8 min read
Combining Datasets is one plus one equals three
Most managers who work with Alternative Data rely on transactional datasets – anonymized credit and debit card transactions – that are aggregated across large consumer panels. Why? Because this data has become widely accessible and consumable thanks to numerous vendors that work with the data and help the buy-side monetize it.
But often you need more data. Think about mobile first companies like social media or streaming services. For those types of businesses you need to track how their users are engaging with their apps (app intelligence data). For ecommerce websites you want to know how customers interact with the brand through their website (clickstream). For large retail and big box outlets, foot traffic data is critical. Each of these data types offers a new and unique perspective into different companies and the behaviors of their customers.
However, combining different types of data is where the real magic happens. Is increased foot traffic actually leading to more sales? Does the time spent on a website influence transaction amounts? Do downloads translate into revenues? How long does it take to materialize? Combining data means not only building better forecasting models and avoiding potential mistakes, but also being able to answer much more interesting questions..
And yet, many managers only subscribe to one data type. Why don’t managers integrate other types of data more often? Price is often cited as a reason, and we are not just talking about the financial cost. Cost includes the time it takes to evaluate, onboard, test, and integrate the data into your investment process. This type of data analysis is a heavy strain on resources. New data is a big unknown for managers that takes upfront investment, with no certainty of value at the other side of integration.
Thankfully, technology can help managers take the leap and significantly reduce costs and time to value. We detailed these challenges and how investors overcome them in our eBook “Top Challenges of Implementing Alternative Data.”
Here are the top reasons why managers are taking the leap into new types of data and using technology to overcome the most common challenges.
If you have followed our blog, you’ll know that we often talk about making better financial predictions. Forecasting models have evolved as new data has become available. Our research has shown that as you add inputs (unique datasets) to your forecasting models, the prediction errors decrease.
This makes intuitive sense. A single data source may have blind spots or biases that other data sources can help offset. Datasets that do not have much overlap to your existing arsenal usually add the most value.
When data seems to tell you an unbelievable story, don’t necessarily believe it. If this is a “six-sigma” event, assume there is something wrong. Question the signal, go back to the data source and perform a post-mortem. If the data is wrong, what could be the cause? Does anyone else see this big move, or is it just me? If it’s just me, why?
Another great way to interrogate data is the look at a completely unrelated data source. If credit card purchases have dipped at Wayfair, what about their app usage? What about their website? There should be tell-tale signs hiding in other datasets that can either confirm or repudiate a large inflection. When you have found something real, build enough confidence to put on a trade by leaning on a new dataset.
Combining datasets not only helps in analyzing a particular company or sector, but it also allows for broader diversification of your portfolio. Different industries and sectors need different data inputs. Searching for signal across these datasets will expand your forecasting universe to a broader range of companies, compared to only using one type of data for idea generation. For example, foot traffic might be critical for retail, while app downloads could be the key for finding inflections for technology companies. By incorporating various types of data into your process, managers can diversify their exposure across industries, while also tailoring their predictive models to the nuances of each business they are analyzing.
Sometimes the most valuable insights aren’t immediately obvious. When managers combine different data sources, they can uncover hidden patterns or relationships between variables that would be missed with just a single dataset. It helps analysts be more creative with their data and ask much deeper questions. For example, an unexpected correlation between app engagement and spending could help identify trends that drive long-term revenue growth. These patterns help managers make smarter, more unique investment decisions when leveraging the interaction of diverse types of data rather than relying on one source alone.
In a market where everyone has access to similar baseline data, such as credit card transactions or earnings reports, it’s increasingly difficult to generate Alpha. That’s where combining different datasets gives you a serious edge. By layering unique data sources—like app engagement, social sentiment, or geolocation data—on top of more traditional transactional data, managers can gain insights that are harder to find, acting before the market catches up and elevating their investment process at the same time. Think of it as moving beyond the “vanilla” data sources and into specialized insights that provide earlier indicators of company performance.
Have you ever thought about leveraging more data for your companies? Talk to us about it.