Question

Topic: Research/Metrics

Sample Profile Conundrum

Posted by Anonymous on 12/19/2008 at 6:16 AM ET 250 Points
I am in bit of dilemma, my client recently did a contact study. Contact studies are usually used to help develop a sample profile for longitudinal studies.
The data from the contact study, seems to be contrary to other studies and a more recent study done for similar product.

It suggest that consumption of the product is amongst a far smaller proportion of the capital city (X) population than the alternative study does?. It seems that other regions interviewed in the contact study have higher consumption.

I am confused as to how I can handle this, and the client want to start a tracking study.

I need your experts advice on:
What are the likely issues regarding the sample profile
What options to the client are they
- Is X an anomaly
- Do we exclude X from the sample points
- Do we ignore X
- Do we replace X
What should I do about these issues?
Which Option presents the best solution and why?

I have to recommend approaches to the issues?
- for each one I mentioned above
- Also established the pros and cons of each one
- determine the favoured solution and why?

I would be grateful to your ideas and thoughts to help me with this sample profile problem.
Thanks in advance
To continue reading this question and the solution, sign up ... it's free!
SUBSCRIBE
Or, SIGN IN

RESPONSES

  • Posted by Candureactor on 12/19/2008 at 12:19 PM Accepted
    I'm not sure I can answer your questions specifically, but here are some questions that may help you with this problem:

    1. The data from the contact study (CS), seems to be contrary to other studies and a more recent study done for similar product.
    Q: Was the sample for the CS different that the other studies? If yes, this is a bias that may partially explain your findings.
    Q: Was the question exactly the same in all studies? If no, this is a bias that may partially explain your findings.
    Q: I assume that you were not expecting a different result -- what theories (beyond sample or survey bias) do you have to explain it? For example, a better/worse product or better/worse distribution/marketing in the stated region.

    2. Is X an anomaly?
    Q: What was the sample size of the CS? Can the difference be explained by sampling error? For example, a sample of 100 has an error of +/- 10%.
    Q: Was the methodology consistent? Various research methodologies can have impact on results.
    Q: Was data collection monitored and quality assured. Interviewers can play a large role in biasing results.

    3. Do we exclude/ignore X?
    Q: What impact does X have on your overall market success? I am guessing that it has quite a bit of an impact or else you wouldn't be asking. In that case, you cannot exclude or ignore as that would be irresponsible.
    Q: Have you ruled out all biases mentioned above? If yes, the you can accept the result or redo that region. That is of course more cost and more time for the client. Is it worth it, can you show ROI?

    4. Best options?
    a) accept the result as valid... based on previous findings/intuition this does not seem a good option.
    b) accept that some sort of bias has skewed your results.
    - You can redo if cost/time is not an issue.
    - You can "smooth" the results by blending with another test cell.
    - You can impute another value (average of other cities) and closely monitor the results for this cell in the longitudinal study.

    I'm sure there are other options but I hope this gives you a starting point to be able to form your opinion about what is best for your client.
BACK TO TOP

Post a Comment

Related Discussions

Community Info

Top 25 Experts
(Research/Metrics)