“Data-Driven Thinking" is written by members of the media community and contains fresh ideas on the digital revolution in media.
Today’s column is written by Ayusman Sarangi, director of product management at YuMe.
Data science touches our lives every day. When we share a picture on Facebook, it suggests a friend’s name to tag. When we binge watch shows on Netflix, it recommends similar shows. When we use Google Maps, it navigates us in an optimal way to reach our destination. Applications of data science are everywhere and the advertising industry is no different.
The advertising business is becoming more intelligent in a rapid way through technology powered by data science applications. The industry is attracting top talent to solve interesting and challenging business problems, as seen through job openings seeking data scientists at a scale and rate never seen before.
Data scientists are working on several fronts in the advertising world. Here are seven key areas that are impacting your business:
Marketers want to reach receptive and attentive audiences across all screens. Understanding consumers’ interests, their purchase behavior, buying intent, what devices they use and which households they belong to requires leveraging various statistics that include first- and third-party data, non-PII-based device identifiers and other data signals to build audience segments for ad targeting and reporting. Automobile brands, for example, would benefit by analyzing data to find useful patterns that indicate auto purchase propensity and creating an audience segment called “auto intender” for ad targeting.
2. Optimizing campaign for performance
Every campaign has a goal, be it reaching a certain target audience at scale or driving engagement and conversions on the brand’s site or app. Optimizing campaigns to hit those goals requires running many experiments to learn what is working and what is not, gathering insights by analyzing performance, and then making predictions based upon learning. Algorithms recommend certain sites, devices or audience segments that are performing well and then can optimize media spending towards them. A retail brand may optimize spending midway through the campaign toward advertising on tablets, driving more app engagement and offline sales. Learning is key here, as often one may not know what’s the best-performing audience and where to find them.
3. Finding the right price
No advertiser wants to overpay and no publisher wants to leave money on the table. Smarter algorithms for pricing optimization can make a big difference on margins for both advertisers and publishers. Price discovery and price recommendation in the ad marketplace looks at historical pricing patterns, customer lifetime value, campaign budgets, competitive pricing and audience data to evaluate the right price for every ad opportunity.
4. Analyzing audience insights for campaign leanings
Insights and learnings from a campaign can be carried over to future campaigns. Finding patterns between various audiences and analyzing and visualizing their performance help marketers and agencies refine their target audience. This can be as simple as adjusting a brand’s advertising campaign with broad demography targeting and refining the target demographics, based on performing audience insights.
5. Classifying inventory for contextual relevance and brand safety
Inventory classification helps in both placing an ad in the right context as well as blocking the ad in unfavorable and unsafe brand environments. Most marketers would like to avoid placing an ad next to content that has negative sentiment. A travel brand does not want to show a video ad with a travel recommendation for a country if the publisher content talks about political instability in the same country or region. At the same time, many would like to advertise based on contextual relevancy to increase the ad effectiveness. To classify and categorize inventory, one needs to analyze billions of sites and apps using learning algorithms.
6. Forecasting inventory for media planning
Media planning needs inventory forecasting to help secure inventory. Publishers need forecasting for yield optimization and guaranteeing media delivery. The complexity to forecast increases with multiple parameters to forecast against. It is relatively simple to forecast supply for US inventory in a month, but very complex to forecast for females who are aged 25 to 54, mother of one to two children, interested in SUVs and live in California, where the ad is viewable in a brand-safe environment. Here, data analysis combines audience data and historical traffic data to do inventory management.
7. Identifying traffic fraud
Fraud follows money. We see lots of fraudulent activities in the digital payments industry and now we see them in the advertising industry. Unfortunately, humans don’t see every ad, which is a big advertising challenge. Identifying such non-intentional traffic needs analysis of the traffic patterns to learn and detect any anomaly. Is there an abnormal amount of traffic coming from a certain region, part of the day, device or browser? It, too, must be evaluated and analyzed at every ad opportunity level because fraudulent traffic is not limited to only a few sites and publishers.
So what is driving this big shift towards data science-driven advertising solutions? First and foremost, digital measurability helps collect data in an unprecedented way. Second, big data solutions provide the necessary tools to store and analyze these massive data sets to help answer business questions in real time – literally in milliseconds. Third, cloud-based solutions provide the data infrastructure and scale at an affordable price that often brings up the comment by data scientists: “Let’s first store the data. We will figure out what to do with it later.”
Like it or not, the entire value chain of advertising is changing through data science. Targeting, campaign performance and optimization, audience insights, brand safety, contextual relevance and the fight against fraud are all being made easier through the magic created by data scientists.
So, like it or not, your life is about to get better. Strap on your seatbelt and embrace it.