Know-How Exchange : Research/Metrics

Hi Omar,

You can use different sampling techniques depending on the nature of your customers and your service level:
1) random sampling is appropriate when you believe all customers are alike and your service level is consistent
Practically, with a number assigned to each customer, you just draw random numbers
2) stratified sampling is appropriate when there are distinct customer groups/segments with different service requirements and/or satisfaction issues. Company size (large versus small and medium enterprises) or industry may be intuitive segment criteria. If this is the case, i would make sure to have a sufficient sample size (see below) for each segment. Practically, you divide your customers in these groups and randomly draw from each group
3) likewise, if you believe service levels may be different at some times because e.g. most customers come 12-1, exactly when most employees are on a lunch break, make sure to sample enough customers from the 'peak' and the 'off peak' group. Retailers have long realized that consumers have different priorities 'off-holiday' versus the Thanksgiving-Christmas period and offer faster, stripped-down service during the holidays.

The required sample size can be determined with a statistical formula (e.g. https://www.surveysystem.com/sscalc.htm) once you decide on the confidence interval, i.e. the range around the sample answer in which you are 95% confident that the true answer falls. For instance, with 300 daily customers, a confidence level of 95% and a confidence interval of 10, the sample size is 73. When this sample reports an average satisfaction of 81, you can be 95% confident that the true average satisfaction of all your customers is between 71 and 91. Of course, the daily number of customers may not be the population of concern; if you want to sample your weekly population of about 1500, you only need 90 customers to achieve the same confidence level and interval. You may even take all your customers over the last year (60,000?) as the population, in which case you can tighten the confidence interval to e.g. 5, which requires sampling 382 customers.

Remember, depending on your answer to my first point, you want to achieve this sample size in every group you believe may have different satisfaction issues!

Cheers

Omar - To the advice you've already received, I woud add two things:

First, you should be sure to identify respondents, or "tag" them somehow so that you can properly segment. This can be done by either embedding a code in their survey response or by sending them different versions of the survey that are identified by a segmentation code. Also you can ask them deliberate questions so that you both a) know how to segment them, and b) can make sure your intended segmentation scheme "fits" with the way in which they self-classify.

Second, make sure that the way you segment fits in with an overall marketing strategy that you can executive effectively.

Good luck!

Omar,

Sounds to me like you have two distinct revenue streams: issuing & renewing licenses.

In which case, you'd need to separate the two streams (and thereby obtain different customer sets) by a basic parameter that is within your control: No. of licenses either issued (fresh) or renewed.

Now, you have two distinct set of people, on which to focus further. Do you have a complaints/suggestion box? If so, how is it structured? If you capture name/time & date....you have three more parameters that you can apply to the two sets already obtained.

After which you can create actual customer segments (both fresh/renewed) like: Green/Yellow/Red (like a traffic light...indicating need for attention). You can take a sub-sample from each of these created segments and contact them for further inputs into various factors & attributes related to waiting/servicing time, competency levels of staff etc. This sub-sample can be M/F:: 50:50 as starters. If you want to be sure that this represents the 'universe' that you are serving, then you'd need to weigh this information back to the total universe (M:F) that you've served under each stream.

If you do not have a suggestion form, then you need to see the total number of people served under (let's say) 'Fresh licenses' issued. I'm sure you'll have some basic demographic information, since they are being given licenses!!

Use demographic info, like: Sex/Address (to gauge location)/Income/Education etc. and tabulate this in excel. Let's say you have served 1000 customers in week 1, of these 1000, 250 are female. Of these 250 female, 25 are between 30-35, of these 25, ten of them stay in area, and so on and so forth....you need to randomly pick 10-15% of each 'type' within a period of review...add it up...and Voila! There's your sample (Of course you'd need to do it among males as well!)

I agree with Irene, you need to set aside some quality time and review your budget if you are doing this on your own...better option would be to engage a third party MR supplier to evaluate and create a proposal for you.

Hope this helps!