SAP Impedance


Black Friday and Cyber Monday puts the maximum stress on the IT systems. Many retailers who have transformed their businesses to complete with Amazon. Have learned directly from ex-Amazon software engineers, that Amazon sets a baseline for testing its online systems at 1.5x last years traffic. Then backs off the number of servers until transactions take up to 2 weeks to arrive at data consistency.

Many retailers using SAP and other ERP's are architected around on-premises batch processing not real-time database processing. Often the front end web systems can scale up on larger hardware, but not scale out automatically to use more servers or virtual services such as AWS or GCP.

Herein lies the rub. Retailers such as Nordstrom are now in the process of transforming their backend ERP systems to scale out to handle a do or die competitive change to omni-commerce. The ability to seamlessly interact with customers across all shopping channels--digital and physical in-store retail including returns--in an integrated shopping experience.

Ironically firms such as Nordstrom, Target and REI hire key staff away from Amazon. Many of these key staff are also practicing engineers in data sciences, machine learning, operational optimizations and distributed systems. Because 77% of the worlds transactional economy touches an SAP system somewhere (@dahowlett). Thus their new roles outside Amazon is to transform SAP impedance for distributed business systems than can scale out to handle busty real-time customer business.


The change to real-time business is also a change to seasonal forecasting with real-time web analytics and sales consumption (physical store and web sales) built around a machine learning data pipeline which feeds optimization algorithms from @gurobi page .

This real-time business architecture is underpinned by a shift to distributed systems computing with similarities and differences with SAPs XO business centric architecture (we'll come back to XO). To keep different system services running and secure. Each service is self-monitored. Online services monitor system back pressure. Services that are failing over are automatically quarantined as new services come online (Workday is underpinned by Scala and Akka built on this internal architecture). More typically apps from the client-server era systems are separated into cloud based API services and monitored by New Relic. Firms serious about real-time business at scale such as Uber and Lyft write new services in Go (golang). Go Google's system programming language has tests built into each function to increase resiliency. Industry analysts often call provisioning and orchestrating services developer operations (devops).

# Takeway Retailers busy transforming their business systems for omni-commerce are more concerned with separating services to remove impedance for real-time business than adding impedance with XO centric business architecture. Charting a course toward real-time business by leaving current SAP (and Oracle) modules in place while connecting up new capabilities is the Occam's razor solution for teams faced with competing with Amazon.

# Coda. No surprise if SAP transforms Qualtrics toward a highly granular feedback system that can survey and capture measurement data as part of a ML data pipeline for realtime businesses.