Finding FAANG: The value of Behavioral Data
TL;DR – High-quality behavioral data represents the most significant growth opportunity for businesses today. The world’s most successful companies are harnessing it to dominate their markets. To compete with them, or drive growth in your own industry, you need to rethink how you view behavioral data; there’s so much more potential than packaged analytics solutions would have you believe.
Most companies have already understood the need to compete with data. Many have built amazing platforms with transactional and demographic data, but are seeing a diminishing return on investment. There’s only so far you can go with these types of data, only so much you can improve on your algorithms, before you reach a limit.
Behavioral data is the missing piece. By combining behavioral with transactional and demographic data, you see a step-change in value.
Netflix, Amazon, and other tech giants have figured this out and as a result, have built pioneering businesses based almost entirely on behavioral data. With behavioral data as its core value prop, Netflix is now valued at $176bn. From the content that they produce to the people that star in it - it’s all heavily influenced or informed by behavioral data. And of course, every user’s Netflix experience is completely unique and personalized. Behavioral data is the reason why it is so hard for anyone to compete with Netflix.
Read on to learn why behavioral data is so powerful, with some examples to demonstrate the step-change in value that it offers in comparison to transactional and demographic data alone.
What’s the difference?
Behavioral data is radically different from both transactional and demographic data:
- Demographic data describes person-level attributes.
- Transactional data tells you that a person has made a decision, to buy or not to buy something.
- Behavioral data allows you to capture the relationship and interactions your customers have with your digital interfaces, in granular detail.
Behavioral data allows you to capture an infinitely broader range of decisions, as well as the contexts in which those decisions were made.
Behavioral data is evolving
Forget click tracking, pageviews, and session count. The old-school understanding of behavioral data does not communicate just how much value can be derived from accurate behavioral signals; as has been the experience of the many companies we have worked with over the last 10 years.
No other data set can provide the insights and the deep understanding that behavioral data can. As a result, new use cases for behavioral data are being discovered every day, and existing use cases advanced upon.
The richness of behavioral data makes it high-performance fuel for AI because it is both explanatory and predictive. With a deep understanding of customer behavior, you can both predict and influence their decision-making process.
Example use cases
Here are three specific use cases that demonstrate the difference in visibility gained with behavioral data, in comparison to transactional data alone, as told through the story of a fictitious sock shop.
Personalization
On the day that a customer buys socks from a sock shop, right before purchase, they also look at four different types of flip flops from the same store. They keep one tab open for seven minutes, scrolling the product image carousel, and reading the warranty. They even close the tab and come back to it again later, but in the end they close it.
This journey, made visible with behavioral data, amounts to a convincing picture that the customer wanted to buy the flip flops, but just wasn’t pushed over the edge to purchase them. We can therefore assume that next time a customer comes along that looks and acts like this customer, there’s a high chance that recommending flip flops at checkout will result in a purchase.
This level of visibility allows the sock shop to build a product recommendation engine to suggest product pairings at checkout. As behavior is the best predictor of behavior, behavioral data is the most reliable informer of these product recommendations.
With transactional data alone the sock shop would have remained blind to the fact that the customer was interested in flip flops at all. They would be limited to recommending products based on what items are most commonly bought together, in no personalized capacity.
Perhaps pink and blue socks are most commonly bought together. But a first-time buyer who doesn’t know the brand, for example, may be more likely to buy pink socks along with the sock shop’s best-selling item instead.
With behavioral data, the sock shop could also see if the customer saw the socks that they bought on the website months before, but waited until Black Friday to actually make a purchase. This is an indication that the customer is price sensitive, which can further inform personalization efforts.
Behavioral data offers much more insight about intent to purchase, and about the needs and wants of individual customers. Deeper audience understanding with behavioral data allows users to be bucketed according to these kinds of factors.
Snowplow BDP for personalization
Personalization is a broad and varied use case, with the above being just one example of how it can look in action.
Since implementing personalized product recommendations on their website, for example, one of our customer’s data team has been able to increase its ‘Add-to-basket’ rate by 200%. The team has also used Snowplow data to uncover non-intuitive associations between different products.
By training an ML model, this data team has created a recommendation engine at the checkout that uses data from customer interactions to provide targeted, highly personalized product recommendations. This is a marked improvement on the previous setup, which was based on a manual, rules-based association model.
The Globe and Mail (Canada’s leading news provider) used Snowplow behavioral data to build a content scoring system, based on value to users, that is used to inform the personalization of 99% of the customer experience. The result so far has been a 17% CTR increase, and a 10% increase in subscriber acquisition rate. These insights are used everywhere, from informing journalists on what to write, to driving low latency decision making in the newsroom.
Software.com identifies the most productive times of the day for individual developers and takes automated action to ring-fence that time (whether it’s by halting notifications or preventing calendar invites). By identifying a developer’s most valuable time in a personalized capacity, Software.com works to protect and enhance it, and ultimately improve the productivity of the team. Since implementing rich, high-quality behavioral data by integrating Snowplow into its data stack, Software.com has seen a 250% YoY increase in its user base.
Fraud detection
The sock shop introduces an incentive to encourage new account sign-ups in the leadup to Christmas. Every new account created gets €5 free credit. This makes the sock shop susceptible to application fraud, with fraudsters creating many accounts in order to take advantage. At scale, this could amount to huge losses for the company.
To create an account, customers are required to fill out a form. One of the use cases for this form is to prevent fraud by recording transactional and demographic data, which is used to assess the fraud risk of the individual. This could mean, for example, asking the customer to provide a mobile number, or an address - if the mobile number or address are the same across multiple accounts, the account may be fraudulent. The sock shop must balance preventing fraud with ease of use so as not to deter people from buying from them altogether.
Using behavioral data to create a fraud score is a much more accurate way of detecting fraud, as signals can be tracked across the entire customer journey. A customer that arrived at the website through a marketing campaign is unlikely to be fraudulent, as fraud is typically preplanned rather than opportunistic. Did the customer spend time reading the landing page? Have they engaged with the brand before? Did they browse products before deciding to make a purchase? If not, the fraud score is added to.
By setting parameters for these behaviors, and calculating a resultant fraud score over time, the sock shop is using behavioral data to determine how genuine an activity is.
Scoring continues once the user reaches the form that gates the account creation page. Rather than analyzing the contents of the application form, the sock shop can analyze the way the form is interacted with. If the form was auto-populated instantly, or if keystrokes are consistent rather than sporadic, this might indicate that the customer is a fraudulent machine. Form-filling behavior is much more predictive of fraud than the values inputted into the form. Behavioral data helps to establish whether this user is a computer, a fraudster, or actually just a daughter in search of a novelty pair of socks for her father for Christmas.
Marketing attribution
As customer journeys become increasingly complex, accurately assigning credit to marketing activities is more important, and more difficult, than ever before.
Another customer of our sock shop first comes into contact with the brand by scrolling past an ad on their social media feed. They don’t click on the ad but instead arrive at the website via the sock shop’s climate pledge a number of months later - having followed a link on the UN’s website, where the shop is listed as a sustainable brand.
After the customer views the climate pledge they go on to read more about the company. The sock shop’s Brand team has been running a brand awareness content campaign. As the only eco-friendly sock brand in the world, they have written a blog series to showcase the fact. The customer flicks through this content, before browsing products. They don’t buy anything but they do sign up for the newsletter. When the sock shop sends them an email they open it and browse socks for a while longer, though still no purchase is made.
Weeks later, Marketing launches a campaign promoting a new t-shirt range. On the same day, an ad from the campaign brings the customer to the site once more, and though they browse t-shirts for a while, they still do not make a purchase. Instead, they click on the socks recommended in the sidebar - the same eco-friendly socks they read about all those weeks before - and buy these instead.
Which campaign gets the credit?
Effectively attributing credit to all of these touchpoints is impossible based solely on transactional data, without making massive assumptions.
Transactional data is only able to track how many socks the customer purchased, and when. From this, the team could make the assumption that the ad for the new t-shirt range can be attributed credit for this purchase, though there is no way of knowing whether the customer who bought the socks actually saw the campaign, as the data is limited. The marketing campaign would therefore get all the credit, but this is entirely wrong.
High-quality behavioral data provides full visibility over the entirety of this customer journey. This allows the sock shop to measure incrementality, and correctly attribute credit to the brand awareness campaign, the link from the UN’s website, and the social ad from months before. Touchpoints that would have otherwise remained unconnected to this customer journey.
With behavioral data, the sock shop can evolve its marketing attribution over time, as the visibility it offers allows the team to account for and ultimately predict user intent. They can deepen their understanding of how factors like seasonality and persona influence attribution, and eventually they can start to use ML to assign credit.
High-quality behavioral data is the key ingredient here. Packaged analytics tools like Google Analytics are black boxes; you have no idea how data is being used and transformed, or what assumptions are being made. As data is third-party owned, ITP/ETP restrictions also mean that client-side cookies expire after 7 days (in some cases even less, depending on a user’s privacy settings), making lengthy customer journeys impossible to track.
Packaged solutions are also not built with AI and Advanced Analytics in mind. The data these platforms create requires a complicated series of transformations to make it suitable for advanced use cases. Investing in behavioral data that is built to evolve is a much more scalable solution.
Marketing attribution with Snowplow Behavioral Data
With Snowplow, Green Building Supply built an end-to-end attribution model covering the full length of its lengthy and complex customer journey. This involved a process of stitching together divergent data sources across multiple channels and devices. The ability to accurately calculate where revenue was generated and how their ads and channels were performing enabled GBS to optimize those channels for success. The result was a 137% boost in revenue, an increase in new customers from ads by 13x, and a 106% conversion rate increase.
Behavioral data answers questions that transactional data cannot
This piece has only scratched the surface of the opportunity that behavioral data presents. Its limitless capabilities mean that every day new applications are being discovered, and existing applications are advanced upon.
Snowplow enables companies to capture high-quality, granular, behavioral data that’s ready for use in AI and ML models, the way that Google, Facebook (Meta), and Netflix are doing. With better behavioral data, on top of your pre-existing transactional data structures, you get better results. By generating data with quality in mind, we’re assuring data quality at the source.