Aware of the Rutgers University study estimating that $7.5 to $16 billion was being wasted every year on soft skills training provided to the wrong people, the senior leadership of an agricultural products company wanted to optimize their training budget. Up until then, they had operated with the typical "spray and pray" approach of providing training to all employees, and hoping that it would "stick" with those who needed it most. Faced with a potential bill of $3 million to provide their sales force with a new training, they wanted to measure whether or not the training had an affect on sales. They were also anticipating a "down" year in total sales, and wanted to be sure that the overall sales cycle didn't affect the study.

First, they decided to pilot the training by rolling it out to half of their 2000+ sales force, and then they contacted us about carrying out an analytics study. If they had contacted us first, we could have saved them approximately $1 million right off the bat by piloting the training with a much smaller subgroup of their reps (and we didn't know it yet, but the results later indicated that this would have approximately tripled the ROI of the training!).

In this case, setting up the study to yield a true financial ROI was fairly straightforward, since the trainees were sales staff, and the company had already established a clear formula for translating sales revenue to profit. We were fairly limited in how we could set the study up, since the training had already been conducted, but the company could provide the three necessary elements for an analytics study:

• an intervention to analyze (the sales training),
• a group of participants who had received the intervention (the sales training) as well as a control group (who had not undergone the training)
• the metric to be impacted by the intervention (sales revenue)

Since we wanted to not only answer the question "does this sales training work" but also be able to answer related questions like "for whom does it work?" we collected two additional data points on each of the participants that could lead to differences in the effect of the training: what region they were based in and their sales performance in the previous year (expressed as rankings in quintiles: the top 20% were the first quintile, the next 20% were the second quintile, etc.) Again, if we had been contacted before the pilot study had been run, we would have suggested other potentially relevant characteristics of the participants and could have sharpened the study results even further.

It was also easy to address their concern that the overall sales numbers wouldn't confound the results of the study. Our statistical model is robust enough that all we needed was a control group and enough participants to study each of the relevant variables, and we had more than enough participants to maintain a very high level of statistical certainty.

All that was left do to, then, was run the statistical analysis, and the results were very clear and very rewarding. Even though the company did indeed experience a down year, overall, in sales, the study indicated that the training contributed $23 million in sales. That is, the sales reps who received the training had cumulative sales numbers that were $23 million greater than they would have if they performed at the same average as the control group. With a total training cost of $1.5 million, this yielded additional profits to the company of just under $3 million, or an ROI of 95% for that year. The study confirmed that the training was definitely worth the investment.

But it gets better: because we had done the further study to ask the questions about "who specifically" benefited from the training, we were able to make a remarkable discovery: 90% of the benefit of additional sales revenue was generated by the sales reps in the lower three quintiles of prior performance! The company had assumed that the higher volume sales reps would benefit more, and had stacked more of them into the pilot study, but the top two quintiles showed no effect at all from the training. The analytics study indicated that they could have cut two thirds of the participants from the training and still yielded virtually the same results! The following year was the true proof of the concept. The study predicted that the company could generate another $25 million in additional sales the following year by continuing to roll the sales training out. The senior team followed our recommendations and did so only to the lower three quintiles in sales performance. They saved $1 million right off the top by not training everyone, and the sales reports for the following year substantiated an additional $28 million in sales.