As the perceived value of enterprise data continues to soar, one challenge remains: how can we engage in a shared economy to use data more effectively? To address this question in this post, we’ll discuss ways of fostering user engagement around data.
The term “system of engagement,” as it applies to IT and data infrastructure, was coined by Jeffrey Moore, who wrote the research paper, Systems of Engagement and the Future of Enterprise IT. Jeffrey noted that while utilizing the system of record as a means of maintaining enterprise IT and data infrastructure has become widely acceptable, the social approach to enterprise IT is still considered somewhat progressive and risky. Although some enterprise data software practitioners treat such approach with skepticism, which is perhaps due to the heavily regulated nature of data management; the fact that there are B2C organizations that have capitalized on it is indisputable.
The success stories of businesses like Amazon, Google, and Yelp inspired us to look into the possibilities of adopting these crowdsourcing practices and applying them to the world of enterprise data management and governance. The passion for building a community around enterprise information has been an integral element of our company’s infrastructure, particularly via Collibra University and Collibra Community. Likewise, our product strategy is intentional about boosting collaboration and user engagement around data and related processes.
In this blog, we’re previewing our latest engagement functionality: ratings for information assets. Enterprise-wide adoption of this functionality can result in improved data literacy, visibility and trust, operational efficiencies, accuracy of insights and analytics, and more.
The ratings feature will enable users to provide subjective feedback on information assets, which is meant to serve as enrichment to the objective asset characteristics such as rule-based data quality, certifications, or articulation scores. This feature could potentially become a game-changer for your data management strategy. If the ratings and reviews can be crucial to finding the trusted products on Amazon, hotels on TripAdvisor, or businesses on Yelp, as studies suggest, it lends strongly to the belief that they could also be deployed to improve internal trust in enterprise data, analytics, and insights. Just as the “Ratings and reviews have become a key part of the consumer path to purchase, second only to price with swooping 95% of shoppers declaring they ‘consult ratings and reviews while shopping,’” could it be that the path to data provisioning and data-driven decision making can be further streamlined if a similar functionality is fully adopted as part of the enterprise information platforms and data catalogs?
Let’s take a look at some examples of how ratings can be applied within Collibra:
Our vision is to enable users to rate assets of certain types, such as data sets (pictured), on a familiar five-star scale, and to write the corresponding free-text reviews in support of the given rating. For example, a certain subject matter expert can give a five-star rating to a certain data set coupled with an explanation of why they found it useful. The process is similar to consumer product ratings on the Amazon website. The user ratings are added to the distribution chart for the given asset, featuring metrics such as average rating and the total number of ratings. Any edits or removal of ratings are captured in the history of the asset, as shown:
By default, ratings are enabled for the Report and Data Set asset types and their derivatives. You can configure this feature by asset type to be further activated across your platform. The ratings information can also be shown in an asset table or a set of tiles:
You can sort in the ratings columns or use them in filters for searching:
The future opportunity for this functionality is virtually limitless. Ratings data is user-generated content that can be aggregated and leveraged by AI and analytics in several diverse ways. For example, it might be interesting to combine such ‘wisdom of the crowds’ with the statistics pertaining to the usage of this data. This would generate data popularity scores, which would refer to the importance of data from both behavioral and user-defined standpoints. Among other purposes, these scores could be used to highlight the most important data and bring it to the surface of the data landscape. We’re testing machine learning use cases that use sentiment analysis to track and verify reviews across the platform. This will help administrators and stewards to flag reviews that contain conflicting information, such as positive reviews with one-star rating and vice versa, and better understand the potential issues that might be causing negative reviews.
We anticipate the adoption of the user-defined ratings features to vary depending on the use cases, governance models, and data culture within a given company. The more federated and decentralized the data governance style, the more important the collaborative, communal aspects would be in engaging with the processes around it.
Besides being beneficial in helping to find and understand the data that users need, these features are crucial for bottom-up adoption of data tools. By providing insights and ultimately generating more traffic and potentially refining usage of data assets, such features can become trustworthy business differentiators. With the regulatory aspect of enterprise data formation being one of the major drivers, the importance of making a way for the ‘wisdom of the crowds’ to be incorporated in the data landscape could be easily overlooked. In the long run, however, it can generate enduring value.
As the first installment of the ratings feature rolls out to our beta users this month, where do we go from there? We’re considering introducing a multi-dimensional aspect to ratings in the future to be able to rate assets separately on different key criteria, such as content or usefulness. The usefulness of the asset could be highlighted through explicit use of an endorsement flag as well, since a five-star rating may not always be equivalent to a recommendation to use a certain asset. Another idea is to simplify review writing by offering the option to select common descriptions that users can leverage to clearly express their opinions, rather than manually typing the entire review. These short descriptions would improve the readability of reviews and provide more information for classifying the assets more appropriately. Such practice, along with the endorsement option, appears to be working well for bol.com, an online retailer from Netherlands:
Finding the most effective ways to properly engage users around information management, and ultimately increase return on data investments, will be a journey.
Thanks to our adoption product team, this feature has been brought into existence and we look forward to more engagement features from them! How would you want to rate your data? We’d love to hear your ideas – leave a comment below or tweet your answer @Collibra.