SAP Analytics Cloud #2: SAP Predictive Analytics vs. Analytics Cloud

13 July 2017

Peter Douglas

Peter Douglas

Senior Consultant

In a head to head predictive analytics demo battle, which tool gives the non-expert user the insight they need as quickly and as accurately as possible?
 

SAP Predictive Analytics vs. SAP Analytics Cloud – demo

Last year I wrote an article about generating marketing insight from data using “SAP Predictive Analytics” (SPA). I built a small demo that used a custom R script and integrated with SPA to generate automated insight. The demo was bespoke and limited to a specific use case. Recently I’ve been (pleasantly) surprised to discover that a similar capability to generate automated business insight is exactly what “SAP Analytics Cloud” (SAC) does out of the box!  This is exciting stuff: it heralds the beginning of putting ‘predictive’ into the hands of the business user.
 
To recap on the data set, it contains time-series data on website visit traffic for each day of a year across 7 brands, alongside data on the number of TV adverts for “Brand A” on each day. Visits to brand “microsites” might give some clues as to the behaviour of consumer demand for the products for each brand, particularly if the microsites are e-commerce enabled. The TV advertising may be a factor in driving demand, not just for that being advertised but maybe also for other brands.
 
I loaded the same data into SAC and selected “Smart Discovery”. The results are shown below.


Automated insight in Smart Discovery 

Pete-D-1-SAP-Analytics-Cloud-SAP-Predictive-Analytics-vs-Analytics-Cloud.png
 
As with my previous demo in SPA, Smart Discovery has picked up the fact that adverts for “Brand A” are having a complementary effect on website traffic to “Brand B” which is perhaps contrary to what one would expect. This could be since brands “A” and “B” are quite similar and interest in “A” may have peaked, but “B” may be new, with a growing market share.
 
To simply provide a measure for how far a user can trust this insight, Smart Discovery provides an “Insight Quality” value out of 5. In this case the analysis has been scored 3/5. Anything lower than “3/5” prompts Smart Discovery to automatically prevent the user from using the simulation tool described in the next section since it would lead to inaccurate conclusions.


Simulation using Smart Discovery results 

We can simulate and optimise the amount of traffic to the “Brand B” website and can see how each influencing factor contributes to the optimised value, as shown in the “Waterfall” chart below. This is a feature that is not available in this form in SPA, although it could be built through another R extension, so we add to the score for SAC vs. SPA: 1 - 0.
 
Pete-D-2-SAP-Analytics-Cloud-SAP-Predictive-Analytics-vs-Analytics-Cloud.png
 
One observation I would make about Smart Discovery is that a more technical user does not have access to the detailed parameters of the statistical model working behind the scenes, which I have read is delivered using the “Key Influencers” service available in SAP Cloud Platform.
 
When comparing the scope of predictive functionality available to a more expert user in SAC it doesn’t have the comprehensive library of predictive algorithms available in SPA, since SAC is a tool that keeps the barrier low for business users to leverage predictive tools. I think therefore it’s a point scored for SPA (now 1 – 1!), as this is something that would appeal to data scientists if SAC was closer to SPA in this respect.


Applying Smart Discovery insights for better predictive forecasting 

Finally, we can use the results of the Smart Discovery to inform a time-series forecast for website traffic to “Brand B” by including the future plan for adverts for “Brand A” at the start of the following year. This chart was produced using the time-series forecasting functionality, described in “Part 1” of this article, and applying an “Advanced Forecast” to include the complementary effect of advertising for “Brand A”.
 
SPA also has the capability to generate time-series forecasts, although the ability to add additional input variables, e.g. TV advertising, to improve the forecast accuracy is not available “out of the box” (can be done using “ARIMAX” algorithm in HANA PAL or R extension, but that’s another post for another day!). So, once again we add to the score for SAC vs. SPA: 2 – 1.
 

 
What all of this means is that a Marketing Manager using SAC can predict consumer demand for “Brand B” in the New Year and around Easter, so that further marketing activities can be planned. They can also inform Demand Planners of the predicted uplift to ensure that enough stock is in place to meet this demand. The forecast and other supporting information can be included in the “Digital Boardroom” for Management level consumption.


Final thoughts

What are the final scores? It’s a win for SAC vs. SPA at 2 - 1!
 
SAP is on a journey with the Analytics Cloud tool, and there is still some way to go to improve certain aspects of its functionality, such as data shaping and blending, which is more mature in other SAP tools like Lumira and competitor tools like Tableau. This “extract, transform, load” capability is better suited to other platforms, for example, BW/4 HANA, which fully integrates with SAC.
 
SAC is maturing and SAP have made solid advances in its functionality since it was released a year and a half ago. I look forward to seeing more predictive and advanced analytics capabilities added to its tool set in the near future, which will help break down the barriers to any business user applying sophisticated analytics to their own business challenges.


POSTPONED: SAP Analytics Cloud: A name changer or a game changer?


 

About the author

Peter Douglas

Senior Consultant

Focused on helping businesses gain valuable insight from their enterprise data through the delivery of solutions in predictive & advanced analytics.

Since arriving at Bluefin, Peter has had the pleasure of being part of challenging and rewarding projects across the Consumer Business, Manufacturing and Pharmaceuticals industries, for some well-known global names. His primary focus is working with clients to understand and exploit the power of emerging capabilities in advanced analytics and apply this to their business needs. His skills and hands-on experience span a range of SAP technologies, such as HANA / BW, Predictive TPM and CRM.

Peter’s passion for all things “Data Science” manifests itself in championing the predictive & advanced analytics ecosystem of SAP tools as a great opportunity for clients to tap into the power of their data to answer targeted business questions with accuracy and insight.

His previous life as a Quantum Physicist and PhD Grad has left Peter with old habits: in everything he does he loves testing theories and experimenting with new ideas and technology. He’s also addicted to continually learning new things.

Outside of work Peter enjoys life in London and, when he can, explores the World.