Blis Forecasts Where Consumers Will Be — Then Targets Ads Based on That Prediction
Blis, a provider of local intelligence, is unveiling a new service today that it claims can figure out where consumers will go — and then target mobile ads based on those expectations.
Blis Futures, an artificial intelligence-driven, machine learning solution, identifies patterns to determine where consumers are likely to spend time, and then focuses brands’ marketing to reach them at optimal moments. When the technology is in place, Blis says it will only charge marketers when consumers visit the chosen locations. This is an example of a growing trend in the mobile ad industry to focus on demonstrable “cost per visit” outcomes and less on impressions and click-through.
Gil Larsen, Blis’s vice president of Americas, spoke about the launch of Blis Futures and how the company is helping brands better determine when and where they should target their customers.
Why did Blis develop these new tools?
Being able to drive people to specific brick-and-mortar retail locations has been a hot button topic for several years; there are several companies that are doing it, including ourselves. We’ve run hundreds of campaigns where we can correlate device IDs that have seen our advertisements and gone to specified locations. Up until this point, we’ve done this on a CPM (cost per thousand impressions) standpoint and backed our way into the metrics of what it takes to get a certain amount of device IDs to a certain store.
What Blis Futures is — we’re basically demonstrating confidence in our technology and the strength of our predictive data analysis. We’re only going to charge advertisers if the customer visits the intended location, using a cost-per-visit metric.
As much as we’d like to think we’re all spontaneous, human behavior is about 93 percent predictable.
How do you forecast where customers might go?
We’re running hundreds of campaigns every day, across 11 offices and 52 markets that we’re in around the world. We’re constantly ingesting device IDs and building audience segments off those IDs. The way it will work is, whether it’s a retailer, a consumer packaged goods (CPG) brand, or a quick serve restaurant (QSR), we will get an understanding of who is their target consumer, what the brand is looking to do, what are the specific locations they are looking to drive to. For the first part of the campaign, we will have our technology figure out based on the key performance indicators and the specific location they want to drive customers to, what are the right device IDs to target.
The beginning of the campaign will be a lot of listening; seeing device IDs, where they go, and what locations they go to. Figuring out what is the right cost-per-visit metric. The first piece of the campaign — on a CPM basis — our machine learning technology figures out who to target, where to target them, and what the right moment to target them is based on the store locations. The second half of the campaign will be where we convert from a CPM to a cost-per-visit campaign. The advertiser will only be charged for device IDs that go to the desired locations for the better part of the campaign.
So this will be able to assess, via device IDs, patterns such as when consumers tend to frequent a particular coffee shop and where they visit afterwards? Is that the understanding that goes into predicting where they will go next?
That’s absolutely right. It’s using predictive analytics. It all depends on the KPI (key performance indicator). If a brand is just looking to get any customer, whether a new customer or a repeat customer, into a store, that example is right. If we see behavior that tends to show this consumer tends to go to coffee shops in this area, we’re pretty confident we can drive that consumer to the specified location.
One of the things we’ve been asked for — and it is a good differentiator — some brands only want to drive new customers and don’t necessarily want to pay for customers who already go to a store. We have a solution for that. Before the campaign starts, we would listen and any device IDs we see going to the location prior to the campaign starting, we would exclude. We could also, during the campaign, look back a week, and exclude device IDs from within the last seven days. It depends on what the KPI of the brand is.
There are certain stores and brands where people are going all the time. This has come up in CPG specifically. If you have someone go to a location where there’s a lot of different things sold, how do we know the ad actually drove that business?
By using device IDs, you are able to make these assessments without revealing personally identifiable information?
Correct. It’s all device ID. It’s opt-in device ID as well. Most of our inventory is in-app inventory, so you have to have your location services turned on and opt-in to share you location within the specific app. There’s no PII data is transferred.
How did Stella Artois make use of Blis Futures in the beta test?
With the Stella campaign out of Europe, they’re looking to understand their customer behavior and real world insights in addition to digital insights. This gives them an opportunity to engage with them in the right moment. They’re looking to drive incremental new customers into bars that serve the product. People who visit bars probably visit a certain set of bars, but not necessarily for Stella Artois. This is something they wanted to deploy to get better insight into their audience and the impact of the brand campaign in terms of people going to bars because they saw the Blis ad versus they were going there to begin with.
Is there a brand segment that Blis Futures especially meets the needs for?
The sweet spot for this is any brand that is brick-and-mortar-centric, where driving to retail locations is a key KPI for them. Stella Artois is a great example. Retail makes a ton of sense; CPG is a big one for us. QSR is another big one. Those are the categories where we’re seeing the most interest.
Read the original article here.
Tags: AI, BlisFutures, CPV, Gil Larsen, Street Fight Magazine