Written by CA88’s Chief Strategy Officer & 数据科学家: 克里希纳Boppana.
One of the promises of buying digital media programmatically is scale, unlimited users at a high frequency and low cost of media. 今天, when running Direct Response or performance based campaigns, retargeting is still the best performing tactic because programmatic is very efficient at finding and converting users who have already shown intent. 然而, retargeting has its limitations, mainly that it doesn’t scale beyond users that have visited your webpage.
So, how can you scale your programmatic advertising? Is there a better way to improve ROI beyond retargeting? How do I find more consumers that will convert efficiently?
历史销售数据:
Since the beginning of marketing science, marketers have built their marketing plans and strategies using research to identify ‘角色s’ of their consumers. The data that goes into building these 角色s is often 一个 dimensional (developed through surveys and by analysts). This means that the analysis considers 一个 or fewer moments of aggregated data and doesn’t analyze the consumers holistically. In the digital advertising world, 我们每个人都是消费者, sit in more than 100 industry data segments (i.e., living in Boston, male, purchase intent, 等.平均). Traditional marketing 角色s only account for a handful of segments like demographics, 地理, 家庭收入, 等..
Marketing Data: Now – 消费者的角色
Through our data science research, we have discovered a way to make marketing 角色s more accurate, using machine learning and applying data science to available 1st, 第二和第三方数据. This is an impossible achievement with human analysts, there are just too many data points for any analyst to process to make a meaningful output in near real time.
How Do I Find New Consumers?
Stop rinsing and recycling your audience! 而不是, you want to first identify all of the data associated with your consumers, this means all of the 100 or more segments based on multiple attributes (e.g., surfing behavior, time of day, type content consumption, 等.) Then apply multiple data science models and algorithms to identify a “Persona” that captures these complex user behaviors and places them in a machine generated audience Persona Graph. The audience Persona created will be based on dynamic, real-time consumer behaviors, unlike the static 角色s or the traditional look-a-like models. Once the users and user behaviors are identified, predictive algorithms can be applied to assess the value of the 角色 and the value of each user in real time. The end result is the ability to identify potential “new consumers,” not recycled users. And knowing the value of the user will allow you to target effectively and efficiently, 提高性能.
图片问题
Humans are really good understanding and remembering pictures. So, logically, advertisers would want to see consumer behaviors associated with their campaigns. 从历史上看, this has been a difficult problem in data science – the ability to visually represent an algorithm output. 为了解决这个问题, we created an easy to understand visualization for consumer behaviors that are important for a campaign or advertiser. Below is a screenshot of our 消费者的角色 visualization.
The left image shows all the 角色s for a campaign. If we expand 一个 of those 角色s, the “young working class family” in red; the right side image shows all the audiences that comprise to this 一个 角色. You can also see the connections that each 角色 has with other 角色s. In other words, they share behaviors.
We also looked at probability of conversion. What we found was a machine generated 角色 captures complex behaviors and responds to data signals in near real time. Thus, increasing performance.
行动的见解
At CA88 we are transforming the static marketing 角色 concept into a data science driven ‘消费者的角色’ using all the data signals available. This is 一个 step further in the direction towards the promise of programmatic — extending audience reach and enhancing campaign effectiveness using what we know best, data.