For most teams, the best opportunities to sell tickets come from your existing fan base. Retention is more successful and five to 25 times cheaper than acquisition. It also strengthens fan relationships, even creating lifelong fans. But you need to get the right message in front of the right people. This is where lookalike lead scoring comes in.
Lookalike targeting isn’t a new concept—Facebook, Google, Twitter, and other ad networks have their own spin on a way to reach “similar audiences,” but don’t offer a ton of insight beyond that. We wanted to pull back the curtain on how we do it, and touch on why relying on a single ad network for lookalikes might not be the best idea.
Getting closer to your ideal leads
Lets consider each person across all of your databases. We have all kinds of data available about these people in one place. You name it:
what tickets they’ve bought
We can visualize the audience as a cloud of individual points floating in space. People with more traits in common will be closer together, and people with little in common will be farther from one another:
Within this cloud, we’ve got different groupings. In one place we’ll have single game purchasers and in another we’ve got merchandise buyers, but for now, we’ll highlight all the people who bought a season ticket this season, in the center of which we’ve got something resembling the “representative season ticket buyer.”
What we’d like to do is find people who are most similar to our core season ticket buyer persona, but who haven’t yet bought a season ticket. This lookalike segment is immediately and obviously valuable: these are exactly the people we want to bring into the fold.
Without getting too deep into the weeds (though if you want to see the data science team’s eyes light up, ask us about the alternating least square matrix factorization recommendation models we use to do this), this process is a good analogy for how Umbel’s lookalike lead scoring works. But we aren’t only limited to finding the Season Ticket Holder (STH) lookalike audience. Any product, any business outcome, any segment of the audience at all can be the basis for our model. Think app downloads, people who buy your T-shirts, etc.
Testing lookalike validity
Earlier this year we ran a few pilot tests with some of our partners. One of our partner NBA teams was looking to sell a miniplan package of five games. By looking at people who had made similar purchases in the past, we found a segment of approximately 15,000 people who looked like the absolute best candidates for this miniplan.
We scored and ranked the list and then split it into three groups: Good, Better, and Best leads. When we gave the leads to the sales team, we named those lists Red, Green and Blue to disguise which was which. A bit sneaky, but anything for science, right?
If indeed we can identify and rank the best leads for this miniplan, then certainly the “Better” list would convert at a higher rate than the “Good” list. And the “Best” list would convert at a higher rate, still. And that’s exactly what happened.
We ran an identical kind of test in a marketing context as well, with one of our membership clients. This particular organization has a big, annual shindig—the premiere event for the their members. They’re always looking for ways to tick up attendance. We looked back at past attendees to create a lookalike lead list for the event and loaded our “Good,” “Better,” and “Best” lead segments into their email marketing platform. Once again, it worked like a charm. Tracking open and click-through rates off that campaign we again saw clear evidence that we’re identifying exactly the right people, and correctly scoring their fitness for these campaigns.
Prescriptive vs predictive analytics
Now, there are a few different applications of lookalike modeling out there in the market. Most prominent among these might be Facebook’s Lookalike targeting that we mentioned above. What separates Umbel’s models from the ones you already know is two-fold.
First, we incorporate data that other solutions simply don’t have—and fundamentally, all that data is yours. Facebook’s lookalikes won’t have any concept of what’s in your ticketing or CRM platforms. But those are the strongest signals for identifying your best leads. And the lack of those signals are why many have found that not all lookalikes “are created equally.”
Secondly, while other solutions may provide descriptive analysis that tells you about how things are now, and some might even throw in some predictive analysis, which makes a guess about how things might look later—Umbel is doing prescriptive analytics. That is, we tell you exactly what to do to make an impact today. Umbel’s lookalikes are more than mere social marketing reach extension. We’re handing over a hyper-targeted list of your most valuable leads for your most valuable products—and it’s actionable now.