There are several kinds of data.
First, there is fundamental data (something directly and intentionally measured, like a list of stock investments or an image file on a camera). Second, there are facets or features of the fundamental data, often known as metadata (data about data). This might record, for example, the purchase price of one of the stocks or the date the camera took the photo.
Some secondary datasets may come from 'data exhaust'—something a digital system records or logs without a specific intent in mind.
Monetized datasets can provide either or both types of data. For example, a video stream of a security camera could be considered fundamental data, and logs of web traffic might be secondary data or metadata. What makes them data exhaust is that they are automatically collected, often for unknown future purposes.
Whether fundamental or secondary, these datasets may be monetized as raw, unprocessed data, without structure or labeling, or it might be aggregated and/or processed into a larger, more 'packaged' dataset.
For more information, read the first three modules of the Data Supply Chain (Acquire, Store, and Aggregate).
Sometimes datasets are distinguished by the terms 'traditional data' and 'alternative data' - a division often discussed in financial services. There, and in parallel industries, traditional data means measurements that directly describe fundamental things about an asset or other item of interest. These tend to be absolute and past-based.
Sometimes an alternative dataset can be used to infer something about a ‘traditional’ dataset. Parallel or secondary data can build on traditional data to infer additional detail or provide predictions.
For example, retail stores traditionally report their sales results in the quarter after the holidays, leaving investors to wonder about performance for a time. However, many retail stores have parking lots that can be monitored by security cameras or by satellite. Parking lot occupancy can be cross-referenced to nearby stores to guess what earnings will be.
If we ask the basic question “What will a retail store’s holiday sales numbers be?” we might have to wait until the next quarter for the report. To get some indication of these results sooner, we can use computational thinking: decomposing the result (sales) into the steps leading up to it (number of potential customers in stores, which can be inferred by how full parking lots are).
By pushing ourselves to see the root cause of or correlations to those sales, like an increased number of shoppers, we can get some indication of the outcome sooner.
Such a mindset lets us get an idea of what sales volume we might expect before we get the quarterly report that happens well after the holidays.
Traditional Data
Directly describes an asset’s market position or fundamentals
Broadly accessible, obvious, usually from within financial markets
Tends to be ‘now’ or ‘after the fact’
Tends to be free or low-cost
Often has a long, consistent history
Alternative Data
Can be used to infer fundamentals or something a/effecting fundamentals
Is ‘discovered’ or ‘mapped,’ sometimes not obvious—usually from outside financial markets
May be used to predict the future
Tends to be expensive
May be shorter or less consistent
Rather than selling complete data sets, some firms create value by providing insights based on data such as analyses of market activity, research or trends.
'Insights' companies include market research firms like Gartner and Forrester, as well as specific insight products, such as investment bank Credit Suisse's insight into the value of pharmaceuticals for industry benchmarking.
Market research firm Community Marketing Insights specializes in research about LGBTQ people consuming mass-market products.
These firms are not differentiated simply through the data they acquire and/or aggregate but through their unique analysis of it.
Their value propositions may include the underlying data or only the outcomes of their analysis.
Some of the insights for sale are basic mathematical analyses of statistics, while other insights firms construct qualitative frameworks (like Gartner's Hype Cycle), taxonomies, timelines, cause-and-effect studies, and/or integrate editorial perspectives.
Algorithms (mathematical models used for analyzing data) are another way to create value. A lot of energy goes into developing strong algorithms, and the differences in quality are enormous. While some organizations may create data science and machine learning teams, many firms just need access to a good algorithm to analyze their own datasets.
Monetization of algorithms generally falls into two categories: direct sale or licensing of an algorithm; or application program interface (API)-based access to it. The direct approach usually accelerates existing work on machine learning or operates inside a highly-secured environment, while the latter provides a commodity service.
Perhaps one of the best-known examples of an API-monetized algorithm is Google Image Recognition. Users of the service do not have direct access to Google's algorithm and need nearly no technical infrastructure to use it. They simply submit an image via an API call, and Google returns a weighted list of likely keywords. The service can recognize text, everyday objects, and brand logos. Causeit itself does this: we pass images of business cards we've received through Google to quickly identify companies and other relevant information for our internal customer relationship management tool.
Algorithms can be sold as a completely standalone service, like Google Image Recognition, or a core part of a more significant value proposition.
WeGlot, for example, helps website creators easily overlay their primary website with translated (or 'localized') editions and edit the various editions quickly. For example, an English-language website may have a Spanish edition. That functionality is useful on its own, but the real value of WeGlot is that they provide machine translation of your website using an aggregate of three different translation algorithms and a way to blend that machine translation with professional human translation where needed. WeGlot did not need to build an algorithm for translation—this would have been expensive and unnecessary—but instead was able to monetize other algorithms uniquely.
One of the ways data-centric products create value is through data-informed personalization. Traditional business wisdom says that an organization needs to choose between personalization and scale. A bank might have high-quality, personal financial advice for their biggest clients but only offer cursory, generic accounts to their broader customer base. Using a combination of big data (trends from their entire user base) and little data (specific data points about an individual user), wise firms can vastly increase their service's actual and perceived value while continuing to operate at digital scale.
Whether it's movie recommendations from Netflix, Google's customized slide shows based on location and contact data, or a calendar app that knows when you need to leave to get to your next appointment on time, the best digital tools already personalize things for us. If you look more deeply at their strategies, you'll find that most balance big data and little data elements.
For example, Apple Watches ask users for simple goals around physical activity and then 'nudge' users to achieve those goals with tailored advice and encouragement. The watch might say, "you only need to take a brisk twelve-minute walk to reach your activity goal!" or "you're usually more active by this point in the day, but there's still time." This 'nudge' mentality is based in behavioral psychology and works because it isn't the same reminder at the same time and in the same way for every user. Generally, more personalized strategies like 'reminders that you don't set' justify an app's request for personal information and reinforce goodwill between the user and their app or device.
Users even gladly participate with firms who are clearly attempting to offer them financial products like credit cards if the strategy is reciprocal enough. Services like Nerdwallet, Mint, and Credit Karma all ask for sensitive, personal information from users (often through secure APIs like Plaid) in order to first feed helpful information back to users, such as tips for improving their credit score or saving money. In that context, recommendations of credit cards and other financial products are more appropriate, better tailored, and more welcome than traditional broadcast ads for credit cards.
Faced with lots of options but limited resources, it can be challenging to know what to invest in. Data can always help these decisions, whether you're working at a financial institution and participating in public marketplaces; or inside a small firm and prioritizing your team's time. Inside an enterprise, business modeling tools can help leaders determine which products and services to expand or pivot and which to discontinue; or which markets to expand into or withdraw from.
When used well, data can build or improve relationships.
The most apparent application might be relationships between you and your customers or users, but you can also use data to enable third parties to connect, as online social networks do. The most basic data can provide incremental lifts to your relationships with customers, as with online shops that offer a special discount to customers on their birthday.
The real opportunity is to use more advanced data about your users to help them learn both about themselves and what you can do for them. Instead of focusing on transactions and sales, focus on personalization strategies, match-making with other users, recommending helpful content, or other valuable 'a-ha!' moments that delight them.
Just like a good friend or colleague, aim to be generous and helpful, anticipate their needs, and assist them in thinking through decisions. Use data to listen to your customers rather than target them.
- United Airlines uses basic data about its customers to express gratitude: phone agents thank callers for how long they've been flying with the company.
- Facebook applies its advanced data capabilities not just to advertising but also to the real needs of its users, like suggesting connections with old friends and fostering good memories by suggesting users revisit old content in which they and their friends are 'tagged' and expressed a positive sentiment.
- Apple strengthens connections between its users by prompting friends and family to congratulate each other on workouts or challenge each other to friendly competitions.
The are many strategies to create value with data. Whether evaluating your own opportunities or products on the market, it can be helpful to think through the elements of a digital value proposition. There are a few examples of digital value propositions below, based on the following ad lib.
1. Our [initiative or offering]
2. help(s) [customer group]
3. who want to [jobs to be done]
4. by using data from [sources]
5. to [reduce verb + customer pain]
6. and [increase verb + customer gain]
7. unlike [competing value propositions].