Online advertising has shifted toward a data-centric ecosystem, where an increasing number of consumer purchases are influenced by advertisements leveraging data from outside parties. This data can make connections that otherwise may be unexpected and significant – say, that drivers of one specific automobile type are likely to own the same kind of computer.
With every transaction, advertisers invest in data to help them understand something they don’t already know, which can increase reach and find a new audience. The primary question then becomes, how good is that data? Is it accurate? Is it reliable? Advertisers should know that a vendor’s marketing materials are not the best source of unbiased information. Despite what the market tells you, good data is very hard to find. Buyers need to look further than marketing collateral if they want data that is actually going to change their bottom line.
Advertisers rarely buy raw, unprocessed data. Instead, they actually buy modeled data, under the assumption that this data provides some insight into consumers, and will greatly improve their ad campaigns. The issue is that this data is very often ineffective.
While brands are under the assumption that they’re chasing qualified consumers, they still frequently fall short of this goal. Data providers often take what few insights they can extract from spotty data and try to find a common thread to attribute to a larger audience. The widely quoted statistic in the industry is that 40 percent of consumers bought and sold as “new moms” are actually male, and there could be larger discrepancies for other profiles.
Modeling is not the problem – the underlying data being modeled is the problem. Modeling is an absolute necessity online, and without the tactic it would be virtually impossible to target consumers on the scale that advertisers want. But in today’s marketplace, data providers often scrape a small layer of data, then find a common thread to build a model. The buyer (the advertiser or agency) very rarely has any insight into the primary source, or the overall sample size that the model is built on. It really is a case of “buyer beware,” where little is known about the quality (accuracy) of the underlying data and therefore the same goes for the quality of the resulting models built on top of the raw data.
Of course, there is no reason not to buy modeled data, but there are plenty reasons why advertisers should ask more questions about models and source data. The best data comes from primary sources, and it’s even better when the party with the primary source does the modeling before it’s passed to the buyer. By nature, this is the most transparent way for advertisers to consume data. There may be an intermediary data exchange in place to handle the financial transaction, but this is still getting data directly from the source.
Advertising with audience insights is a necessity online, and doing so effectively is a simple case of better data in, better results out. If a data provider is using their primary directly measured data, it will be better than blindly modeled data. It doesn’t matter if we’re talking about financial data, shopping histories, or interest data. Segments based on primary data deliver the most effective results.