In this piece, we'll tackle guidance on the use cases for identity graphs in consumer packaged goods.
Consumer packaged goods companies are, in droves, working to generate a much deeper understanding of their customer. They are increasingly bringing their media planning and campaign activation in-house, and building direct relationships with the adtech community.
As brands make these seismic switches, consumer packaged goods companies face a number of challenges, but chief among them is not owning large swaths of the funnel.
Instead, much of the customer interaction is owned by retailers and eCommerce vendors, with brands left guessing which messaging is actually driving product adoption.
Whether you are selling M&M's or mobile phones, if you don't own the last mile of the path to purchase (aka the execution of that purchase), it's significantly more difficult to:
- Understand who your most likely prospects are (and are not)
- Identify the most relevant messaging for each prospect type
- Deliver the correct message to each consumer at the right time
Identity graphs can help with each of these challenges, but most brand marketers have difficulty implementing the tools to address them at scale, especially since many brands see very little direct interaction with their customer.
In this article we will examine some critical use cases of identity data and identity graphs in consumer packaged goods (aka the CPG space) and how they relate to the goals above.
The objective of this overview is to give CPG marketing teams concrete steps they can work on together to unlock the value of understanding identity at scale, without overwhelming their (usually) small data analysis teams.
How does a brand build from the data they have?
In CPG, the data we see is often related to the truly "faithful": the consumers who are most brand-loyal.
It's pretty rare for someone to go to the M&M's website, even if M&M's are their favorite candy. Those who do go are the brand's most ardent fans.
Add to this that people aren't likely to visit the M&M's website often, and it puts brands on their heels when they are trying to figure out who their customer really is.
Identity can help.
The faithful are a great group to use as a seed audience, but data about them needs context and stability.
Identity can help with both context and stability by translating those cookies into Mobile Ad IDs (MAIDs) and Hashed Email Addresses (HEMs), each of which can be easily related to deep demographic and psychographic data, interests and more.
Both are also notably more stable than the ephemeral third-party cookie, so storing data against those stable IDs makes it much more valuable in the long run: you'll be able to use it for months, maybe years, rather than weeks.
This allows marketers and brand data teams to make decisions about their audience over time, and act on those choices through things like lookalike audience generation and demographic/psychographic profiling.
It doesn't end there, though. Through a combination of identity matching and detailed attribution through partners, we can match SKU-level conversions in a way that lets us build more and more data about brand-loyal customers, which we can push back into the model.
As those initial lookalikes see ads for M&M's, we can watch for people in the target group using an attribution partner that has access to SKU-level data via loyalty programs, and see that the person we're advertising to is buying M&M's rather than Skittles.
That knowledge feeds on itself over time and becomes an increasingly valuable dataset to drive sales by better understanding the customer, even when you don't have access to a SKU-level attribution partner.
How identity builds insight and revenue, step by step (with a snowball effect)
★ Customer Segmentation
Step one is to stabilize the audience we have into something we can act on: HEMs and MAIDs. To do that, we take the cookies that come to the M&M's site and match them to data coming from other sources to the identity graph provider. This gives you a match between what you've learned about a user and a MAID and HEM that can be easily uploaded to a DMP or CDP to generate more information on that user.
Functionally, you build a segment (or segments) in your CDP containing the IDs representative of your faithful audience.
Now that we understand our audience and have built it into segments we want to target, based on our own knowledge of the user, we want to add what we can from DMP or CDP partners.
For this, we restrict our segments to MAIDs and append demographic and psychographic characteristics to further assist in segmentation. Once that further segmentation is complete, we re-attach all the non-MAID IDs representative of our target segment via the identity graph.
★ Lookalike Modeling
Once we have a good handle on our base segments, we can apply machine learning to those segments to build similar audiences that don't include your faithful, but look very similar to them. This lets you take your known group of, say, 1 million faithful customers and expand it to 5 million people who look like them.
From there, you'll want to pass that larger lookalike audience through your identity graph again to expand the number of IDs in your lookalike target segment to something like 12 million.
At each step, you're really just bouncing each segment you build off the identity service before you take your next action.
★ Detailed Attribution
The snowball effect starts to take shape when you add detailed third-party attribution to your toolset. These providers have access to panels of people who share detailed loyalty and/or credit/debit card transaction data, which they can match to your lookalike target segment.
Basically, if you reach an ID in real-time bidding, you send the attribution provider all the IDs known to belong to that person.
If any one of those IDs is in their panel, they will watch for conversions and tally those that happen. They will also build a projection for the full reach of your campaign based on the ratio of their panel to the population in your campaign area.
Those projections differ if you send all of the IDs related to a person rather than just the one you actually reached them on, so be sure to work with your attribution provider and make sure they understand that you're using an identity graph.
Conclusions
Adtech companies sometimes make big data projects feel out of reach for smaller CPG analytics and marketing teams.
But they don't have to be.
Yes, an identity graph is a raw data asset, and it takes some massaging to make it really work for you. But the power you gain in each use case makes it well worth the effort, and the massaging is often much less significant than it appears at first glance.
More importantly, when you work with Roqad, we've done it before. Our team has built and implemented identity graphs for use cases as simple as segment extension for a DSP and as complex as measuring mobile phone churn for a hedge fund, and everything in between.
We've seen it all. Don't fear the unknown when it comes to identity graphs in consumer packaged goods. We can help.
Questions?
Contact us. We're here to help.