The next BriefingsDirect digital business innovation interview explores how the powerful combination of deep analytics and the procurement function makes businesses smarter and more efficient.
When the latest data science techniques are applied to more data sets that impact supply chains and optimize procurement, a new strategic breed of corporate efficiency and best practices emerge.
Listen to the podcast. Find it on iTunes. Get the mobile app. Read a full transcript or download a copy.
To learn how data-driven methods and powerful new tools are transforming procurement into an impactful intelligence asset, BriefingsDirect recently sat down with David Herman, Chief Data Scientist for Strategic Procurement at SAP Ariba. The discussion is moderated by Dana Gardner, Principal Analyst at Interarbor Solutions.
Here are some excerpts:
Gardner: Why is procurement such a good place to apply the insights that we get from data science and machine learning (ML) capabilities?
Herman: Procurement is the central hub for so many corporate activities. We have documents that range from vendor proposals to purchase orders and invoices to contracts, and requests for proposal (RFPs). Lots and lots of data happens here.
So the procurement process is rich in data, but the information historically has been difficult to use. It’s been locked away inside of servers where it really couldn't be beneficial. Now we can take that information in its unstructured format, marry it with other data – from other systems or from big data sources like the news -- and turn it into really interesting insights and predictions.
Gardner: And the payoffs are significant when you're able to use analysis to cut waste or improve decisions within procurement, spend management, and supply chains.
Procurement analysis pays
Herman: The very nature of spend analysis is changing. We implemented a neural network last year. Its purpose was to expedite the time it takes to do spend analysis. We dropped that time by 99 percent so that things that used to take days and weeks can now be done in mere hours and minutes.
Because of the technology that is available today, we can approach spend analysis differently and do it more frequently. You don’t really have to wait for a quarterly report. Now, you can look at spend performance as often as you want and be really responsive to the board, who these days, are looking at digital dashboard applications with real-time information.
Gardner: How is this now about more than merely buying and selling? It seems to me that when you combine these analytic benefits, it becomes about more than a transaction. The impact can go much deeper and wider.
Herman: It’s strategic -- and that's a new high plateau. Instead of answering historic questions about cost savings, which are still very important, we’re able to look forward and ask what-if kinds of questions. What is the best scenario for optimizing my inventory, for example?
That's not a conversation that procurement would normally be involved in. But in these environments and with this kind of data, procurement can help to forecast demand. They can forecast what would happen to price sensitivity. There are a lot of things that can happen with this data that have not been done so far.
Gardner: It's a two-way street. Not only does information percolate up so that procurement can be a resource. They are able to execute, to act based on the data.
Herman: Right, and that's scary, too. Let's face it. We're talking about peoples’ livelihoods. Between now and 2025, things are going to change fundamentally. In the next two to three years alone, we are going to see positions [disappear], and then we're going to have a whole new grouping of people who are more focused on analysis.
The reality is that of any kind of innovation -- any kind of productivity -- follows the same curve. I am not actually making this prediction because it’s the result of ML or artificial intelligence (AI). I am telling you every great increase in productivity has followed the same curve. Initially it impacts some jobs and then there are new jobs.
And that's what we're looking at here, except that now it’s happening so much faster. If you think about it, a five-year period to completely reshape and transform procurement is a very short period of time.
Gardner: Speaking of a period of time, your title, Chief Data Scientist for Strategic Procurement, may not have even made much sense four years ago.
Herman: That's true. In fact, while I have been doing what I'm doing now for close to 30 years, it has had different names. Sometimes, it's been in the area of content specialistor content lead. Other times, it's been focused on how we are managing content in developing new products.
And so, really, this title is new. Yet it’s the most exciting position that I've ever had because things are moving so much faster and there is such great opportunity.
Gardner:I'm sure that the data scientists have studied and learned a lot about procurement. But what should the procurement people know about data science?
Curiosity leads the way
Herman: When I interview people to be data scientists, one of the primary characteristics I look for is curiosity. It’s not a technical thing. It’s somebody who just wants to understand why something has happened and then leverage it.
Procurement professionals in the future are going to have much more available to them because of the new analytics. And much of the analytics will not require that you know math. It will be something that you can simply look at.
For example, SAP Ariba’s solutions provide you with ML outcomes. All you do is navigate through them. That’s a great thing. If you're trying to identify a trend, if you're trying to look at whether you should substitute one product for another -- those analytic capabilities are there.
SAP Ariba's solutions provide you with ML outcomes. All you do is navigate through them. That's a great thing.As for a use case, I was recently talking to the buyer responsible for staffing at one of SAP’s data centers. He is also responsible for equipping it. When they buy the large servers that run S4/HANA, they have different generations of hardware that they leverage. They know the server types and they know what the chip lifecycles look like.
But they've never been able to actually examine their own data to understand when and why they fail. And with the kinds of things we're talking about, now they can actually look to see what's going on with different chipsets and their lifecycles -- and make much more effective IT deployment decisions.
Gardner: That's a fascinating example. If you extrapolate from that to other types of buying, you are now able to look at more of your suppliers’ critical variables. You can make deductions better than they can because they don't have access to all of the data.
Tell us about how procurement people should now think differently when it comes to those what-if scenarios? Now that the tools are available, what are some of the characteristics of how the thinking of a procurement person should shift to take advantage of them?
Herman: Anyone who's negotiated a contract walks away, glad to be done. But you always think in the back of your head, What did I leave on the table? Perhaps soon the prices will go up, perhaps the prices will go down. What can I do about that?
We introduced a product feature just recently in our contracts solution that allows anyone to not only fix the price for a line item, but also make it dynamic and have it tied to an external benchmark.
We can examine the underlying commodities associated with what you are buying. If the commodities change by a certain amount – and you specify what that amount is -- you can then renegotiate with your vendor. Setting up dynamic pricing means that you're done. You have a contract that doesn't leave those what-ifs on the table anymore.
That's a fundamental shift. That’s how contracts get smart -- a smart contract with dynamic pricing clauses.
Gardner: These dynamic concepts may have been very much at home in the City of London or on Wall Street when it comes to the buying and selling of financial instruments. But now we’re able to apply this much more broadly, more democratically. It’s very powerful -- but at a cost that's much more acceptable.
Is that a good analogy? Should we look to what Wall Street did five to 10 years ago for what is now happening in procurement?
Herman: Sure. Look, for example, at arbitrage. In supplier risk, we take that concept and apply it. When trying to understand supplier risk, begin with inherent risk. From inherent risk we try to reduce the overall risk by putting in place various practices.
Sometimes it might be an actual insurance policy. It could also be a financial instrument. Sometimes it’s where we keep the goods. Maybe they are on consignment or in a warehouse.
There are a whole host of new interesting ways that we can learn from the positives and negatives of financial services -- and apply them to procurement. Arbitrage is the first and most obvious one. I have talked to 100 customers who are implementing arbitrage in various forms, and they are all a little bit different. Each individual company has their own goal.
For example, take someone in procurement who deals with currency fluctuations. That kind of role is going to expand. It's not going to be just currency -- it is also going to be all assets. It is ways to shift and extend risk out over a period of time. Or it could even be reeling in exposure after you have signed a contract. That's also possible.
Gardner: It seems silly to think of procurement as a cost center anymore. It seems so obvious now -- when you think about these implications -- that the amount of impact to the top line and bottom line that procurement and supply chain management can accomplish is substantial. Are there still people out there who see procurement as a cost center, and why would they?
From cost to opportunity
Herman: First of all, it's very comfortable. We can demonstrate value by saving money, and it goes right to the bottom line. This is where it matters the most. The cost is always going to be a factor here.
As one chief procurement officer (CPO) recently told me, this has been a kind of a shell game because he can't actually prove how much his organization has really saved. We can only put together a theoretical model that shows how much you saved.
As we move forward, we are going to find that cost remains part of the equation -- I think it will always be part of the equation – yet the opportunity side of the equation with the ability to work more effectively with sales and marketing is going to happen. It's actually happening now. So you will see more and more of it over the next three to five years.
We can demonstrate value by saving money, and it goes right to the bottom line. This is where it matters the most. The cost is always going to be a factor here.Gardner: How are analytics being embedded into your products in such a way that it is in the context of such a value-enhancing process? How are you creating a user experience around analytics that allows for new ways to approach procurement?
Herman: Again, supplier risk is a very good example. When a customer adopts the SAP Ariba Supplier Risk solution, they most often come with a risk policy in place. In other words, they already know how to measure risk.
The challenges with measuring risk are commonly around access to the data. Integration is really hard. When we went about building this product we focused first on integration. Then we came up with a model. We take the historical data and come up with a reference model. We also really worked hard to make sure that any customer can change any aspect of that model according to their policy or according to whatever scenario they might be looking at.
If, for example, you have just acquired a company, you don’t know what the risks look like. You need to develop a good look at the information, and then migrate over time. With supplier risk management, both the predictive and descriptive models are completely under the control of our customers. They can decide what data flows in and becomes a feature of that model, how much it is weighted, what the impacts are, and how to interpret the impact when it's finished.
We also have to recognize when you’re talking about data outside of the organization that is now flowing in via big data, that this is an unknown. It's not uncommon for somebody look at the risk platform and say, Turn off that external stuff so I can get my feet under the table to understand it -- and then turn on this data that’s flowing through and let me figure out how to combine them.
At SAP Ariba, that’s what we are doing. We are giving our customers the tools to build workflow, to build models, to measure them, and now with the advent of the SAP Analytics Cloud be able to integrate that into S/4HANA.
Gardner: When we think about this as a high-productivity benefit within an individual company, it seems to me that as more individual companies begin doing this that there is a higher level of value. As more organizations in a supply chain or ecosystem share information they gain mutual productivity.
Do you have examples yet of where that's happening, of where the data analytics sharing is creating a step-change of broader productivity?
Shared data, shared productivity
Herman: Sure, two examples. The first is that we provide a benchmarking program. The benchmarking program is completely free. As long as you are willing to share data, we share the benchmarks.
The data is aggregated, it's anonymous, and we make sure that the information cannot be re-identified. We take the proper precautions. Then, as a trusted party and a trusted host we provide information so that any company can benchmark various aspects of their specific performance.
You can, for example, get a very good idea of how long it takes to process a purchase order, the volumes of purchase orders, and how much spend is not managed because you don't have a purchase order in place. Those kinds of insights are great.
When we look at analytics across industries we find that most supply chains have become brittle. As all of us become leaner organizations, ultimately we find that industries end up relying on one or two critical suppliers.
For example, black pigment for the automotive industry was provisioned for all of the major manufacturers by just one supplier. When that supplier had a plant fire and had to shut down their plant for three months it was a crisis because there was no inventory in the supply chain and because there was only one supplier. We actually saw that in our supplier risk product before it happened.
The industry had to come together and work with one another to solve that problem, to share their knowledge, just like they did during the 2008-2009 financial crisis.
In the financial crisis, we found that it was necessary to effectively help other company’s suppliers. Traditionally that would be called collusion, but it was done with complete transparency with the government.
When you look at such ways that information can be shared -- and how industries can benefit collectively -- that's the kind of thing we see as emerging in areas like sustainability. With sustainability we are looking for ways to reduce the use of forced labor, for example.
In the fishing industry, shrimping companies have just gone through their industry association to introduce a new model that collectively works to reduce the tremendous use of forced laborin that industry today. There are other examples. This is definitely happening.
Gardner: What comes next in terms of capabilities that build on data science brought to the procurement process?
Herman: One of the most exciting things we’re doing is around contracts. Customers this quarter are now able to evaluate different outcomes across all of their contracts. A prominent use case is that perhaps you have a cash flow shortage at the end of the year and it’s necessary to curtail spend. Maybe that’s by terminating contracts, maybe it’s by cutting back on marketing.
We picked an area like marketing so that we can drill down to evaluate rights and obligations and assess the potential impact to the company canceling those contracts. There is no way to do this today at scale other than manually.
If the chief financial officer (CFO) were to approach someone in procurement and ask this question about cash flow, they would bring in your paralegals and lawyers to begin reading the contracts. That's the only way today.
Customers are now able to evaluate different outcomes across all of their contracts. We are teaching machines to interpret the data, to evaluate cause and effect and then classify the impact so decision makers can act quickly.What we are doing right now is teaching machines to interpret that data, to evaluate the cause and effect -- and then classify the impact so that the decision makers can take action quickly.
Gardner: You are able to move beyond blunt instruments into a more surgical understanding -- and also execution?
Herman: Right, and it redefines context. We are now talking about context in ways that we can't do today. You will be able to evaluate different scenarios, such as terminating relationships, push out delivery, or maybe renegotiating a specific clause in a contract.
These are just the very beginnings of great use cases where procurement becomes much more strategic and able to respond to the scenarios that help shape the health of the organization.
Gardner: We spoke before about how this used to be in the purview of Wall Street. They had essentially unlimited resources to devote to ML and data science. But now we are making this level of analysis as-a-service within an operating expense subscription model.
It seems to me that we are democratizing analysis so that small- to medium-size businesses (SMBs) can do what they never used to have the resources to do. Are we now bringing some very powerful tools to people who just wouldn’t have been able to get them before?
Power tools to the people
Herman: Yes. The cloud providers create all kinds of opportunities, especially for SMBs, because they are able to buy on demand. That’s what it is. I am able to buy what I need on demand, to negotiate the price based on whether it’s on peak or off peak and get to the answers that I need much more quickly.
SAP Ariba made that transition to a cloud model in 2008, and this is just the next generation. We know a lot about how to do it.
Gardner: For those SMBs that now have access to such cloud-based analytics services, what sort of skills and organizational adjustments should they make in order to take advantage of it?
Herman: It’s interesting. When I talk to schools, to undergraduates and graduate students, I find that many of those folks are coming out of school with the right skill sets. They have already learned Python, for example, and they have already built models. There is no mystery, there is no voodoo about this. They have built the models in the classroom.
Just like any other business decision, we want to hire the best people. So, you will want to maybe slip in a couple of questions about data sciences during your interviews, because it’s the kind of thing that a product manager, an analyst, and an IT leader need to know in the near future.
With the transition of the baby boomers into retirement, Millennials are coming up as this new group which is extremely talented. They have those skill sets and they are driven by opportunity. As you continue to challenge them with opportunities, my experience is that they continue to shine.
Gardner: David, we have talked about this largely through the lens of the buyers. What about the sellers? Is there an opportunity for people to use data in business networks to better position themselves, get new business, and satisfy their markets?
Discover new business together
Herman: We need a good platform to discover these kinds of opportunities. Having been a small business owner myself, I find that the ability for me to identify opportunities that trigger business is really essential. You really want to be able to share information with your customers and understand how you can generalize those.
I recently spoke to a small business owner who uses Google Sheets. At the end of every call, everybody on this team writes down what they had learned about the industry so they could share it among themselves. They would write down the new opportunities that they heard in a separate section of the sheet, in a separate tab. What were the opportunities they saw coming up next in their industry? That’s where they would focus their time in building a funnel, in building a pipeline around it.
When looking at it from that perspective, it’s really useful. Use the tools we have to get into these new areas of access -- and you win.
Gardner: What should people expect in the not too distant future when it comes to the technologies that support data science? Are there any examples of organizations at the vanguard of their use? Can they show us what others should expect?
We now have to look at it differently. We need to look at how to use ML to validate your risks and assumptions and then concentrate investments. ML is going to help you find your answers faster.Herman: Here’s the way I look at it: If we are going to think about how much money you could invest and bet on the future, maybe we have 7 percent of operating income to play with, and that’s about it. That has been in the common in the past, for us to spread that spending across four, five, or six different bets.
I think now we have to look at it differently. We need to look at how to use ML to validate your risks and assumptions, of how to validate your market and then concentrate investments. We can take that 7 percent and get more out of it. That’s how ML is going to help, it’s going to help you find your answers faster.
Gardner: How should organizations get themselves ready? What should organizations that want to become more intelligent -- to attain the level of an intelligent enterprise, an intelligent SMB -- what do you recommend that they do in order to be in a best position to take advantage of these tools?
Collaborate to compete
Herman: Historically we asked, What is your competitive advantage? That’s something that we talked about in the 1980s, and then we later described learning as your core competency. Now in this time, it’s who you know. It’s your partnerships.
Going back to what Google learned, Google learned how to connect content together and make money. Facebook one-upped them by learning about the relationships, and they learned how to make money based on those relationships.
Going forward, customer networks and supply chains are yourdifferentiation. To plan for that future, we need to make sure that we have clear ways to collaborate. We can work to make the partners strategic, and to focus our energy and bets on those partners who we believe are going to make us effective.
When you look at what are the key enablers, it’s going to be technology. It’s going to be analytics. To me that’s a given in these situations. We want to find someone who is investing, looking forward, and who brings in these new capabilities -- whether it’s bitcoin or something else that is transformative in how we make companies more network-driven.
Gardner: So perhaps a variation on the theme of Metcalfe’s Law -- that the larger the network, the more valuable it is. Maybe it’s now the more collaboration -- and the richer the sharing and mutually assured productivity -- the more likely you are to succeed.
Herman: I don’t think Metcalfe’s Law is over yet. We are going to find between now and 2020, that’s where this is at.
Listen to the podcast. Find it on iTunes. Get the mobile app. Read a full transcript or download a copy. Sponsor: SAP Ariba.