Applications of sampling

Data Science, AI and ML
#1

  1. Validating assumptions through market research
    Suppose your company wants to launch a bike ride-sharing service. This service relies on people having smartphones with sufficiently charged batteries and sufficient mobile data. Now you wish to evaluate the market size. To do so, you will have to get a sample that represents people from various income levels, mobility needs, access to data, type of devices, willingness to adopt the bike-sharing model, etc. By doing so, you can arrive at a reasonable estimate of the overall market size of your offering.

  2. Quality control
    Extensively used in the manufacturing industry. Suppose you wanted to check the quality of injections produced in a factory. Let’s say, the company produces 1 million injections a month. In this case, quality assurance becomes critical. However, it may not be possible to check each injection manufactured. So the company will sample a proportion from each batch and, based on the results, make an inference on the quality of the whole quality produced.

  3. Uses in new product development
    Suppose you are working on a new service, say a new bike-sharing service. The typical process with you will follow will involve four steps:

Concept creation & testing
Pilot testing
Beta Testing
Launch
In most of these stages, you would find good use of sampling techniques. Essentially, you want to draw inferences about the whole population by studying the responses of the sample. It becomes vital that you steer clear of any biases and under-representation of the population in your sample.

Concept testing: Before starting the development, you might want to know the appeal of such an offering. We can accomplish this by asking a few prospective users of such a service. However, a better approach would be to scientifically go about surveying the people. This way, you can ensure that you are getting representation from all groups, both those that are comfortable with newer modes of transport and those that are apprehensive. You might want to understand who much people are willing to spend on such a service. During interpretation of the findings, we can ensure that each stratum of the society finds representation in the sample and also there are sufficient people in each stratum.

This will lead to meaningful feedback and eliminate the scope of false confidence you might get if you say surveyed only people who are in the pro-sharing economy.

Pilot testing: This is the phase just before the beta launch, and you want to factor in as much feedback as possible. Here, using the same principles of testing, you can have useful feedback by ensuring that you have factored in cultural and behavioural patterns from your study, by using the sampling techniques.