Exploring old warehouse data can help you spot unseen influences on your business processes. Here are some methods you can use to find solutions for improving business.
Data warehouse information can unveil hidden connections between business and environmental events, influencing the success or failure of business processes. Graphing chosen factors over time and comparing them can yield valuable business intelligence for your client. Here's how you can use your organization's database to find real solutions to improve business.
Exploring data warehouses
Here’s a quick review of the important points in exploring warehouse data:
- Have a process-oriented wish list in mind before you start; know where your operation needs improving, and focus your search on factors you know to be relevant.
- Select a key item—market share, manufacturing costs, advertising costs, turns per quarter—for graphing over a selected number of time periods, measuring fluctuations in meaningful increments.
- Check your search setup by graphing your key item’s changes over time, then graphing some variable known to influence that key item, and comparing them. Effective guesses on time intervals, number of periods, and correct unit of measure of the key item’s fluctuations are critical to the meaningfulness of your results. If it’s off, work with these variables until you see results. Your primary goal in these searches is to graph variables that rise and fall, or otherwise change in tandem with one another, indicating a possible relationship between them.
Now we’ll look at more specific methods of analysis.
What patterns are interesting?
If you graph total regional sales for one of your client’s products across several quarters, and then graph the occurrences of promotions in that region for that product, you could reasonably expect to see periodic increases in sales corresponding in time to those periods of promotion. This is a simple and obvious pattern—key item and variable fluctuate together over time.
But there are other patterns that can imply a relationship between your key item and obscure variables. These patterns include:
- Key item and variable separate widely, but then tend to converse at a uniform rate.
- Key item and variable separate widely, but then tend to diverge at a uniform rate.
- Key item and variable fluctuate at different rates and over different intervals, but show simultaneous “spikes”—sudden divergence from their normal graphs.
- Key item and variable fluctuate at different rates and over different intervals, but show spikes offset from one another by short time intervals (i.e., when one variable spikes, the other will invariably also spike shortly thereafter).
- Temporary periods, either sporadic or irregular, of in-tandem fluctuation between key item and variable, indicating a possible third factor’s influence.
There are other patterns that can indicate a relationship between events and conditions, and your client will become adept at spotting them. But these patterns will get things rolling.
Remember the company business models
There are invariably some ideal models that senior management uses to guide business processes. Such models, if they are correctly constructed and sufficiently detailed, will demonstrate patterns of their own—expected fluctuations in most areas of operation, within most every business process, and at almost every level of detail. That’s the nature of business, the character of a surrounding economy, and competent, realistic management will expect adherence to models that reflect this.
How can you put this to use? The modeling of processes and influences on a company’s performance is a tricky but necessary undertaking. You can use the techniques of warehouse exploration and analysis to add useful detail and extract corrective intelligence for the company’s business models. Of course, the data warehouse exists to facilitate the use of analytics, which yield metrics to check company performance against its business models at various levels. But these analytics and metrics, by definition, measure only relationships that are already known and understood.
Consider that models never contain everything. If they did, the unexpected influences on your business process would be expected, which is clearly impossible. The emergence of hidden relationships between variables in your data exploration represents divergences from, or enhancements of, your business model. By seeking out and documenting these influences, senior management can add useful detail to the company business models.
You can also use data warehouse exploration and analysis to generate models for senior management, if no models exist or if they're insufficient. This involves a major commitment of time and effort but could well be worth it.
Beauty in chaos
It's a staple of chaos theory that if you stare into chaos long enough, patterns, sometimes beautiful ones, will begin to emerge. It’s your task to spot them.
There are two primary results to be seen in the corresponding variable behaviors you’re going to dig out. The first is a causal relationship. In such a relationship, the unknown variable you’ve compared to your key item is a direct causal influence on the key item’s fluctuation. An example of this is the dramatic increase in turkey sales that immediately precedes Thanksgiving.
The second result is a correlative relationship. This is a relationship where neither factor necessarily influences the other, but one always accompanies the other. An example of this is the observation that lipstick tends to be purchased by females. Either of these relationships, when discovered, may be used to your company's competitive advantage.
Turning your results into client profitability
The goal is, of course, to turn the discovery of previously unseen influences into strategic advantage. To this end, you need to classify your results:
- Factors you can control—If you discover that sales for Product B rise when a Product A promotion is in effect, implying some subconscious association in the mind of a typical customer, then marketing strategies for both products can be combined to optimize the advertising dollar.
- Factors you can’t control but can anticipate—You notice that certain diet foods increase in sales with the advent of the new year and the beginning of swimsuit season. But sales also increase slightly in the weeks after the Grammies and the Oscars. Marketing people can exploit this trend, increasing sales.
- Factors your client can neither control nor anticipate—If you're a publisher and your exploration uncovers that the residual sales of a particular presidential memoir tend to rise somewhat whenever a national political scandal erupts, you can still take constructive action. Hastening reprints of this book at the first sign of such a scandal boosts product availability and subsequent sales.
A final word
While the point of this style of data analysis is to put cold, hard science underneath business process re-engineering, it's not to besmirch the executive’s instincts. While you’re handling an objective analytical tool for digging out some extra efficiency, you’re also putting a resource into management’s hands that is ultimately intuitive. As you work with these techniques and refine them, you’ll find yourself developing a sixth sense, sniffing out the unexpected.