Why Predictvia for political research?

Predictive Analytics + Dynamic-Adaptive Sampling:

  • Fast: We have the data sources and technological process in place to deliver results about voting tendencies and public opinion in few days, sometimes even the next day.
  • Cost-efficient.
  • Laser targeted: We can reach small/remote regions and very specific locations as well.
  • Precise: We consistently beat traditional pollsters in terms of final, observed sampling error.

Download some of our success cases here.

Polling through Internet: Can it be efficiently done?

Internet population is different than general population or from the electorate, so Internet-based polling has been naturally questioned. Here is how Predictvia overcomes that obstacle:

  • Predictvia “slices” the sample, during its design, in hundreds or even thousands of different strata.
  • Such micro-segmentation allows us to focus on obtaining sample even in those sub-segments where Internet penetration is very low.
  • We then use targeting capabilities (mostly Google’s, but also Facebook’s and others’) to be able to reach all of those sub-segments and obtain responses.
  • Finally, our Dynamic Adaptive Sampling (DAS) technology, determines if the responses from every stratum are “homogeneous” enough or there is need for more responses according to a quality criteria. (DAS is explained in more detail below).

In the end, our methodology adjusts the analysis of data according to the characteristics of the general population (or the electorate), so as to obtain a representative picture of what the population (or electorate) is thinking.

To be clear: results derived from our methodology are as susceptible to sampling error as those from any other methodology; however, our methodology has proven to be effective at “skipping” the so-called Internet bias.

Why Predictvia can do it?

Dynamic Adaptive Sampling (DAS):

DAS is a Predictvia technology that evaluates in semi-real time how good a sample is and, based on such evaluation, make automatic decisions about whether to stop or continue a data collection process. These are DAS key facts:

  • Dynamic Adaptive Sampling (DAS) is an intelligent sampling technique that optimizes a sample design as the data is being collected.
  • DAS is capable of evaluating hundreds and even thousands of sampling strata (crossings between key research variables).
  • The method analyzes data from every stratum of a sample design and then decides if it needs more sample, according to how “uniform” the whole strata set is.
  • Since this analysis occurs simultaneously with the surveying process, DAS is able to “ask” for more filled-questionnaires on an as-needed basis. It only stops when its quality criteria are satisfied.
  • Therefore, this technique improves the sampling process by reducing collection time, sample size and cost.
  • But more importantly, DAS significantly improves the chance of having an efficient sample by the end of the data collection process.

Google Consumer Surveys (GCS) partnership:

Predictvia is one of only six companies worldwide with total access to advanced features of Google Consumer Surveys (GCS), the powerful Google survey platform.

Moreover, GCS’s and Predictvia’s technical teams continuously work together to develop integration of both companies’ platforms to such an extent that technologies like Predictvia’s DAS become seamless.

Some of the capabilities we can enable through GCS features are:

  • Target small geographical regions (such as small towns and villages) to poll potential local voters.
  • Dynamically generate panels of people who have a characteristic (for example: X political party leaning).
  • Target potential voters based on their occupation.
  • Target potential voters identified among the traffic of a Client’s website (for example: a political party website).

Google Consumer Surveys is extensively used among political pollsters in highly developed markets. Here is a reference of its effectiveness from the accredited New York Times' FiveThirtyEight blog (Nate Silver):


Be Frank application:

Be Frank (https://www.facebook.com/befrank00) is a survey application developed by Predictvia:

  • Manage the distribution of polling questionnaires to micro-clusters or specific social media users.
  • Gather data about the Facebook and Twitter profiles of respondents.
    • Such data is anonymized and then used to train mathematical models of potential voters who have a specific characteristic.
    • Those models, in turn, can be used to target and reach people that, predictably, have the same characteristic.

Option of service:

  • Regular (daily, weekly, monthly) polling, mounting up to election’s day.
  • Geo-political reach at multiple levels: nation, state (province), city, town, small town, neighborhoods.
  • Multiple segments: from voters in general to people who have a given political leaning, we can target and collect data from different segments, as defined by the Client.

General process and deliverable:

  • Sample design.
  • Modeling (optional).
  • Questionnaire design.
  • Data collection: polling.
  • Dynamic Adaptive Sampling (DAS): Explained above in detail. During execution of the polling process, the initial sample design is “intelligently” modified by DAS in its search for more efficiency.
  • Quality assurance: Responses are eliminated if they are flagged as “illegal” by a quality control process. Some instances of “illegal” responses are: repeated participation and bot-like behavior.
  • Raw data: Clients receive data in a structured format, in case they want to run their own analysis of the collected responses.
  • Interactive dashboard: Data is processed and analyzed as it’s received. The results are displayed on an interactive dashboard which allows for an easy navigation through the results. Importantly, the user of this dashboard can generate crossed-visualizations as required. Example below:

Contact Us