Random Sampling allows you to specify and target an interaction to only a sample percentage of eligible customers (e.g. “I only want 5% of all my customers that meet the targeting criteria to be able to see this survey”).
Below will detail how to use the feature, and this article will contain some common use cases.
How To Use
Within your WHO targeting segment, this targeting option is listed as “Random Sampling”. You are able to add this rule on all interactions.
Once you have selected that rule, you can then specify from all eligible customers that met the criteria for that segment - what random percentage of them can see the interaction while getting responses from a representative cross-section of customers.
You have the flexibility to specify a different random sample percentage for each segment within an interaction, or set them all the same in order to only target a percentage of your entire interaction population.
How Does Random Sampling Work
Alchemer Digital randomly chooses a number between 1 and 100 and then assigns that number to your survey/interaction based on the conversation id. It then looks at the percent defined to see if the assigned number is less than the percentage chosen. It is random each time, just like a coin flip: you may have 56 heads and 44 tails over 100 flips. The percentages would increase in accuracy over a larger sample of responses (1000 responses would be a closer distribution than 10 responses).
Changing Random Sampling on a Live Interaction
When the random sampling percentage is changed, the audience for the survey will include the previously considered audience.
Example:
Sampling ID:
id1
, Customer-defined sampling percentage: 20%Assume the SDK-generated random number as 15
Since 15 < 20, the user is selected and their responses are collected.
If the customer later updates the sampling
id1
to 10%:The SDK random number (15) remains unchanged.
Since 15 > 10, the survey will no longer be shown to this customer.
However, any previously collected responses are retained.
If the sampling percentage of
id1
is then updated to 25%:The stored random number (15) is again evaluated.
Since 15 < 25, the survey becomes eligible to be shown again, depending on other criteria.
Previously collected responses are preserved.
The random number is assigned based on the survey, not the segments in the survey. If you have multiple segments in the survey based on random sampling, the same generated random number is applied to both segments.