10 Questions Every Marketer Should Ask Before Buying Audience Data
by Adam Paulisick
Sometimes, investing in efforts to reach target audiences can seem like a big gamble, with the odds stacked against you…
Here you are, charged with exciting media buys, and you find that there are literally thousands of data providers who tell you using their insights will allow you to find the perfect audience. But which choice will make you a winner?
Assuming you don’t have enough time or money to try them all, here are 10 questions to ask before buying data.
- Does the data reflect actual buyers, or are they merely look-alikes?
How someone communicates with the US census every 10 years is a far cry from providing a complete picture of how loyal of a consumer that person is.
The data should reflect consumers who have purchased the advertised product/service or taken an action that would instigate a purchase; the dataset should not merely represent a stereotype of people who “likely” will react to your ad because of their age, sex, or income.
Challenge a dataset to reach more than a legacy demographic segment such as “women 18 to 54.” Find out whether that dataset actually measures past consumer behavior.
You want to reach an actual consumer of your product or service, not just someone who looks like one.
- Where does the data come from (and since when)?
Too often, marketers don’t know the source of the datasets they use. There are myriad reasons for knowing where your data is coming from, not the least of which are matters of privacy or liability that can have consequences for the person buying the data.
But understanding the source of your data is also important for the sake of consistency. Have the same sources been used for the past year? If not, you run the risk of having different results for different years and no way to really understand what worked or didn’t.
If a seller won’t disclose the origins of its data, there’s likely a reason… and it probably isn’t to your benefit.
- How do you know your data is representative (smart) and not just big?
It’s crucial to ensure your data is truly representative of the consumer you’re hoping to reach.
Consider data on consumer packaged goods sales; that’s an example of an industry dataset that is now available to digital marketers. If you’re using retail sales data to prioritize which households will see your ads, you’ll need to be completely certain the dataset you’re using accounts for purchases made in all outlets—not just grocery, big box, or convenience stores.
Let’s take the CPG example further: Imagine you’re a toilet paper marketer. It might seem obvious to use offline sales data to create precision marketing audiences that will see your digital campaign. But what if that data is derived only from a set of offline grocery sales data? You’d miss all the potential consumers who buy any of their toilet paper at a big box or convenience store. If your data isn’t representative of all the outlets where toilet paper is bought, your segmentation strategy will miss certain customers and understate the total incremental sales made at those locations.
Finally, it’s worth confirming that all your datasets are linked (for the household or person you’re serving the media to). If the datasets can’t “see” each other, it’s the same as only having a single dataset. Commingling of datasets is one of the major differences between “big” and “smart” data.
- How often are your sources updated?
Some data ages quickly: Markets can be volatile, and consumer attitudes can shift rapidly. It’s especially important to consider how often the source of your data is updated if the category is prone to heavy seasonal swings, frequent recalls, disparate trends based on geography, or availability issues.
- What control do you have of your sources?
Be wary of datasets that come and go. It’s not uncommon for new data to be created and older data to be worth letting go, but having a choice about it certainly is uncommon. Find the source, and make sure it can’t evaporate without your consent.
- Is your data surrounding the individuals or the total household?
Every marketer has different goals. Telecom advertisers are likely to be more interested in an individual’s profile, whereas CPG marketers may consider purchase behavior in terms of households. Either way, you want to make sure the data gives you a complete picture of who might buy the product and whether they have already bought it
- What level of granularity can you get to?
Understanding the granularity of a purchase dataset is crucial for understanding its usefulness.
Consider the example of SKU- vs. UPC-level data. SKU (short for “stock keeping unit”) data doesn’t differentiate between a brand and sub brand. UPC-level data makes key distinctions between related products. Said differently, an SKU dataset would include all forms of XBOX sold at a retailer, whereas a UPC set would specify which variation of XBOX was sold (e.g., number of controllers and the like).
- How much history do you have for the data (either household or individual)?
Robust historical data is necessary for any marketer who’s trying to overcome challenges of seasonality or understand macro trends in the market that may be slowly taking place over time. New jobs, kids, marriage, and infinite other possibilities also affect how and whether consumers buy a product.
It’s also important to monitor data over time to determine whether a behavior is truly new or it just surfaces episodically.
- Where can the data be activated, and how?
Understand, up front, what limitations there are to activating a particular dataset. Be sure that you know the specific places and use cases where the data may or may not be applied.
- What’s the typical response or reaction to the data?
Norms and benchmarks are critical. Though it may not seem fair to ask this of every dataset, it’s necessary to ask for what results are available. Case studies and evidence of any sort should be available to prove you aren’t the first person taking the car for a ride.
* * *
Here are the key takeaways:
- Not all datasets are created equal; know what questions to ask before investing in enterprise level data sources.
- Don’t use datasets from companies that won’t share their sources or collection methods.
- Big data can be misleading if it isn’t smart data (tuned by smaller, balanced datasets).
- Historical collection, granularity of the data, and how representative a dataset is affect both effectiveness and cost, not just cost.