Monday 3 July 2017

Sampling in Evaluation Studies

When one wants to know the exact number of defective pieces a box may contain, we need to test and count them all. However, when in comes to large quantities, sampling works the best.

Many times HR professionals question how do we do sampling for our training programs. 

Case in Point
Name of the training: Leadership training
Number of people to be trained: 90
Departments: All
Estimated cost: Rs. 75 lakhs

Rs. 75 Lakhs is a major investment. For all our ROI studies and before investing so much money we suggest our clients, to do a pilot test and take a sample before deciding if the training would give any ROI. Following are the steps that we take1.    
  1. Decision to do a pilot test.
  2. Identify a maximum of 10 people for the pilot.
  3. Out of the 10 people, one person must represent each department. So we have a heterogeneous group.
  4.  Identify at least two impact measures per participant.
  5. Conduct the training with action plan documents to be submitted post one month.
  6. Measure reaction and learning.
  7. Make the necessary changes in the content based on the reaction and learning feedback.
  8. Wait for a month.
  9. Conduct focus group/ interview the participants involved.
  10. Check if impact measures match the action plan documents.
  11. Forecast training impact.
  12. If the forecast is in the positive direction, launch the training for the remaining 80 people.


Please note one need not take the sample of 30 people that is considered a minimum required in statistics. When it comes to training, 10 people are a good representation to check the impact of the training.

For all our impact studies, we emphasize the need to do a pilot and check a sample. Unfortunately, this is overlooked on account of the following reasons
  1. The urgent need to conduct training, so there is no time to wait and do a pilot.
  2. There is a mandate from the top management to launch training immediately.
  3. HR people are co-coordinating for various things and hence need for sample test is not felt.
  4. Post-pilot, what if we realize that the training is not giving the impact then how do we deal with it.
  5. We have to use the sanctioned budget.
  6. We have developed the content so 10 people undergo the training or 60 people, in any case, we have invested in the content.
  7. We are short-staffed to keep doing pilot testing.


There is a need to go beyond our traditional way we conduct training and not just experiment with the training format (gaming, activities, and simulation) but also the way we execute the end-to-end training process.

Happy Sampling
Sonali