values and historical growth. All of these are controlled for the category that the brand belongs to and adjusted to different rates of digital brand engagement that we have seen across different categories.

We want to know not just what brands consumers are familiar with, but what brands they are familiar with in a positive way. This helps us solve for issues of positive recognition (e.g. Apple) vs negative recognition (e.g. Camel). We can also quickly determine what brands have picked up momentum lately.

Above we noted that Gibson Les Paul is regarded as a great brand in the guitar community, but it wouldn’t make sense to compare that against a brand like Nike which is well known even by people who don’t care about shoes. Luckily, with Helio we can sort this problem out. Because it classifies companies according to 103 primary-level categories (baked goods, cheese, backpacks, etc.), we can make sure that we’re comparing apples to apples (or guitars to guitars) when it comes to brand.  We establish brand ranks by taking the brand scores of companies within a certain category and calculating how that score ranks against other companies in the same category. For example, companies with a brand rank of 10, will be in the top 10% of brand in their primary category and companies with a brand rank of 1 will be in the bottom 10% for the category.

Let’s look at an example of this in action. Below we scored the brands of around 700 eyeglass companies and tagged some for illustrative purposes.  Helio identifies around 1,200 eyeglass companies, so around 500 companies were left out for readability. The X Axis is deliberately left blank here because we can filter by any other Helio feature we want. For example, we could look at the top ranked brands based in California, or those that are designer, or those that make under $5 million in revenue and we’re founded in 2013. This allows us to be very granular in the filters we run.

Brand Score of 700 Eyeglass Companies Screen Shot 2018-06-11 at 11.40.58 AM.png

You can see that scoring like this helps us quickly and efficiently surface those companies with the best brand elements.

Brand score matters to us because we’ve found it to be an attribute that is very predictive of future revenue growth. To validate this, we back scored our brand metric one year prior to the actual revenue growth data we get on consumer companies. In essence we try to recreate what if anything the brand score would tell us before a company has actually displayed high future growth that all investors look for in their investments. For this, we took actual revenue growth data from June 2016 onwards and back scored the brand metric one year before the revenue growth data, aggregating the results across multiple vintages (sliding windows) starting from June 2016. Overall the results are promising with the top decile of the brand score having a much higher median expected revenue growth than other deciles. We found that those companies with a brand rank of 10 had 16X the expected revenue growth of companies with a brand rank of 1 and about 3X the revenue growth of brands with a rank of 5, as seen below.

Relationship between Brand Rank and Median Future Revenue Growth

Screen Shot 2018-06-11 at 11.44.09 AM.png

How we use brand score

The great thing about brand score is that it can surface companies that weren’t even on our radar before, and because it is based on consumer sentiment about a company, brand score can pick up signals that a company may be interesting to look at before the company even has strong offline distribution.  In an increasingly omnichannel world where a lot of companies get their start selling direct to consumer, we think this is really important and helps us identify great breakout brands early.

Today, we use Helio brand scores in a lot of targeting work. Our credit and equity funds may reach out to companies that our model identifies as having a strong brand that also look attractive across a variety of other metrics we track. We may also use brand scores when diligencing a company and deciding whether or not we should make an investment in the business.  We talk about the strong brand elements of 4505 and Kosas in our write-ups about our investments in these companies.  Sometimes we will use brand score to examine new trends within a category to determine whether or not we think those trends have wind behind them; for example we recently looked at the brand score of functional mushroom companies.  

We get a lot of use out of our brand model today, but like every model in Helio, it’s not done yet.  Our data science and engineering teams are hard at work feeding new data into the model and adding new features to capture more of what makes a brand strong. Some key areas we want to improve on in the future are better understanding and quantifying the consumer perception of the brand, capturing more data sources that could potentially reflect how a brand portrays itself, and incorporating more effective frequency and recency of data that are good proxies for brand. More to come.  

%d bloggers like this: