As the line between offline and online becomes ever more blurred, reaching a specific user becomes a more complicated task. Traditional offline marketing techniques (i.e. surveys or census information) focus on who a consumer is in their day-to-day life, from where they do their shopping, to how many children are in the household, to what car they drive. Conversely, online behavioral data is based on concrete user action rather than self-declared survey or census information, and enables granular audiences to be targeted at scale across the internet. This brings us to a modern marketing challenge: how advertisers can accurately reach people online based on who they are offline.
The solution is called “onboarding”, and the basic principle is matching an offline consumer database to an online set of cookies via a shared ‘match key’ (most commonly an email address, although no personal information, or PII, is actually transferred in the process). There are various companies specializing in data onboarding (such as LiveRamp and Datalogix), and then through integrations with other platforms, the onboarded audiences can be transferred for use in analytics and targeting. For example, using loyalty card data, we could build a set of users who have purchased breakfast cereal three times in 30 days offline in a grocery store. Then, this segment can be onboarded to allow a major cereal brand to target these regular cereal buyers with customized ads, encouraging them to buy their brand.
There are also companies who started with offline household data and have brought their data online using the same match process to create equivalent audiences online. Some examples of these companies are Experian, Acxiom and Nielsen.
The benefits of onboarding are obvious. As the world becomes increasingly multi-channel, brands need to be able to communicate with their customers in the right place at the right time. It opens up the opportunity to reach users based on their offline preferences, from purchase data to travel to lifestyle or household info. This means a brand could target consumers they know are their customers offline, but have no way of finding them online. Also, if a company has already done expensive segmentation work offline, this allows them to make use of it and be consistent with their offline and online strategies. The most likely candidates are those who have strong offline data, for example large retail, cable or telco companies. They already have databases of customers with phone numbers and email addresses, but no easy way of linking them to online users.
However benefits aside, there are still a number of challenges we face when bringing an audience online.
Challenges Around Privacy, Time and Cost
The first hurdle with bringing an offline dataset online is ensuring consumers’ privacy. Those who run the advertising only want to see anonymized, aggregated segments with no PII attached. This means the onboarder needs to match, then strip out, aggregate and anonymize the audience segments, before passing them elsewhere to be used in campaigns. On top of that, all parties need to ensure they are compliant with industry regulations. Ultimately, it increases turnaround time and the overall cost of onboarding.
Secondly, there needs to be a large enough set of online cookies with email addresses linked to them, and this is why the practice is relatively new in the US and only just starting in other major English speaking markets. Onboarding requires a dataset offline stored with email addresses and/or other personal information, and this has to match with the online PII in order to connect the two. As a result, we see a huge drop off when going from an offline dataset to online; an average match rate is 35% to 45%. This means that if you start with an offline CRM database of 5 million users, you’d expect your online user pool to end up at around 2 million.
Then there’s the inevitable drop off as the segments are transferred from the onboarder to a platform for targeting, and on top of all that, if you add in the fickle nature of cookies (i.e. multiple cookies per user, how often they are cleared, etc.), you could end up with a significantly reduced targeting pool.
Then, there’s the issue of data freshness. For example, just because a user bought breakfast cereal once last week doesn’t mean he/she does so regularly. By nature, the onboarding process is relatively slow, so it becomes difficult to ensure that a user’s in-store purchase gets translated to an online audience before it becomes irrelevant.
Performance and Measurement
Finally, just because the user is a customer offline does not mean they are going to convert if you target them online. Offline audiences are best used for branding activity; they are not as efficient in the last-view performance world. This is not because the offline audiences are not accurate or valid, we know they are made up of the right users, but if advertisers are looking for an online action at the time of serving the ad, audiences built from online behavioural data will ‘perform’ better. A better way to measure offline audience data is to look at the offline impact of the online ads, for example, to see if the users who’ve seen the ad online have gone on to spend more in-store. Companies such as Datalogix and MasterCard can do this by matching users exposed to the ads back to offline purchase data, and then assessing the impact against a control group.
Developing online audience segments that are accurate, relevant and respectful of a user’s privacy at scale from offline data is a challenge, but ultimately is a capability that will continue to grow and evolve. Meanwhile, the ability to target specific audiences across device and media is the future. For now, it is key for a company to use their own first-party data – should they have enough of it that is reliable, accurate, granular and scalable.