Sample blending (using sample from multiple sources) broadens your reach while collecting data, and allows you to get more information faster because answers are coming from multiple sources instead of one, overall increasing the quality of your study. The concept of blending keeps the sample mix consistent over time because you are able to more accurately represent and monitor a specific target audience, making your data reliable. Get the most out of sample blending with these tips:
1. Use blended samples on the appropriate study type
Not all panels are created equal. Quality and sample compositions vary in behaviors and attitudes towards brands, tech, gaming, shopping, or media consumption, to name a few. Representation from multiple panels can help prevent bias because you are considering the different details in the composition of each panel, and using that to create the perfect blend to represent your target audience. Therefore, blended sample is most appropriate for ad hoc and new wave/tracker studies which aim to measure trends, brand consensus, or consumer behavior. A blended sample will help achieve the most representative data for the targeted population.
With a blended sample the data is more consistent because not one supplier is sending a majority of respondents compared to another supplier. Because the majority of the population spends their time on social media, this is a new viable resource to blend panelists from online panels with traditional panelists. The behaviors and opinions of online consumers are changing and are not always reflected in a traditional panel, so by incorporating online respondents you will be able to more consistently measure brand opinions over time.
2. Ensure quality by limiting the percentage of respondents each supplier sends
When using a blended sample, there is a concern that data quality will decrease. One of the easiest ways to control quality is to limit the percentage of respondents each individual supplier sends. A best practice is to keep the blend conservative and not to have any individual partner provide more than 25-30% of respondents to the study. This reduces dependency on a single partner and avoids causing the data to skew. Different suppliers panels or online respondents have different profiles, and by having one specific group send a majority of the sample the data might skew. With a diverse amount of suppliers the population you are targeting can be better reflected in your data.
Another reason quality comes into question when blending is the belief that respondents who are not from a traditional panel are not screened the same way or as thoroughly. However, this should not be a concern, because when done correctly, online respondents can be screened in real-time, and blending will still deliver high-quality results.
3. Monitor incentive structure of panels
Each sample source uses a different incentive structure to motivate their respondents. This can be through cash rewards, a charitable donation, online rewards, or other incentives. Monitoring incentives is necessary because past studies have shown that different incentive methods produce different levels of motivation among panelists, and certain types of motivators have respondents who are more prone to speeding through the survey and providing poor responses. With blending be sure to monitor how different panel suppliers incentivise their panelists in reference to the study you are running to make sure that you aren’t introducing bias regarding awareness of a specific brand or product. For example, if you are trying to measure opinions about a certain brand or product, you would not want panelists to be incentivised to take surveys with that same product.
4. Be wary of the incidence rate on wave/tracker studies
When conducting a wave/tracker study with a low IR the sample blend has to be more flexible because fewer people fit the qualifications you are testing for. For surveys with a high IR, it is easier to have a more consistent blended supply of respondents. Depending on the IR and any applicable exclusions, a small blend variance allows wiggle room for difficult-to-reach audiences.
Remaining within an allotted blend variance is important because respondents from different sources can exhibit different characteristics. For example, respondents from social media sites are more likely to be gamers, but respondents from more traditional panels tend to be older women. So if you’re running a video game project and you have 100% social media sourced respondents one month than 100% traditional panel respondents the next month, your first month will likely see a high awareness level of a video game, while the next month your respondents will see a much lower awareness level. But, when you have a very low IR project, the likelihood that the characteristics of said respondents will change from source to source is less likely.
