Today's database access tools, with their powerful GUI interfaces, provide marketers with the hands-on ability to access, manipulate and draw conclusions from their own data. And, that has been a wonderful development! However, as we discussed in the August 6, 2007 e-Letter, these access tools also have great potential for misuse when they fall into the hands of individuals with little or no analytical training. The result can be false insight, which is worse than no insight at all.
In the August e-Letter, we noted that marketers should think twice before tackling data investigation that requires the definition, creation, manipulation and comparison of multiple past-point-in-time views. However, the unwary can also get into trouble with database queries that have nothing to do with multiple past-point-in-time views. Although the following example involves a business-to-business scenario, the concepts are universal:
The goal is to calculate re-order rates for certain types of consumable merchandise, so we can understand when to send targeted promotions. Intelligent discussion will raise important issues that will make the difference between true versus false insight:
- The average re-order rate will be misleading if the variance is excessive. Therefore, distributions should be calculated, in the search for an actionable critical mass of re-order behavior.
- It will be important to determine if there are overlapping re-order distributions that cluster around multiple critical masses. For example, if small businesses re-order every five months on average, and large businesses every month, then sending a promotion corresponding to the overall average of three months will be perfectly timed to interest almost no one.
- Extending this line of thinking, we should take care to find out if re-order rates vary by other factors besides how large the business is, such as the size of previous orders. Calculating the ratio of the two might prove insightful. Also, seasonality must be considered.
- Single-buyers should be eliminated from the analysis because, by definition, they have no re-order rate. Including them will guarantee false insight. But, perhaps there is something we can do so they can be included in the promotion. Also, how should residential customers be handled?
- Finally, what about production considerations? If, for example, postcards will be employed to generate additional demand, how will the drops tie into the existing production schedule?
I learned this last lesson earlier in my career, after analysis indicated that a specific line of highly-profitable merchandise had high re-order rates. So, I recommended aggressive cross-selling and creative line extensions. The problem was that the specific merchandise line was manufactured in-house, and required custom-built equipment that was very expensive to manufacture and would require additional factory floor space that was not available at the time.
