More Straightforward Analytics: Defining the Objective
by David Loshin
Originally published March 24, 2011
In last month’s article, I suggested some high-level practical steps for introducing a framework for collecting data, analyzing the data, and then applying the analytical results by changing business processes. This month, I’d like to dive deeper into an area that will ultimately blend aspects of performance, technology, and information together. Basically, I want to start with my first step in process preparation, namely “defining the objective.” I pointed out that in order to improve a business process, you had to have more than just an understanding of the process – you also had to select a specific value driver along with a specific method.
As an example, let’s look at reducing spending by selecting the vendor who provides the optimal price for the most frequently purchased categories of items. This objective seems to meet our criteria since it refers to a specific value driver (“reducing spending”) along with a specific method (“by selecting the vendor who provides the best price for the most frequently purchased categories of items”).
Actually, this provides a good template for generally framing the definition of a key business objective for analytics:
Do you know what the value drivers are within your organization? Most frequently, they roll up to high-level concepts such as increased revenue, decreased costs, increased profitability, increased productivity, and reduced risk. The opportunity is understanding how each of these high-level drivers is perceived within the organization, who the point people are in the organization for those value drivers, and most importantly, how those drivers are manifested within specific areas of the business, or even which business processes (and business process participants) are motivated or incentivized by those drivers. By the time you get into the weeds, there may be measures that are taken that provide little insight into their use, yet as they are combined and rolled up with other metrics, provide an aggregate view of organizational performance.
Next, you’d have to get a handle on both the existing and the needed metrics that measure performance for any specific value driver, especially at the business process level. Knowing the metric and getting its measure establishes the baseline for performance, and also guides the specification of a performance objective. Going back to our example, reducing spend might be related to lowering the average price for some number of key supply items. Knowing that you want those prices lowered leads to the third aspect of the template, optimization.
An optimization is a process change that exploits an existing process gap, thereby leading to a performance improvement. Going back to our example, we know we want to lower the average price for key supply items. Determining the optimization means evaluating what variables under your control are impacting product prices and then figuring out how to adjust those variables accordingly. For lowering the average price, this might suggest a few approaches, such as:
Even though these are only three suggestions, note that each of the three involves a process change. In essence, even before you begin the data modeling, integration, and application development, you have already identified the area of the business that is impacted and you’ve provided approaches for exploiting the analytical result.
Basically, you are preparing yourselves for the next stages in the straightforward analytics process, so that you can answer these types of questions:
So in preparation for transitioning to the next step of establishing measures, review your objective to understand the explicit and implicit data dependencies early in the process. To go back to our example, the specified objective sounds good, but requires some additional consideration about feasibility of achievement, especially in relation to the data concepts referenced. This objective explicitly involves two master data concepts (“vendors” and “items”), three quantification measures (“current spend,” “item price” and “item purchase frequency”), and one hierarchy (“categories of items”). There are also some implicit data concepts as well: another master data concept (“internal groups with budget authority”), more hierarchies (“vendors and their corresponding ownership chains,” “internal groups with budget authority”), and at least one other quantification (“spend measured by internal group with budget authority”). This will help you to recognize if there are any potential data hurdles to be jumped before investing the time and budget in building the end-user application.
Recent articles by David Loshin
Copyright 2004 — 2020. Powell Media, LLC. All rights reserved.
BeyeNETWORK™ is a trademark of Powell Media, LLC