Data Collaboration Platform Playbook On-Demand Get Yours Now

How to Judge Data Quality in 5 Steps

How to Judge Data Quality in 5 Steps

You would be forgiven as a programmatic trader, searching through hundreds of different data segments from dozens of data vendors within your various DSP’s, for thinking that all third-party data is the same. However, understanding the differences between data providers and the segmentation they offer to the market can prove critical in ensuring the best performance for a given campaign KPI. To help guide you in making sense of this data deluge, here are 5 areas you should focus on, and questions you should ask, when selecting a data provider.

Data Collection

Perhaps the most basic, but most important question of all, is how a particular data vendor collects their data. Does the data come through web properties? Through mobile apps? Is it offline census data, survey data (online or offline), or something else? Does the data provider have exclusivity on the data they collect, or do they have any specific partners and contributors that they are able to disclose transparently? Is there anything unique about the data collection methodology that the data vendor has?

There are also questions of compliance — is the data provider OBA compliant? Do they work with any self-regulatory bodies? Do they hold membership with industry bodies such as the IAB?

Data Segmentation & Modeling

Once raw data is collected, the next question is how is the data bucketed into market-ready segments. Does the data provider use recency and frequency of action? Are they working only with declared or demonstrated behaviors?

Importantly, does the data vendor extrapolate data segments through building look-alike and act-alike models from seed sets? If so, how are they performing this modeling, at what rate are they extrapolating the segmentation? How frequently are the models updated?

Data Onboarding Methodology

If the data is collected from somewhere other than a digital medium, it will need to be onboarded, meaning mapped to digital device IDs in order to make it available within the programmatic ecosystem. This onboarding process may be different from one provider to another. Some use lat-long data or zipcodes, others use deterministic (PII) keys such as email address or phone number. These methodologies differ in accuracy and the scalability they are able to achieve. Understanding how offline segments have been onboarded is a key consideration in understanding the accuracy of the data segments.

Data Scale

Data scale is an important consideration, particularly for those looking to deliver high-volume campaigns within more niche environments where data volume and density tend to be low, such as mobile data or across smaller markets such as MENA, APAC, or Eastern Europe.

The majority of data vendors do not break down segmentation by market, making it difficult to forecast availability prior to launch. However, some do break down taxonomies by market and provide specific data sets for each region.

Similarly, with mobile, there is often limited availability of data segments within a mobile environment, due to the nature of 3rd-party tracking within certain mobile operating systems. If you are planning to activate your data on mobile, you will need to understand mobile-specific data volumes as well as segments available.

Data Quality Initiatives

The discussion around data quality and data fraud cannot be avoided in 2017. Reliance on cookie data leaves data vendors open to challenges on data accuracy due to false input, device sharing, and bot traffic. It is important to understand what measures data vendors are taking to mitigate the risks against data quality in order to ensure accuracy and efficacy.

Asking tough questions of your data provider can be awkward, and some vendors may even get defensive with the questions you ask. But for the most part, they should be used to these conversations. It is up to you, the buyer, to make sure you are investing in data that will positively impact your campaigns, so be bold, and ask as many questions as you dare.