Driving Analytical Insights

Summary: Generating credible insights is more than crunching numbers and creating attractive visualisations. Ineffective insights, arguably, cause more harm than no insights.

Case in point: ‘My Electricity Provider’ shared this insight on my energy use.

 

 

At a glance, these insights appear, well, ‘insightful’ coming from a smart meter, till you dig a bit deeper. The insight is for a property address that is still under construction and is currently at the frame stage!

 

It means there’s no hot water system, cooking is incorrectly assumed to be electric and there’s no lighting installed. The electricity usage is, most likely, from the workers on-site!

 

Laundry and dishwasher are out of the question, as there’s no one living there, and the plumbing has not been established.

 

Further, how were these insights derived: ‘Energy Insights uses information you provide, data from your smart meter, weather metrics and other data sources to estimate where electricity is being used across your home.’

 

Difference between Data, Information, Insight and Analytics.
Before we delve deeper into insights, let’s establish a fundamental difference between data, information, insights and analytics.

 

 

In simple terms, data is a collection of facts. Data can be raw as in the original form as collected, and data can be cleansed or augmented, where data is standardised and new data generated based on existing data. In these scenarios, the original data does not change. Data, in its pure form, lacks meaningful structure or context.

 

When data is processed, organised, interpreted and structured it becomes information. Information is, essentially, data made valuable and accessible. Information is refined structured output providing a business context.

 

Where information helps observe what is happening, insights tell us why. Insights delve deeper into data and connect dots that may seem unrelated. Insights provide a clearer understanding of the underlying causes and implications of information, often leading to more informed decisions and strategies.

 

Analytics is the discovery of patterns and trends from the data and therefore Analytical Insights helps to test new variations in business rules and generate better insights. Thus, analytics is the systematic computational analysis of data or statistics.

 

What are Analytical Insights?

Analytical Insights are a fusion of analytics and insights, where complex data sets are examined to uncover patterns, trends and relationships, and these findings are then interpreted to provide a deeper understanding and practical guidance.

 

The insights from ‘My Electricity Provider’ are a combination of statistical data and trends. Even with the best intentions, the insights have missed the mark. So, what are the salient features driving analytical insights?

 

Any analytical insights need to address the following common concerns from users:

 

• Is there any data missing from the underlying data set?
• Is the data accurate?
• Is data consistent?
• How recently has the data been updated?
• Does the data comply with the company’s formatting and other standards?

 

These queries can be classified across the following data quality categories:

 

 

Contextualisation:

Contextualisation involves adding related information to make data more useful. Data contextualisation allows understanding information within its relevant context, extending the boundaries of conventional data analysis. Deeper relationships and connections that underpin data can extract further value for business.

 

In the example of ‘My Electricity Provider’, a key contextual information missing was the house is under construction and so the standard background information, patterns and trends needed re-evaluation. House under construction is an outlier and therefore the trends and patterns need re-adjustment.

 

Data accuracy:
In simple terms – data is accurate if the data describes the real world. Data accuracy is the fundamental attribute of data quality and data integrity. Key characteristics of data accuracy are:

• Correctness: data is factually correct and free of ‘bad data’. Either the bad data is corrected or excluded.
• Completeness: provides a full data set without any omission.
• Consistency: data is uniform across data sets over time.
• Validity: data conforms to correct formats, ranges and data types.
• Precision: data is exact without unnecessary approximations.

 

Actionable:
Actionable data presents in a way that can be easily leveraged to drive business decisions. This means the data needs to be in proper context, accurate and visualised in a way for easy decision making. This is where BI tools have a role to play. Any good BI tool turns raw, unfiltered data into actionable data that drives insights. Actionable data can be used to draw clear implications.

 

Granularity:
Granularity in data refers to the level of detail of the data. Data with lower granularity is summary data that has been ‘rolled-up’ or summarised using certain criteria. Using data with lower granularity may lead to approximations and imprecise analysis and trends.

 

High granularity, also called fine-grained data, has greater details and provides for more precise and detailed analysis. It also gives the capability to undertake trends and analysis using different pivot points.

 

Interestingly, right granularity is directed to different uses. Strategic information can be derived from higher grained or lower granularity data. Finer grained data with higher granularity is better suited for operational analysis and trends.

 

Relevance:
Data should be relevant to the insights being generated. It should indisputably include all dimensions – positive and negative. Only relevant data can transcend from great hypothesis to strong strategy. A strategy based on relevant data will provide the metrics to track and measure, providing a foundation to respond to changes.

 

Typical challenges for data relevance are:

 

• Data volumes – large data volumes will require data analysis to identify relevant data sets.
• Data silos – isolated data systems will require further data integration work prior to any analysis
• Dynamic environment – rapid changes to business environment and focus will challenge the concept of what data is relevant. This will require continuous update and re-establishment of business rules.

 

Timeliness:
Data that is outdated, delayed, or inconsistent can lead to poor decisions, wasted resources, missed opportunities and potential reputational damage.

 

Another term for data timeliness is data currency – how current is the data. It refers to the potential lag between the time data is recorded to when the data is considered for action. Timeliness or data currency is important as it adds value to information that is particularly time sensitive. In addition, data quality may diminish over time or data granularity may change over time.

 

Conclusion:
Tying all these critical characteristics of data quality in view of the insights from ‘My Electricity Provider’, it’s clear the insights algorithm has failed on multiple fronts. There’s missing contextualisation and accuracy. Since there are no appliances yet connected, the algorithm uses assumed granularity and relevance. All this is then pulled together in a data set that is clearly not actionable due to inherent flaws.

 

Even though ‘My Electricity Provider’ tries to impress their customers, their insights are causing reputational damage rather than growth.

Sachin is a seasoned data advisory and delivery professional with two decades of experience at the forefront of data innovation. Over the course of 20 years, Sachin has been instrumental in orchestrating and delivering a wide array of data projects, encompassing data warehouses, insightful analytics and transformative data solutions. With demonstrated expertise in data strategy and implementation, Sachin has consistently demonstrated his thought leadership for turning complex data challenges into streamlined, actionable insights.

 

See Sachin’s profile here.

Get In Touch