My last post (The New KPI’s of Retail) asserted that the KPI’s upon which retailers run their businesses must be modernized. This post broadens the position to retail analytics in general. The potential analytical value of big data has been booming, and related technology has evolved dramatically in recent years, so why are so many retailers so far behind in exploiting this?
The answer is simple – retail analytics has historically been both difficult and expensive, and for the most part still is. To the extent that hundreds of millions of historical sales and inventory transactions are housed, and that artificial intelligence and machine learning are applied, and that unstructured ‘big’ data is incorporated, the potential cost and complexity can simply be prohibitive.
Truly modern retail analytics has indeed been an elusive goal for most. With that in mind, consider the following top ten requirements of modern retail analytics, followed by my recommendation for resolving all of these within a course of just a few months.
Modern retail analytics must be:
#10 - Comprehensive. Both raw and aggregated data, along with the metrics and analyses that leverage that data, must go wide and deep, across the breadth of retail subject areas, with conformed dimensions to ensure consistent views across items, customers, suppliers, etc.
#9 - Open. No matter how comprehensive your solution is per the graphic above, other “reservoirs” of data with potential analytical value are sure to exist. These could be spreadsheets on someone’s hard drive, flat file extracts from a third-party, or a legacy database in your data center. Your solution should be able to easily and quickly mash this data up for additional insights.
#8 - Integrated. From storage-to-scorecard (i.e. from cloud infrastructure to network to hardware to operating system to database to middleware to software) all components should be seamlessly integrated, and optimized to scale. After all, with analytics, performance is king!
#7 - Secure. Data is gold, so needs to be locked-up, and then locked again and again. Software security assurance processes, data encryption both in transit and at rest, advanced authentication, regular third-party security audits, PII/PCI/ISO/GDPR certifications. It also needs to be highly-available, with at least 99.5% system availability and full redundancy should there be a catastrophic failure of one instance.
#6 - Trusted. Analytics is a garbage-in, garbage-out concern. So don’t let the garbage in. This goes beyond data quality – data and business semantic governance is crucial as well. After all, even the highest quality of data and the most insightful analytics are worthless if users across LOB’s don’t have an accurate and consistent understanding of their meaning.
#5 - Timely. Data pertaining to sales transactions, inventory movements, social media interaction, etc. needs to be consolidated for analytics as rapidly as is necessary to support related decisions. In some cases, next day should suffice. In other cases, real-time is warranted.
#4 - Scalable. Your users will expect at least the most recent 2-3 years of historical data to be readily available to analyze. And their queries, well, they need to be reasonable, but they need to be fast, and handle peak volumes like those often experienced on Monday mornings during the holiday season.
#3 - Scientific. Artificial intelligence and machine learning are no longer “special” – they’re table stakes now for any modern analytics solution. And data science should not occur on its own data silo – the same data warehouse that supports descriptive analytics should also support predictive and prescriptive analytics, and the user experiences of all three types of analytics should be federated.
#2 - Economical. The total cost of ownership of a bespoke solution, especially one that cobbles together components from multiple vendors, can take years to build and return meaningful value to your users, and can require an unacceptable investment in support and administration. Cloud services can be the most cost-effective and the most expeditious path to modern retail analytics, and further economies are enjoyed to the extent that one solution can address all of these requirements.
And, drum-roll please...
#1 - Adaptive. Adaptive intelligence is where artificial intelligence and human judgement intersect. To that end, insights need to be surfaced, properly-filtered, to whomever, however, wherever and whenever desired, and the insights-to-action loop must be seamlessly applied within a framework that fosters continual improvement, through user-assisted machine learning.
Oracle is in the very unique position to be able to satisfy all of these requirements with one cloud service subscription.
Oracle Retail Insights Cloud Service Suite enables adaptive intelligence for the retail enterprise, and includes:
It is pre-integrated with key Oracle Retail applications like Oracle Retail Merchandising, Planning and Customer Engagement, and can be fully-deployed within just a few months (two points that, among others, tempted me to expand the list above to be a ‘Top 20 List’).
With Oracle Retail Insights, there are no longer any excuses to not fully exploit the analytical value of your big data, and inform your decisions and innovations with modern retail analytics.