Big Data scenarios are now much easier to implement
Big Data is becoming smaller. And smaller. Not in volume – but in the difficulty of handling. This makes it fairly easy to run scenarios which two years ago seemed incredibly difficult to implement. These scenarios of course come with significant business benefits. However, they are often hyped, along with the technology you need to implement these scenarios.
Often, vendors of the these new-age technologies such as artificial intelligence, machine learning, pattern recognition, and predictive miracles will show you use cases which sound great, but then comment them with two or three sentences, making them seem common-place. If such uses cases are commonplace, then why gloss over them? Also, many SAP experts keep asking themselves “what’s my role in this?”. How does structured SAP business data fit into this context? And – more importantly – “what’s in it for me?”. How can an SAP expert gain benefits from this new world of Big Data?
Three Use Cases With a Big Impact On Your Business
Let’s take a step back and consider real life, out of this world use cases which you can implement easily using your SAP data.
For this, I will consider the three top areas which McKinsey identified in terms of business impact: Predictive Maintenance, Customer Service Management, and Next Product to Buy. In these simple use cases, I will assume you have the following available:
1. An SAP system (obviously)
2. The Big Data platform of your choice
3. An integration solution such as Datavard Glue
Datavard Glue is a software-only solution, running directly in your SAP system. As you will see with these use cases, you can easily and inexpensively run it and use it for both rapid prototyping as well as for production implementations in record time.
A Big Data platform comes in handy when implementing Big Data. Of course, for prototyping a simple Python installation may be all you need. The picture below illustrates our reference architecture for such an implementation:
For prototyping you will not need all of these components of course. I’d like to show you how using a simple data extraction you can start with SAP data (e.g. provided by Glue or a simple download) and (for example) Python. For the productization of any scenario, a standardized technology with flow control, lineage, authorizations, SAP integration etc. is of course vital – I would recommend this to be the latest stage where Glue comes into the picture.
The story of the use case of Big Data for Predictive Maintenance is so old by now it’s growing a beard. However, it is still extremely relevant and valid. Not only does Predictive Maintenance help to detect faulty parts before they break, but also it can help to detect parts before they become faulty and cause harm before actually physically breaking! Just imagine a production line in the pharmaceutical industry where a critical part to measure quantities imprecisely, and causes whole batches of medicine go bad at random intervals.
From an SAP perspective, there are three things to be done:
- You need to make SAP equipment data available to your Big Data platform. This is done in the blink of an eye using Datavard Glue – either from scratch or by simply using one of the out of the box SAP PM (Plant Maintenance) data scenarios in Glue
- You need sensor data from your machines – our extensive maintenance history from SAP. Ideally you have both, and can combine them. Taking sensor data and some readily available open source libraries for pattern recognition may take you quite far. In Python, you will find that a script to chew through the data may have less than 100 lines of code. Heck, if done by an experienced Python developer, the code may be some 20 lines of code only, even if it needs some twisting and tweaking to get all parameters right initially.
- You will want to consume the results of this in SAP. You may want to integrate this into a dashboard for reporting purposes, or even better, directly generate PM maintenance orders in SAP from the results of your pattern recognition.
Using Glue, you can pick up the result data and generate SAP PM orders easily.
Customer Service Management
No area may be as big as Customer Service, and its IT landscapes so heterogeneous as in this business area. Between CRM, call centers, classical customer service, and home-grown applications you will find all kinds of applications. From an SAP perspective there is the new area of “X”-data (referring to experience data) to complement operational data. However, let’s again take a step back and look at this from a general SAP professional’s point of view. How about figuring out customer relevance as a simple KPI to see where premium service is actually worth the additional effort.
Customer Lifetime Value (CLV) is a useful KPI in finance. Why not correlate this KPI with the effort you pour into customer service? This KPI can be calculated based on SAP finance data. The relevant SAP data can be easily provided from SAP’s accounts receivable and general ledgers, and SAP master data. Using Datavard Glue, identifying relevant data and making it available on your Big Data platform can be a matter of minutes. Using this SAP FI data, and any SAP or non-SAP reporting tool, you can easily visualize the data.
The picture below illustrates an example for how the visualization of the DSO KPI might look like (in this example based on PowerBI).
Similarly, you can derive the KPI of “Days Sales Outstanding” (DSO). The combination of DSO and CLV is a very powerful toolset for measuring your sales, your process efficiency, and evaluating your customer base.
Next Product to Buy
NPTB (Next Product to Buy) is an interesting challenge in cross-sales. If you have ever shopped online at the leading online marketplace, you will know this “other customers who have bought this product also have bought that product”.
Imagine you’re a salesperson, no matter if in B2B or B2C. Using the same idea, you can propose additional products from your portfolio based on current or past buying decisions of your customers.
The screen shot below shows a sample implementation based on Datavard Glue and PowerBI of this cross-selling (or up-selling) potential:
To implement you will need:
- Past sales data for your customers to analyze and to feed into a recommendation engine
- A recommendation engine… Don’t be scared by the “engine” part, these things may be very complicated, but there are simple open source scripts available to get started with. Especially if you’ve never looked at your data from this perspective, you may get very good initial results.
- An integration solution for implementing a continuous improvement of the data, and a feedback loop of the data into your sales business process. This would of course be Datavard Glue.
Questions, comments? Leave me a message: