Daily Archives: May 6, 2008

Spend Analysis Meme Busting, Part II

Today’s post is again courtesy of Eric Strovink of BIQ (acquired by Opera Solutions, rebranded ElectrifAI).

9. Spend analysis is a “big iron” problem requiring large servers and databases.
Nonsense. Global 100 datasets fit easily onto ordinary laptop computers, and a modern laptop computer can deliver near-instantaneous drill-down times.

10. Static reports can deliver profound insight, replace procurement analysts and sourcing consultants, and even direct the course of entire sourcing programs.
This idea surfaces every now and then in the writings of pie-in-the-sky analysts and on blogs like Spend Matters. Most recently, it is being promulgated by spend analysis vendors who have run out of new marketing ideas. Of course, if it were really true, those reports would have existed long ago, and we’d all be out of a job.

11. Extracting data from accounting/ERP systems is difficult.
In fact, it’s easy, and usually doesn’t require IT resources to accomplish. The only difficulty I’ve ever experienced was a situation where a customer’s head of IT folded his arms and swore up-and-down that it simply wasn’t possible to dump data from their accounting system. One telephone call to the accounting system vendor (a helpful Canadian firm) provided the command necessary to extract the data trivially.

12. Accounts payable data is all you need for spend analysis.
This meme is on the decline. By now, many practitioners know that A/P spend visibility will find low-hanging fruit, and that it will identify buckets of spend that might be worth a closer look. But they also realize that in order to get to the next level of savings, it’s necessary to build commodity-specific analyses that take into account existing contract terms and invoice-level data, as well as balance-of-trade and demand-side considerations. And, this is only the tip of the iceberg when one considers the analysis possibilities buried in HR data and in ops data such as service repair records. Experienced practitioners build data analysis cubes whenever necessary, throwing them away when done, or retaining them if it is useful to do so.

13. Real data analysis requires statistics and applied mathematics.
Some consultants argue that unless rigorous statistical analysis is applied to a dataset, no meaningful conclusions can be drawn. For sourcing, though, the data visibility provided by a spend analysis system is usually more than sufficient. For example, try loading invoice data for a particular SKU over an extended time frame, and scatter-plotting the price. Is the price the same? Many times it isn’t, even if you have a contract that should have locked it down. When you find this pattern — and chances are you will — you’ve just written yourself a check, with no applied mathematics required.

14. Spend analysis systems should react in real time to real-time data.
This meme is becoming popular amongst armchair analysts and bloggers, but it’s scoffed at by practitioners. What does real time mean? When the requisition arrives? When the PO is issued? When the invoice arrives? When partial payment is made? When full payment is made? And what is anyone going to do about it, anyway, in “real time?” This ties into the notion of the “executive dashboard,” where the idea seems to be that an errant transaction will set off some sort of red alert. Even intrusion detection systems, which actually can justify a real-time component, have largely given up on this idea; the “fact” of an alert is not necessarily indicative of anything, and only in-depth after-the-fact pattern analysis can distinguish signal from noise.

15. “We already purchased an enterprise license for [xyz BI system or OLAP database], so we don’t need yet another analysis capability.”
This is the classic argument that the head of IT feeds to the CFO, which enables the CFO to kill the incipient spend analysis project, as she often does when fed such an argument (CFOs being people who really dislike spending money). Everyone feels good about saving money, and they move on to the next item on the agenda. Procurement doesn’t dare argue with IT, typically, so that’s the end of the story. Unfortunately, IT and the CFO have just doomed their company to another N years of zero spend visibility, because (a) nobody in Procurement knows how to configure either a BI system or an OLAP database, and (b) nobody in IT is going to take any time out to help them — assuming, that is, that IT themselves understand the BI system or OLAP database, which in many cases they do not.

16. “It’s important to deploy a spend analysis solution enterprise-wide, immediately.”
Here are three reasons why this can be a poor idea.

  • Procurement is a tricky business with lots of opportunities to overlook things even when work has been done competently and professionally. A spend analysis system, unfortunately, does a great job of pointing out those things. Why should Procurement hang its dirty laundry out the window for the whole company to see? Wouldn’t it be a better idea to find (and correct) obvious errors quietly and privately, before publishing a company-wide spend cube — if a company-wide cube is even necessary?
  • Company-wide initiatives create company-wide inertia, starting with IT sticking its thumbs in its belt and making trouble about how much time/money/effort it’s going to take to “set up a server” or, worse, how much time and effort it’s going to take to “evaluate the vendor’s remote hosting site for security compliance.” Wouldn’t you rather be running the spend cube immediately on your laptop, rather than waiting months for IT to get its act together? You can always deploy the cube company-wide, later, without any pressure.
  • Company-wide spend analysis initiatives create company-wide opportunities to kill the initiative. These days, spend analysis can be obtained inexpensively and set up quickly, within your discretionary budget and without asking anyone’s permission.

17. It will cost a lot to outsource cube construction, or to train internal resources on the spend analysis system.
This situation has improved substantially, and it continues to improve. New services offerings from some vendors have cut this price substantially from the levels of a few years ago; and contingency-based sourcing consulting firms will not only build your spending cube for free, but also return solid savings to your bottom line.

In summary:

9. Spend analysis can be done on a modern laptop computer.

10. A static report is not capable of providing much in the way of insight.

11. Data extraction from the vast majority of accounting / ERP systems is quite easy. (It might take some mapping to get it into cube form, but that’s why spend analysis tools come with good E”T”L tools.)

12. Accounts payable data is just the beginning. Every database is its own gold mine!

13. Data analysis starts with well mapped and relatively complete data – not advanced statistics or applied mathematics. Visibility is key.

14. The spend analysis system should be capable of accepting updated data when the data – and the analyst – is ready for new data. The answer does not lie in “real-time” or “monthly updates” because every organization, and every data set, is different.

15. BI & OLAP is not spend analysis.

16. As with any solution, initial deployment should be limited in scope to that which is controllable while the learning curve is overcome.

17. Training and consulting is a lot more affordable than you think, and if deployed on invoice data first, will return immediate ROI.

Thanks again, Eric!