Today’s post is by Eric Strovink of BIQ.
As an engineer who originally entered the supply management space in 2001 to build a new spend analysis system, over the last 9 years I’ve watched marketing departments consistently “dumb down” the original broad and exciting definition of spend analysis that I remember from those days, to something really quite ordinary. For example, here are the steps required for classic data warehousing:
- Define a database schema and a set of standard reports (once, or rarely)
- Gather and transform data such that it matches the schema
- Load the transformed data into the database
- Publish to the user base
- Repeat steps 2-4 for life of warehouse
And here are the steps required for what has come to be termed “spend analysis”:
- Define a database schema and a set of standard reports (once, or rarely)
- Gather and transform data such that it matches the schema
- Load the transformed data into the database
- Group and map the data via a rules engine
- Publish to the user base
- Repeat steps 2-5 for life of warehouse
Not much difference.
You might ask, how can spend analysis vendors compete with each other, when the steps are so simple, and when commodity technologies such as commercial OLAP databases, commercial OLAP viewers, and commercial OLAP reporting engines can be brought to bear on any data warehouse? Well, it’s been tough, and it’s especially tough now that ERP vendors are joining the fun, but they compete in several ways:
- Our step 4 is better [than those other guys’ step 4].
- [briefly, until it failed the laugh test] Our static reports are so insightful that you don’t even need anyone on staff any more.
- [suite vendors’ (tired) mantra] “Integration” with other modules
- “Enrichment” of the spend dataset with MWBE data, supplier scoring on various criteria, and any other ways that might exist to try to add checklist features for analysts that may broaden interest in the spend analysis dataset beyond simple visibility.
It’s all very discouraging, but the doctor and I will continue to point out that spend analysis is not just A/P analysis; it can’t be done with just one dataset; and it’s not a set of static reports or a dopey dashboard, even though some vendors and IT departments would like to think it is. Spend analysis is a data analysis problem just like any other data analysis problem, and it requires extensible and user-friendly tools that empower people to explore their data for opportunities without third-party assistance. Those data come from multiple sources, not just the A/P system; many datasets will need to be built and analyzed; and from them, hugely important lessons will be learned.
The above notwithstanding, building a single A/P spend cube is a useful exercise. If you’ve never done it before, you will find things that will save you money. But that’s just the tip of the iceberg.
Previous: Analytics IV: OLAP: The Imperfect Answer
Next: Analytics VI: Conclusion