Cindi Howson, Chief Data Strategy Officer, ThoughtSpot

Cindi Howson is the Chief Data Strategy Officer at ThoughtSpot and host of The Data Chief podcast. Cindi is an analytics and BI thought leader and expert with a flair for bridging business needs with technology.  As Chief Data Strategy Officer at ThoughtSpot, she advises top clients on data strategy and best practices to become data-driven, speaks internationally on top trends such as AI ethics, and influences ThoughtSpot’s product strategy. Cindi was previously a Gartner research Vice President, as the lead author for the data and analytics maturity model and analytics and BI Magic Quadrant, and a popular keynote speaker.  She introduced new research in data and AI for good, NLP/BI Search, and augmented analytics and brought both the BI bake offs and innovation panels to Gartner globally.  Prior to joining Gartner, she was founder of BI Scorecard, a resource for in-depth product reviews based on exclusive hands-on testing, contributor to Information Week, and the author of several books including: Successful Business Intelligence: Unlock the Value of BI & Big Data, Analytics Interpreted, and SAP BusinessObjects BI 4.0: The Complete Reference.



The economy continues to falter and yet the one consistent message from CEOs in spite of this is continued investments in analytics and AI. In a digital economy, leveraging data is the difference between those who have transformed and those who have simply automated processes. But leveraging data for analytics is only a first step.  To derive the full value of data – whether to improve working capital or reduce stock outs or personalize customer interactions – we need to close the loop from data collection to generating insights to acting on those insights.

Some call this insight to action “decision intelligence,” but I am wary of some vendors who have been “decision intelligence washing” their products without having a closed loop solution.  Decision intelligence as a concept was once the realm of rules-based models: Approve a loan if the credit score is above 700, decline it if the score is below 700. Interview the candidate if they went to an Ivy League school, reject if there is a career gap.

With more data and modern cloud ecosystems, decision intelligence can now be smarter, fueled by the best combination of human insight and technical advances.

The combination is critical. Blackbox AI that lacks transparency risks bias at scale. There are too many decisions where an exception to the rule may yield a better result. These decisions require a human in the loop. Should Apple Pay really have given a wife a higher credit rating than her husband? Would better insights have addressed this biased AI and prevented a PR blunder?

As organizations rush to leverage generative AI in all its forms – ChatGPT being just one – lack of trust is already causing some organizations to ban its use outright.  ChatGPT gives authoritative sounding answers that are utterly incorrect to the point that some AI experts dub these “hallucinations.”  The other challenge facing ChatGPT is that OpenAI has yet to reveal on which public data sources the large language model (LLM) has been trained, further eroding trust. Trust in the data and the insights generated is critical for anyone being able to act on the information. In the absence of this trust, people will revert to what they know and what they trust more: intuition and inflexible rules.

As we think of the evolution of technology to enable insight to action, the cloud ecosystem from business applications to data storage to insight engines are facilitating this end-to-end workflow.

For example, bringing analytics to bear on decision intelligence, one might run a query to ask, “How do credit scores between male and female applicants compare when salaries are the same?” A following query could be, “Give me a list of those applicants whose direct deposits and on-time payments have been consistent for the last 3 years.” With the results of this query, the list of customers to flag in the system for deeper review could be automated by reverse ETL. With reverse ETL, insight-related data is taken out of the cloud data platform and exported to a spreadsheet or written back to the operational business application. Alternatively, the analytics platform relies on an open API to communicate from the analytics platform to the operational application. The shift of work to cloud-based business applications with open APIs enable a more closed-loop insight to action process.

Business Applicationson-premisescloud-based
ETLextract only, source to targetsource to target, back to source
APIsclosed, proprietaryopen
Modelsrule-basedexplainable and self-learning

Even with the most modern technology and valuable insights, if you do not have the right culture of trust and transparency, data will not be acted upon. Further, when data is used to punish people as a first response, versus to learn and improve, then analytics will be manipulated into vanity metrics.  This is where organizations must look at the formal and informal incentives of democratizing insights. Picture the sales leader whose team is underperforming due to things beyond their control (supply chain issues, failed marketing campaigns, etc.): do they get ridiculed or immediately fired? If so, data will be suppressed. A positive, data-driven culture will use the negative KPIs to inform improvements and corrective actions. A Gartner survey found 67% say people “cherry pick” data to tell an alternate version of a story. Countries, as well as organizations manipulate data when the incentives are strong enough whether its dictators to exaggerate  economic growth, Russia to undercount war casualties, or China to undercount Covid cases. ESG and DEI reporting are notorious examples of company reported vanity metrics to the point that ESG manipulation has its own term – greenwashing.

As you strive to move your analytics from backward looking descriptive analytics to full prescriptive, insight to action, follow these best practices:

  • Evaluate analytics workflows that include subsequent manual processes for automation opportunities. Look specifically for manual exports to spreadsheets that then involve re-entering data in cloud-based business applications.
  • Identify high volume decisions that could be augmented with AI. Include diverse stakeholders earlier in the design processes to minimize the risk of unintentional biases.
  • Recognize the degree that culture, poor data fluency, and incentives interfere with the desired action being taken.

For more top trends and best practices see Top 9 trends in data and analytics for 2023.

Content Disclaimer

Related Articles