Category Archives: Spend Analysis

Procurement Trend #06. Data-Based Predictive Analytics

Three annoying anti-trends remain. We’re so close to the end that we can almost taste the bitter-sweet victory, but the sour taste in our mouths still remains as we must continue to provide those fashionably-challenged futurists with counter-examples to the trends of their fore-fathers that no one who didn’t lock themselves in a windowless padded room would try to pass off as a trend of tomorrow. We want to shame them for their stupidity, but we will leave their hard-earned humiliation for LOLCat, who is obviously quite fed up at having to spend yet another life listening to their ludicrousness, but still finding the time to point out how LOLCats have been sustainable at least since the first corrugated cardboard box was created.

So why do these pit-dwelling prophets from Hawalius keep pushing us trends from the rubbish pile? Besides the fact that some of them obviously spent the best part of last decade in a rancid cave, probably because they look around, see the laggard organizations still struggling with last decade’s technology, and assume they can still sell last decade’s leftover snake oil in today’s marketplace. Thus, if most organizations are struggling with proper historical spend analysis, data-based predictive analytics is obviously a future trend, and

  • good decisions require good data

    and so few organizations have good data

  • inventory forecasting is getting harder and harder

    as sudden changes in unemployment rate, interest rates, and brand sentiment as well as unexpected supply chain delays or competitive product introductions can all have a large impact on demand

  • market prices are getting even harder to predict in volatile markets

    and profitability often depends on slim margins

Which would be great reasoning if leading organizations hadn’t figured this out over a decade ago and moved on to doing something about it a while ago!

So what does this mean to you?

Clean and Enrich Your (Master) Data

Dirty data dictates dastardly decisions. And those never end well. But don’t go crazy trying to do it. 100% clean data is a pipe dream, and, as with most situations, the 80/20, or, to be more precise, the 90/10 rule applies. Clean and enrich as required to confidently map 90%+ of spend, including 90%+ of the spend for the top 90%+ of suppliers and the top 90%+ of products. Stop when the effort exceeds the return. With a good mapping tool, the mapping can be done for even the largest Fortune 500 by hand in a week. Depending on how good the data is, the analyst might even get to 95% or even 98%. Then, identify any glaring weaknesses (such as supplier financial or risk data, market data, or cost breakdowns relative to a Bill of Material) that are important from a spend analysis or should cost modelling viewpoint, and get that data.

Put Protocols and Safeguards in Place to Keep your (Master) Data that Way

It’s going to take time, money, and manpower to map, clean, and enrich the data. This will be time, money, and manpower wasted if protocols aren’t put in place to make sure not just anyone can update master data, or at least not without review and verification. Put workflows and approvals in place to minimize the chances of bad data getting into the system or data getting out of whack too quickly.

Automatically Augment Your (Master) Data with Market Data

Good historical data is good. But current market data is better. With past and current data you can not only know current conditions, but with current market data, updated regularly, you can compute trends.

Use All the Data to Predict Trends and Make Sourcing Decisions

Use the computed trends to predict likely future conditions based upon the trends and current market movements. Based on this data, you can judge whether or not it is a good time to source a category and lock in long-term pricing.

Optimize, don’t Compromise!

Continuing on our theme of analysis and optimization, every e-Sourcing suite on the market will support your organization in its sourcing activities, but not every product will allow your organization to optimize it’s sourcing activities.

Optimization requires advanced sourcing capability, and advanced sourcing requires the ability to analyze data, not just collect and report on data.

This means, that at the very least, you will require:

  • true spend analysis,
  • true category analysis,
  • true cost-based bidding, and/or
  • true bid optimization.

Without at least one of these capabilities, you’ll never optimize your spend. So don’t even both to try without them.

If I Succeed in Destroying Dashboards and Razing Report Writers, What Next?

In yesterday’s post, where I responded to the smart alecks, I noted that, once dashboards are destroyed and report writers are razed, there was about a half-dozen next logical steps that could be taken to improve today’s spend analysis solutions, even if that solution was BIQ.

Should cost modelling, award optimization based on historical data and business rules, and federation across related data sets for deeper dives are pretty obvious. Are there somewhat less obvious advancements we should also be thinking of?

Of course. One rung up the ladder, three of them are:

Predictive Modelling

Once you have should-cost modelling, the next logical step is predictive modelling. Use historical data to extract pricing trends and predict likely future prices for the commodity. Use this to determine not only the best time to (re) source the category as well as using deep-dive analysis to determine the best strategy.

Optimize Supplier Relationships

Once you have optimized all of the awards based on historical data and business rules, you also have the optimal allocation by supplier. Once you have the optimized set of awards for each supplier, you can optimize the re-order schedule, shipping arrangements, and even production and sourcing schedules on behalf of the suppliers and take costs out one level down in the supplier chain. Helping your suppliers help you goes a long way to building good supplier relationships and increasing supplier performance.

Simultaneous Drill Across Multiple Data Sets

Once you have true federation, you want to split the screen and update the views to only contain the relevant data in each data set as you drill down through the data. Going back to our previous example, you start in the Payment cube drilling into the goods receipts associated with the wonky widgets, then switch to the Order History cube to find the initial requisitions, but when you drill on the user in the second cube, the first cube is updated to contain only those goods receipts associated with the user. The user can drill through either cube to find the data she wants, whichever is easiest, and both cubes update. She doesn’t have to go back and forth.

These are just a few more things that can be done, and all would simplify the life of an analyst. More to come at a later time but first, this time I’m going to insist that you tell me what you would do. :-;

If I Succeeded in Destroying Dashboards, How Else Would I Improve Spend Analysis.

The smart alecks are correct — technically destroying dashboards is not adding anything to spend analysis so I didn’t actually provide a way to improve spend analysis technology, just the results you get from using it.

So if I succeeded and dashboards bit the dust, what would I do? (Besides banning integration points for report writers for all OLAP-based spend analysis products?*) Good question. Especially since there’s about a half dozen logical next steps.

Three things that would be useful if you had a true spend analysis product like Opera’s BIQ would be to:

  • Integrate Easy Should-Cost Modelling CapabilityThis way you can define a cost breakdown for a product or service you are looking to source and have the tool automatically generate an expected cost based upon current data, as well as a price-range, with confidence, based upon low, average, and high prices paid for the raw materials, energy, labour, etc. (provided that the should-cost model permitted base-cost definitions for any cost components you weren’t buying that were bought entirely by your supplier)
  • Optimized Awards Based on Historical Data and Business RulesYou don’t have to send out an RFX to get base market pricing if you are already buying a product, it’s in your transaction store. Nor do you have to run a complex event to determine the lowest cost providers for a market basket. Moreover, if you are buying commodity products and services with list prices, and all your suppliers do is give you a discount of X% for a guaranteed award, you don’t really need optimization to determine the lowest cost as it’s just a simple formula against current pricing. And if your only business rule is 2 or 3 way split, it’s just the 2 or 3 lowest cost suppliers with the appropriate risk mitigation. In this situation, it would be easy for spend analysis tools to build in some simple optimization capability to tell you your lowest cost buy, and if it’s close to your should-cost model, you can just cut a contract without going through a time-consuming sourcing event.
  • True Federation across Related Data SetsMost spend analysis tools are only capable of working on one cube built on one data classification at a time. This means that even though a user can pick the drill dimension order, only one set of data can be viewed at one time. But sometimes you want to drill into greater detail (such as who requisitioned all those widgets from the wonky supplier), and that’s not in the transaction file — so you need another cube with more detail on the invoice (history). Then you drill in on the augmented AP (cube) data until you get to the invoices associated with the supplier, switch over to the new cube and drill down to the line items of interest and retrieve the requisitioners. Another situation is where you are getting a lot of warranty returns, and you want to figure out what batches the returned items are in so you can determine whether or not the batches were bad and it will be cheaper to do a mass replacement (by just putting out a recall) than dealing with one breakdown at a time. In this case, you need to drill into the warranty cube and then branch over into the invoice cube to get the batch numbers associated with the appropriate goods receipts that are associated with the invoice.

These are just a few things that can be done, and all would simplify the life of an analyst. More to come at a later time but first, what would you do?

* If you don’t know why, you don’t know your spend analysis product limitations!

In What Way Would I Improve Spend Analysis?

When it comes to spend analysis there is at least one particularly powerful tool out there that will meet the majority of the needs of any organization and probably at least one tool that will do, with elbow grease, just about any analysis an analyst can think of. Since businesses have wanted reports and analytics since the days of the first spreadsheets, analysis tools are always advancing and most are beyond the ability of the average user to fully utilize their functionality.

So, given this fact, how would I improve spend analysis? And given that this question may imply that I may only make one improvement, just what would that improvement be? Especially since most tools don’t do (true) federation, don’t support full reg-ex (regular expressions), don’t understand semantics, and don’t run fast enough on large data sets — indicating that, as a PhD in CS with deep expertise in analysis, modelling, optimization, and semantics, there are theoretically a number of advancements I could bring to the table if I put my mind to it?

Despite the plethora of options available, today there is only ONE thing I would do to improve spend analysis. I’d make it impossible to do anything but spend analysis. Specifically, I’d make it illegal to include dash-boarding capability in any (spend) analysis product.

Why would I do such a thing? Besides the fact that I’ve been ranting since 2007 that dashboards are dangerous and dysfunctional, I would do such a thing because, among other things, they give you a false sense of security that, if mismanaged, could be so grave that, like the myth of Nero, you would fiddle while the factory burned.

Why would I ditch the dashboards and make it a crime punishable by any fate one could devise that was worse than death to include any capability whatsoever designed to support a dashboard? Because I just read this post on Purchasing Insight on “the inordinate cost of poor spend analytics” that said that it’s reckoned that more than 50% of businesses employ between 2 and 5 people to prepare and create procurement dashboards and spend reports. This is ludicrous. (No, not Ludacris.) If these people are senior analysts, then a large organization is spending more than 500,000 a year on salary and overhead to create dangerous and dysfunctional dashboards that spit out shiny spend reports that, after being analyzed the first time for inefficiencies, provide zero value to the organization. Once the report is analyzed, the inefficiency identified, and the problem corrected, and once this is verified in the next report, no subsequent report is going to tell the analyst, or management, anything new.

As SI has said again and again, the value of spend analysis is actually doing spend analysis, again and again, testing new hypothesis every time they pop into the analyst’s head. Yes, most hypotheses will yield nothing, but that’s not important because it only takes one insight to yield 100,000 worth of savings. If the tool is flexible, powerful, and configured appropriately, the user will be able to explore dozens of different analyses in a week, and if even one yields 10,000 of savings, that’s an (amortized) ROI of (at least) 5X. Spend analysis is analysis. Not dashboards and reports.

So if you really want to improve spend analysis — ditch the dashboards and focus your talent on real analysis. Otherwise, just download a free reporting engine off the internet. You’ll get the same worthless result, without forking out six figures for a tool you’re not really using.