What is MRP? Besides the exciting new way to get reliable brand tracking. All your frequently asked questions are answered here.
Brands can’t get enough of MRP - and we don’t blame them! They are glad to finally have something on the market that provides them with reliable and accurate brand insights. However, they do also find the concept quite tricky to grasp.
A frequent questions asked is "what is MRP?". MRP is advanced data science but, because we don’t know many marketers who double up as data scientists, we wrote a comprehensive guide to MRP tracking. However, we thought it would also be useful to have an explanatory article focused on answering the top questions we receive.
We hope you find the answer to "what is MRP?" here. If not, don’t hesitate to send your question over to email@example.com. We’d be happy to provide an answer.
MRP, or to give it its full name, Multilevel Regression and Poststratification, is a form of advanced data science made popular by Professor Andrew Gelman. Professor Gelman first used it for election forecasts, while Latana used MRP for brand tracking.
MRP creates a model and uses this model to generate estimates for responses in a survey. This model, when given a set of respondent characteristics, can produce an estimate for how a certain respondent would answer a survey question.
Following that, MRP organizes the respondent’s characteristics into groups. By doing so, they can better capture how the variables interact in real life.
Finally, MRP takes weighted averages of all the predictions. This is to ensure that the model has a fair sample of respondents.
Traditional brand trackers are unable to accurately measure opinion in small target audiences. This is because they narrow in on specific respondents within the target audience. Unfortunately, this method of narrowing in results in a large margin of error.
The MRP method does the opposite and achieves precision even in niche audiences. It does this by not restricting itself to the small number of respondents. Instead, it uses information from the entire sample to create a model that can predict brand awareness based on a respondent’s characteristics. Therefore, it can provide reliable insights with higher precision and a smaller margin of error.
This is another nice thing about MRP, it provides direct information regarding the number of samples needed to get good estimates. The logic is rather simple: we start drawing samples and once our model finds the effects of all characteristics, we are done. This notion of 'convergence' can be tracked dynamically during fieldwork and allows us to achieve much better estimates at much lower sample sizes compared to classical quota sampling.
MRP uses a Bayesian model to predict brand awareness based on a respondent’s characteristics. The Bayesian framework gives us an advantage in that we get a measure of the uncertainty of our estimate for free. These Bayesian ‘error bounds’ get smaller the more information we provide to our model e.g. by including prior information from the past or larger sample sizes.
A MRP model needs a certain amount of information to find the effects of characteristics in the population. For hard to reach audiences e.g. people who know a rather new brand, it is hard to get enough information during the first sampling waves. However, our model will accumulate information over time and after a couple of months, it will have enough information to tell us the difference in brand associations, even for small brands.
As mentioned above, MRP is a Bayesian framework which always comes with an estimate of uncertainty in our prediction. This really helps with detecting changes over time since it allows us to make statements like ‘With a probability of 83%, there was a change in brand awareness from between March and April’.
Strong YES! Since the MRP model learns over time, brand KPIs with only a little information at the beginning will improve significantly.
We hope you found the answer to "what is MRP?" here. If not, remember we would be happy to provide an answer over at firstname.lastname@example.org.