Analytics news and expert advice every Thursday.
Third-Party Data Is Awesome, But Maybe Too Powerful
Columnist Alistair Dent explains the ins and out of third-party data and how — when used the right way — it can help marketers gain powerful customer insights.
Third-party data is the most controversial data source available to marketers. The market for data is in constant flux, and we can expect it to change a lot in the future.
Third-party data is data you can act on that comes from other providers. It’s not your own user data, nor is it coming to you from direct relationships or publishers. A media owner might segment its readers into “entertainment” vs. “politics” vs. “gossip” and make those segments available to buy against.
Somebody like Experian might take its detailed financial understanding of people, combine it with their location and browsing behavior, and try to create a “best guess” profile for other users. You can then layer that data over your inventory to buy ads in a more targeted fashion.
Like first- and third-party data, third-party data acts like a series of filters. Each additional data source you apply helps you to narrow down your targeting, improve your bidding accuracy, or write ads that are more relevant to the people seeing them.
There are two main sources: public data and data brokers, and two main places to use it: Demand Side Platform (DSP) and Data Management Platform (DMP). Let’s take a look at each in turn.
Enriching your existing and incoming CRM (customer relationship management) data is as simple as doing a lookup against public records. In the UK, that includes data held at the Driver and Vehicle Licensing Agency, Land Registry, Electoral Roll and the phone book. These public databases exist to allow people to check into a car’s history, to check anything that might impact a home purchase, etc. You can use that access to learn more about your existing users.
This is very much a melding of first- and third-party data. Unless you know users’ details to begin with (i.e., they’ve already given you their information) you’ve got nothing to look up. Once you do, you can start understanding users better. If the electoral roll shows they have dependents, you can infer they’re more likely to be a family. If it indicates their car was purchased 11 months ago, they might be in the market for car insurance now, as their original policy will be up for renewal.
One of the most powerful public data sources available are social listeners. With a person’s email address you can see what publicly stated interests they have on their Facebook profile, which Twitter hashtags they interact with, what organization they work for and at what seniority from their LinkedIn profile.
By combining all of these data sources together you build a deeper understanding of your existing customers, either to help your sales teams, your eCRM (electronic customer relationship management) and comms strategy, or your media buying and lookalike modeling.
There are lots of companies offering data to the market. These typically consist of segments of users that have been cookied, based on browsing history or other data.
Core data companies tend to be tracking companies: firms like ComScore have their pixels on a lot of sites in order to just generally monitor how websites are used. They can then say, “This user has visited five-plus news sites in the last week” so that person can be in a segment as being interested in current affairs. If they can say, “This user has read car reviews and visited classified ad sites” then that user might be part of an “in market” segment for a new car.
On top of browsing history, companies with other kinds of data might offer other things. Credit card companies can use their online portals to drop a cookie on a user’s browser if that person is into fine dining; Experian knows the typical demographics and income levels of different geographies, so it can offer to layer demographic data onto your geographical targets.
Even better than this sort of implicit data from monitoring behavior is explicit data. You could partner with a company that runs “Which Harry Potter Character Are You?” quizzes. Each time a user takes it, he or she is answering a raft of personal questions that get stored in the cookie from that provider. Quizzes, questionnaires and competitions help companies enrich this data further.
By combining this sorts of data sources together, data brokers build detailed and granular customer segments.
You’ll never see this data. As an advertiser, you won’t know anything about how this information matches to your users. What you can do is use those segments in your DSP or DMP.
Using Data In Your DSP
Next month we’ll discuss DSPs in detail, but for now it’s relevant to discuss exactly how third-party data becomes available.
In any DSP, you’ll see a long list of possible customer segments. Some are free, provided by your own first-party data. Some are free or paid, provided by the DSP owner. Many more are paid for datasets, provided by data brokers.
When you build a campaign (let’s focus on display campaigns for now) you choose your targets. You might know what site content you want to appear alongside and have a list of keywords to help define that.
If you offer luxury skiing vacations, then showing your ads alongside articles about skiing seems natural. But you don’t know which people seeing your ad are likely to look for a luxury vacation vs. a bargain vacation. By scanning the list (they’re long, but searchable) you can find a segment that’ll match, with an associated CPM.
When you apply that segment, you’re adding that CPM onto your advertising cost, but limiting your ads to only people who match that segment. In this case, you might have two segments: “Users in market for skiing vacations” and “Users who enjoy luxury experiences”.
You’ll pay more per person who sees your ad, but you’ve filtered your advertising down to only be seen by users you think will buy a vacation from you. Each filter saves you money, and your balancing act is to save more money than the extra per person cost, for the same number of users buying vacations.
The next step is to move beyond using data as a filter, and use it as a bidding signal. If you saved money for the same number of purchases, then your ROI has improved. But if your previous ROI was profitable, then your motive probably wasn’t to improve ROI further — it was probably to sell more vacations. You can take the savings and reinvest it, increasing your bids and showing to a higher proportion of users in your targeted segments. Amazing.
It’s easy to see why third-party data can improve your buying by layering it in through your DSP, but you can do some other interesting things, too, using a DMP.
Using Data In Your DMP
Your DMP is the place you’re trying to link datasets together. You’re storing first-party data in your analytics platform and your CRM. You’ve got third-party data to buy against in your DSP. But for real insight, you want to start pulling these together.
Your DMP is the place for you to segment your own first-party data. By creating rules like “people who have viewed more than five pages in category A” to define users interested in A, or “people who have purchased more than two items in the last 12 months” to define users who return to purchase from you, your understanding of how user behaviors differ improves — which is great, obviously.
It’s even better when you start combining those segments with third-party information. If your DMP has access to the major data brokers, you could ask, “What demographic does Experian think my segment A users are in?” Immediately you’ve added something extra to your segment, something that helps you predict conversion rates and decide ad messaging.
It can get more interesting, particularly when you’re trying to look at cross-sell and up-sell opportunities. Your “users with house insurance from me segment” could be combined with a third-party “users in market for a new car” segment, to identify people who’ll be looking for car insurance quotes soon. That’s interesting beyond what you can buy in your DSP.
By having this in your DMP you can ask, “How can we email our existing customers about car insurance?” and know who you should email. You can look for new business opportunities, you can help inform creative comms decisions, and you can rework your branding to try to appeal to the best part of your customer base, or see where your customer base has gaps compared to the market.
The single most important part of filling extra data into your DMP is to enable high-quality lookalike modeling.
Lookalike modeling is already a pretty descriptive name. It’s easy to understand what we’re trying to do: target people who are not yet customers, but behave similarly to our existing/best customers.
High-quality lookalike modeling turns prospecting from a “let’s advertise everywhere with a wide net and see what works” mechanism, to a targeted and high return campaign, the way digital marketing should be.
But you can only find people who look like your existing customers if you know what your existing customers look like.
Every single extra data point you have associated with a segment gives you better insight into what that segment looks like. The third-party data you have access to in your DMP can be used to enrich that view of your segments, and start leveraging that out in the market.
Whether the data is based on demographics, browsing history, transaction history, explicitly collected personality insights or any other mechanism, you can do more when you have more.
This all comes with some privacy and data collection concerns, so let’s have a look at what that means.
If I learn that a service I use is selling my data, I get worried unless I’ve specifically said that’s OK. If I learn that services I’ve never even heard of are selling my data that they’ve aggregated by tracking several sites I use … that’s even worse.
As an advertiser, you have two major things to navigate: privacy regulations and consumer sentiment. The latter is far more important, far more restrictive, and far quicker to change. You don’t see any detailed user information beyond your first-party data. You never get to see people’s browsing habits.
You are still going to pay for that access, though. You want those filters, and you’re willing to pay third-party data brokers to improve your targeting. Users end up being targeted based on their behavior, by companies who they have never had a dealing with before.
You know that you’re doing it to improve relevance and save money, and that the users get a better experience because by using data, you’re showing fewer ads to people who won’t appreciate them. Users certainly won’t see it that way when they get told you’ve bought access to their data for targeting purposes.
Third-party data is a fantastic addition to your marketing and your customer insight. Your goal is to use it to improve your buying and optimization, your creative and messaging, your customer understanding and therefore prospecting. You also need to be gentle with it, and careful about how you’re treating customers and potential consumers.
Some opinions expressed in this article may be those of a guest author and not necessarily Marketing Land. Staff authors are listed here.