The Future Analytics: Part 4 – Suite and Actions
This is part four in a four-part series. You can find the other blogs in the series here:
The Future Analytics: Part 1 – Overview
The Future Analytics: Part 2 – Big Data, Predictive and HANA
The Future Analytics: Part 3 – Apps and Visualization
Watch this video to learn more about the Future Analytics: The Future of Analytics & Big Data (sapserviceshub.com)
In earlier parts we’ve talked about the combination of Big Data, Predictive Analytics, HANA, and how analysis through this is visualized and used by users through Apps. In this part, we’ll talk about what you can do with that beyond analytic use cases and actually change the way your business operates. Already, 70-80% of Fortune 500 companies run on SAP’s Business Suite for their business operations. What we can do now is further enrich what is inside the Business Suite with the other elements we’ve discussed earlier, and find additional efficiencies and business opportunities.
To see what impact this would have, I’d like to work through an example which I believe is transferrable to lots of other business use cases, and illustrates how this can change business operations. Let’s assume that we manufacture washing machines. While the manufacturer might themselves create the casings and other moving parts, there are microcontrollers and other electrical components to run the washing machine and allows for different washing cycles and programming options, as well as sensors to correctly handle water temperatures, etc. These components we source from component manufacturers. Washing machines and other durable “white” goods can be a significant investment for a customer (and would be in use for many years) and can easily cost in the range of 500-1,000 Dollars or Euros. Competing on price, therefore, is an important component, especially where retailers might match prices for customers between competitors. Finding ways, therefore, to reduce the price could have a significant impact on the volume of sales you make.
So, let’s imagine we have the choice of multiple component manufacturers that submit us with electronic bids on any components we need. What we need from our suppliers is driven by the component inventory in our manufacturing plants. By purchasing those components centrally, using predictive analytics, we can evaluate which bid best matches our need, based on price and volume and how quickly the components can get to our manufacturing plants. If we need 10,000 microcontrollers, a bid of cheap components but in too high or too low a volume makes no sense and would either mean we have too many components in inventory, or we need to get another batch from elsewhere at potentially a far less compelling price. Moreover, if we can get a cheap deal, but the delivery takes weeks leaving our plants without components, that wouldn’t be optimal either.
If this works well, though, we should be able to source our components as competitively as possible while keeping the production line running, and manufacture our washing machines at a lower cost price.
We can now offer our model at a cheaper cost price to retailers who can now provide discounts to their customers without hitting their margins, and whoever is first, or is willing to offer a better price for them, we ship our new washing machines.
But this only goes one direction. Washing machines break down, are on warranty, etc. so there is a particular failure rate. The customer returns to the store where they bought the machine from, and this provides us with information what exactly the problem was. We can analyze this over time, and figure out if there are patterns in this. Maybe some of it is a design flaw we are responsible for ourselves and we can correct, but in other cases it may be a faulty microcontroller or other components we sourced from our component suppliers. If we find that components from a particular supplier have a higher failure rate than others that is not acceptable, we can drop that supplier from our preferred supplier list and no longer evaluate their bids. That is, the actual performance of the sourced components will have consequences in the supply chain.
This creates a feedback loop of increasing efficiency. The inventory in the manufacturing plants dictates the bid that goes out. The optimization of component sourcing drives the cost price we offer to the retailers and the volume shipped there again drives our inventory and component needs. Any returns and failures are analyzed to ensure we source quality components from suppliers we can trust, thereby also reducing the frequency with which such failures occur. We win, because we sell – over time – more washing machines at lower cost that are more reliable; the retailer wins because they can offer the same washing machines at lower price and get more people in their stores, whereas the consumer wins by ending up with a higher quality washing machine that they can afford. Suppliers win, as long as their components are high quality, can ship quickly, and at an attractive price.
This brings us to recommended and eventually autonomous actions. To evaluate the bids, we would use a predictive model that scores the bid against what we need. This could then rank the bids and offer that up to a central purchaser to make the final decision. That would likely mean that in many cases the top recommended bid would be chosen, but there may be reasons not captured by the model (initially) to divert from that from time to time. However, we could use that override again as input into the model to make the recommendations better if we can identify the reason why that bid was preferable. Once we find that over the course of time such overrides occurs less and less (or preferably, not at all), we could consider turning it from a recommended action that requires approval to an autonomous action that chooses the bids automatically, as long as a clear ‘bid winner’ can be identified, and only request approval by the purchaser in case two or more bids are scored very similarly. Another predictive model could evaluate component suppliers on the basis of the failure reports, initially ‘supervised’ by someone reviewing the recommendations, but over time more and more autonomously.
All of these things, we can do today. We already have over 1,200 Business Suite on HANA customers. (See Business Suite on HANA on saphana.com for much more information on this) We have the predictive capabilities and integration that come with the HANA appliance. We can pull in non-traditional data through Smart Data Access from Hadoop or SAP IQ. And we have the SAPUI5 web application framework to build apps that work on desktop, tablet and phone and allow us to visualize any analysis and results. The Future Analytics, therefore, are already here. Let’s make it happen.
To learn more about how SAP HANA Services can help you throughout your Analytics journey, please visit us online.