Should building decision intelligence skills be a priority in realising the potential from data?

Most organisations have recognised the transformational potential of advanced analytics to improve outcomes – especially through the smarter use of data to make better and faster decisions.

To progress these goals, much has been made of the importance and need for organisations to invest in AI capabilities and cloud technologies. Yet repeated research studies show the key challenges to realising the potential of data are more cultural and organisational than technological. So what can data teams and business users do differently to help overcome these cultural and organisational barriers?

Taking the business on the journey – from end to end

Great analytics teams should already recognise the importance of taking business stakeholders on the journey – involving them in discovering the need, thin slice solution development and rapid experimentation, through to solving the capability process and technology challenges to embed and scale the solution. As well as this business involvement being key to testing and ensuring the data solution is fit for purpose, it is also invaluable in building business trust and familiarity with the solution. Rather than analytics teams just focusing on improving the explainability of models, a recent Harvard study [1] indicates that involving recognised business SMEs is a much more effective way to build senior stakeholder trust and so engagement.

Being empathetic to how data can challenge stakeholders’ expertise, judgement and so sense of self.

Organisational leaders have reached positions of seniority through years of successfully deploying their judgement and experience. Yet there is the risk that the perspectives and recommendations resulting from a new data solution may challenge deeply held beliefs or years of hard earnt experience. Whilst we would hope that leaders are enlightened enough to willingly embrace the challenge, we shouldn't take it as a given. Analytics teams need to demonstrate empathy and humility in how they engage stakeholders in the development of the solution. For example creating a safe environment where stakeholders can test this new perspective against their experiences and so evolve their views and judgement. This step is also super-useful for analytics teams to triangulate the solution against these experience points to ensure there are no over-looked nuances, unforeseen errors or negative implications.

Recognising that data is only one factor in how humans make decisions

Typically, analytics teams by nature are logic driven – if the evidence supports a certain conclusion or outcome, then they think it should be obvious for the business to just adopt and embrace. Yet this overlooks the emotional and judgement dimensions that are at least equally important in how stakeholders makes decisions.

As the illustration[2] shows, for a long time, casinos have been experts in using (and exploiting!) an understanding of how humans make decisions.

If data teams have a better understanding of the thinking and motivations behind how stakeholders make decisions, they can then be more nuanced in how to evidence, position and communicate the approach and so be more effective in engaging users in its adoption.

This understanding of the thought patterns behind decision making can also extend to helping to improve the adoption or even compliance to new initiatives or processes. In the field of cyber security, the risks and so need for compliance should be obvious. Yet the illustration below from The Alan Turing Institute[3] demonstrates how behavioural science provides valuable insight as to why as to individuals take different approaches towards cyber-security. This insight can then be used to shape the user engagement and compliance approach for each cluster.

Moving forward: Building decision intelligence awareness and capabilities

The capability to understand how humans make decisions is rarely considered in the engagement of stakeholders in data/ AI transformations. I am not advocating businesses should add yet another role to their data teams as per Instagram's marketing decision scientists[4] or Google's Chief Decision Scientist.

Instead, firstly we need clear acknowledgement that the best outcome is not about being purely data driven or just using judgement & experience, but the best of both. For a given context or problem scenario, this should evolve to consciously framing what is the right balance between the use of judgement and data - with analytics teams being transparent as to when there is insufficient data or model accuracy to be confidently relied upon.

A next step is to build awareness and make the time for conversations that help surface and reflect the different thought patterns and perspectives that shape decision making. For example:

-??How has judgement and experience historically been used in making a decision, and how can this be explored and tested through data and experiments as part of developing the new data solution?

-??Where might a data decision solution fundamentally challenge the perceived experience, or judgement and so how to position and explore this with stakeholder in a trusted manner?

-??Where seeking to influence decisions, what might be the human decision traps that need to be navigated and so explored through discussions and evidence?

A third step would be to expand data literacy programs for technical and business users to grow awareness and skills in how human make decisions, build trust based influencing skills, alongside how to frame and manage the benefits and limitations in the use of data in decisions.

Without a clearer focus on building the capability to harness the potential of combining judgement, experience and data in making smarter, faster decisions, the risk is that the sophistication of advanced analytics and AI continues to be beyond organisations' capability to absorb and embed.

[1] https://hbswk.hbs.edu/item/what-makes-employees-trust-vs-second-guess-ai

[2] https://www.datasciencecentral.com/data-literacy-education-framework-part-2/

[3]https://youtube.com/clip/UgkxTSmyHrr_YpmNixdFJ-neEK_m0ppspNyc

[4] https://towardsdatascience.com/data-science-vs-decision-science-8f8d53ce25da