Data collection is no longer an issue thanks to modern advances in the tech stack. However, making use of this data and translating it into action is a big challenge that a lot of organisations are still grappling with. While massive effort is spent in trying to translate this huge amount of data into insight, there is often little capacity left to further convert these insights into tangible operational action.
Kshira Saagar, Head of Data Sciences and Analytics at THE ICONIC presented at CX Innovation & Tech Fest this year. During the event he shared some of his insights into how this data-action loop can be closed, allowing operational action to follow from data analysis.
Generating Action, Not Just Insights
In the Data > Insights > Knowledge > Actions pyramid, where companies start at the bottom collecting data as a foundation and work their way to top of the pyramid to translate that data into action, many organisations just move from data to insights and stop there.
There are three common challenges that stop insights from becoming actionable:
- Analysis paralysis: being paralysed by over analysing insights and not making a decision in time for those insights to become actionable
- Broken relationship between systems: most often the analytical systems that aid in decision making don’t communicate properly or at all with the consumer-facing systems, leading to a huge gap between decision making and actual actions
- Lack of an appropriate tiered action plan: not all actions are the same and not all decisions from the data should be actioned in the same way
Realising, addressing and overcoming these challenges will allow companies to move forward with decision making and ultimately enable them to see tangible results. Below are some of the mindset shifts and technology considerations that companies should tackle so that they can move into a better “insights to action” state.
Creating a Culture of Discovery-Driven Problem Solving
Today’s problems and business challenges are uncertain from the start, thanks to the ever increasing volumes of data pouring in. The assumptions one makes at the outset of decision making initiatives aren’t likely to hold up as new information and new data sources emerge, requiring substantial pivots along the way.
And so, in some cases, the past methods of linear problem solving – operating on the premise that managers can extrapolate future results from well-understood and predictable platforms of past experience – needs to be replaced by a more agile method, requiring frequent reassessment and adaptation of plans.
Discovery-driven planning and problem solving offers a lower risk way to move a muddled problem forward towards solution in the face of factors that are unknown, uncertain, and not yet obvious. Applying this agile approach allows companies to learn as much as possible, as cheaply as possible, while pursuing new ventures.
In today’s agile and ever changing information landscape it becomes important to create a culture that accepts this new approach to problem solving and decision making. Some of the key principles of successful discovery-driven problem solving include:
- Choosing the right problems to solve using this approach
- Defining clearly what success looks like
- Benchmarking that success definition against common sense and market information
- Setting the limit where failure would be declared to prevent unprecedented losses
- Clearly communicating the operational requirements (tools/techniques) to solve the problem
Enhancing Customer Experience in the Age of Instant Gratification
In the day of instant gratification driven businesses like Uber, Spotify and Whatsapp, customers have moved into a “I want what I want, when I want and where I want it” mentality. This means that companies need to switch to a new way of operating – not only hyper-personalising marketing of individual offerings to the customers at the right place and the right time, but also personalising individual experiences and moments of truth along the whole customer journey in a way that makes as much sense and causes as little friction as possible for a customer.
This can be facilitated using an Always Learning Training Instantly (ALTI) framework, supported by real-time streaming platforms and event-triggered personalised customer experience offerings. Not only do descriptive dashboards have to become real-time and actionable, but personalisation algorithms on the backend also need to get more real-time and as expressly relevant as possible. This can be enabled by upgrading the data pipelines and processes to reflect this new paradigm. It is also necessary to spend time and effort on developing event-driven algorithms to inform personalising decisions on all crucial customer touch points – from marketing to final product delivery.
Always Learning Training Instantly (ALTI) is a framework or an approach to having real-time models listening to data and customer interactions continuously and improving the models and algorithms continuously from what they have learnt. This contrasts with the current approach in many machine learning applications where models are built once using old data and the same model is used for a year or more without much change – an approach that just won’t work in a dynamic setting like the online retail world for example.
Structuring the Unstructured
One of the biggest stumbling blocks preventing data from becoming tangible actions is “paralysis by analysis” – and a major component of this block is the inability to work with the onslaught of data, leading to death by drowning in the data ocean. A lot of new data sources come in varied shapes, sizes and formats, and attempting to fit all of them into a common standard and format would be a big mistake.
Instead, the fundamental principle should be to look at unstructured data using the discovery driven approach in its natural habitat. There are a variety of programming languages and constructs beyond SQL that enable data scientists to explore, analyse and identify features in unstructured data without having to reformat them.
A technical example: analysing millions of faces on live streaming videos real-time isn’t possible by translating them into SQL tables and then running SQL queries on top of them. Instead, writing a deep-learning CNN or GAN code on top of the streaming video pipeline will result in massive benefits real time and help solve problems that are beyond the human grasp.
For more insights from the boldest CX innovators, join us at Customer Experience Innovation & Tech Fest.
About the Author
Kshira Saagar has been in the analytics and decision sciences industry for almost a decade, having worked across Americas, Asia, Europe and Australia. In his current role at The Iconic, as the Head of Analytics and Data Sciences, he’s responsible for understanding and enabling data driven decision making. Previously at Datalicious and prior to that at Fairfax, he was responsible for institutionalising data-driven analytics across the company’s core competencies and building new-age analytical products for the organisation.