Analysts and pundits love to argue. Unfortunately, these arguments are too often semantical, with little relevance to end customers.
The analytics debate is an exception. Should enterprise analytics be approached as a pure-play, or is analytics best consumed in the context of enterprise applications?
This issue matters. Why? Because it informs buyer choices - and data platform roadmaps. It comes as no surprise that Oracle has a clear position on this topic.
I picked that up with Steve Miranda, Oracle's EVP of Applications Development, on November 2, the day of Oracle's latest analytics announcements. That news features the debut of new Oracle Fusion Analytics for Supply Chain (see: Oracle Fusion SCM Analytics Helps Organizations Build Resilient Supply Chains with Improved Visibility). As per the press pitch, Oracle Fusion SCM Analytics will help customers:
- Gain visibility into supply chain performance
- Correlate supply chain processes into business goals
- Detect, understand and predict supply chain issues
On that third point: "Machine learning and predictive analytics provide automated insights into supply chain data to identify cost-saving opportunities, surface operational bottlenecks, and predict future outcomes." There is also a data warehouse component, adding to the analytics already developed for ERP and HCM:
To help customers discover unique insights faster, Oracle Fusion Analytics Warehouse enables organizations to break down organizational silos and analyze financial, supply chain, and HR data on a single integrated cloud platform. Fusion SCM Analytics is the most recent Fusion Analytics offering for the Fusion Cloud Applications suite, adding to Fusion ERP Analytics and Fusion HCM Analytics.
Analytics in action - Oracle customer scenarios
For context, Oracle held an Oracle Live session on November 2, Connect Your Enterprise with AI-Driven Analytics (replay available). I asked Miranda if he could preview the customer examples. He responded:
So we have two great interviews, one with the University of Pittsburgh Medical Center (UPMC), and the other with CORSAIR Gaming, which will not only showcase their use of Fusion Analytics and the Analytics Warehouse, but also their use of our SaaS applications.
Oracle has offered analytics and planning tools for quite some time. What's different as of today?
With these announcements, the two big things are: one is just the out-of-the-box integration now with Fusion Applications, so having a SaaS-style warehouse, to replace the custom ETL that you've written to do it.
Then the second one is some of the machine learning. The machine learning you'll see in the demonstration is going to be expanding the digital assistant to the BI tool so that you can interact in voice or conversational UI with BI, and ask business-style questions in the way you ask business questions, and get the right response in the right format, or right report - as opposed to searching.
What is driving these customer analytics investments? Miranda says it's about moving away from the canned reports of old, with their labor-intensive prepwork:
What you're going to hear [our customers] talk a lot about is replacing custom reports and custom extracts, for what I would call metrics reporting. CORSAIR brings up the example: they had a request for 400 reports to be opened up for transactional ERP, and essentially replaced all of that with the out-of-the-box reports and the configuration we get in the data warehouse. That's the first part, and that's the immediate gain they're going to talk about now, where they're starting to go.
Then there is the AI component:
What they're also talking about is machine learning or AI, the real analytics. And so, in both cases, you're going to hear interesting examples of the impact of the last 20 months... For example, CORSAIR has really seen both ends of their business [change].
Supply chain disruptions forced the issue:
Number one is because of the work from home... CORSAIR has seen a spike in demand. At the same time, they're dealing with the chip shortage and the manufacturing shortage that many organizations are dealing with. So they've had to rethink their products, their supply chain, and how to better fill the demand to their customers. And they're using the predictive analytics and supply chain analytics for that.
The purpose of the Oracle Fusion Analytics Warehouse is to support customers' needs across functional areas, with predictive and AI capabilities included:
With UPMC, it's a little bit different. They are focused on the HCM side of things. Their biggest pressure has been in the workforce. Whatever term you want to choose, whether it's the great resignation, or whatever you want to label it, they've got a lot of pressure on attrition, and a lot of pressure on rehiring. So they've started to use predictive analytics. On the HCM side, we have prediction of performance, prediction of attrition - that's together with the machine learning we have on the recruiting side, to rank best candidates, etc.
Facing predictive and SCM analytics challenges
Hold up - predicting attrition is not necessarily easy. Even if we identify a risk, by the time we intervene constructively, it's too late to retain a happy employee. I asked Miranda: are we getting any better there? Short answer: yes. To do that, he says, Oracle tracks "dozens of attributes."
We can piece together common patterns. For example, in the HCM case, obvious things like: 'When's your last raise? And when's your last promotion?' But also things like, 'How many managers have you had over the last three years? How many performance appraisals have you done? Have you moved address lately? Have you relocated and changed your home address?' So all these signals come in, and then we measure those against other people in your company in that situation. And we're able to give a prediction on the likelihood of attrition, based on a variety of different factors.
Now we have the introduction of Oracle Fusion SCM Analytics. Let's face it - companies could have used more advanced SCM analytics the last couple of years. Then again, supply chains are still disrupted. Miranda:
We've always had supply chain planning tools. But now the level of importance, and I think the risk factor, and the number of variables, has just increased dramatically.
Miranda cited transportation planning as an example. The need for sophisticated analytics surged:
Transportation planning didn't used to get that complicated. Now, it's very complicated in terms of the cost, the lead time... This leads into potentially planning on: 'Do you have local warehouses where you can store things? Or do you have more distributed manufacturing? Are you an organization where you manufacture closer to where it's distributed, so you avoid that transportation?'
My take - analytics should be embedded and collaborative
Do cloud ERP vendors have an analytics edge against pure-play vendors? Miranda obviously thinks so, but why? One huge issue is external data sources, weather data being a prime example. To be fair, even pure-play analytics vendors were historically weak on external data sources. That's changed.
Effective supply chain planning requires the external, whether it's weather data, or demand signals from customer-facing apps. If ERP vendors want to solve analytics problems, they better be data platforms. Miranda responded:
You are 100% correct, whether it be weather data or demand signals, or if you're a retailer, you need data from your merchandising application. Or if you're in banking, and you don't use our banking application, [the ERP vendor still needs that data for analytics].
Miranda says analytics vendors struggle with the reverse:
The non-ERP vendors don't have the out-of-the-box integration with ERP, which gives two challenges. One you mentioned, which is what you call data cleansing. [In our case], that comes directly out of ERP; it's mapped directly to dimensions that we understand. So there's no confusion in terms of the definition. And two, it's timely, because you don't have this kind of batch ETL running between the systems; we take care of that integration for you.
To make a legit case, ERP vendors must do more:
We also have the ability to load third-party data, not only things like weather data... Let's say you use Oracle HR, except you didn't use time and labor, if you use Kronos or somebody, you just bring that in from Kronos as an external data source.
External data sources - check that box for Oracle. But I believe the future of analytics is embedded. Analytics and transactional systems should fuse, pun intended. In theory, cloud ERP vendors should have an edge embedding their own systems, but that isn't always the case. In terms of who is doing the best job embedding analytics, I'd say the jury is still out. Miranda:
If you think about transactional screens, you're usually looking at [something like] an invoice... or a figure that applies to that context. And that's what we're starting to blend in. That's what we call embedded BI.
Analytics should also be collaborative. When we meet and talk numbers, we come to decisions. Or we message back-and-forth about a store that is underperforming, and why. That conversation soon fades into the messaging feed; the online meeting is archived. Yes, collaborative tools are getting better, but we still have a problem. The context of our data discussions is quickly lost, Slacked into obscurity if you will. Collaborative analytics in context will advance this. Oracle isn't making any news announcements on this topic as of today, but I'd like to hear more from them on this.
Finally, there is the digital assistant question. Miranda and I have spoken on this before; I'm particularly skeptical about where conversational UIs are today. That's ironic; anything that makes enterprise search better should be welcome. Oracle approaches digital assistants with both text (chatbot) and verbal interfaces, and yes, they have customer references. Miranda says JPMorgan Chase is using Oracle's digital assistant as the front end to their recruiting, so if you're applying for a job there, you can give it a spin. Hertz is also rolling it out "pervasively" across HR and finance.
Miranda points out that the job of an enterprise assistant is narrower than Alexa, who must be able to answer just about any question I might ask (and often does a horrible job, especially with search queries). So, I ask Miranda, what types of questions can Oracle's digital assistant answer today?
What was our revenue last quarter for applications in North America?
Okay, that's pretty useful. But perhaps more useful is the ability for a contextual follow-up, without going back to square one:
It knows the context that I'm asking for, as opposed to having to restate the entire sentence. So that's the kind of thing that we can zero in on.
So, you could follow up that general revenue question with a geographic or product dimension.
We actually know the values of those dimensions in your system, because the data is there.
Alexa can't do that - so I'll call it progress. Maybe I'll apply for a job at JPMorgan Chase and try it for myself - if I can do it without diginomica finding out. Wish me luck...
End note: prior to the Oracle Fusion Analytics announcements, Steve Miranda published this post on LinkedIn, See your business in new ways with AI, Analytics, and Fusion Applications together.
Updated, 6:30 am UK time November 4, 2021, with the addition of subheading and a number of smaller tweaks for reading clarity.