The digital media and advertising industry has witnessed an explosion of innovation that has embedded artificial intelligence across a suite of buying and measurement technologies.
While these developments have driven undeniable value for brands, we still get many questions about “what the AI is actually doing” or “why the technology made certain decisions.” Deconstructing machine learning and big data remains a formidable challenge in an industry where millions of buying decisions are made each second.
最近，Copilot团队投资于弥合这一鸿沟，以更好地回答 "Copilot在做什么？"以及其他客户经常问到的问题。
In October of 2020 the Copilot team embarked on a journey of reimagining insights through the lens of machine learning and data visualization. It was important for our team to find innovative solutions to showcase the value of the Copilot decisions being made behind the scenes. With a team of Data Scientists, Engineers, and visualization experts, we looked at the common threads from which different algorithmic components derive decisions. Our goal was to compartmentalize the various ways campaigns are optimized and visualize them in artistic and relatable representations.
The challenge was to communicate how an ad campaign performs and delivers and explain how our technology, Copilot, adds incremental value with machine learning optimization. Adding to the challenge, the project took place during COVID-19 and required collaboration across various teams during an extended period of remote working.
To design Insights, we first connected with Copilot users across many countries to identify and organize pertinent campaign information and analytics. Through this process we discovered three key categories of questions that surface when running any Copilot campaign:
- What did Copilot do?
- How did Copilot perform?
- What can I learn about my audience?
Next, our product and visualization teams organized “sketch parties” to engage a diverse group of experts and crowdsource hand-drawn ideas. Our own Katherine Mello walks through the processes and thinking behind this virtual brainstorm here.
Each sketch party presented a specific design challenge to the team and allowed them five minutes to sketch their ideas. After short presentations and discussions, those ideas were collated, evaluated, and merged into low fidelity mocks for further input from our users.
Copilot素描派对，来自数据科学和市场营销的例子 
通过在亚洲、拉丁美洲、欧洲和北美洲不断进行的用户访谈，我们将这些低保真度的模型改进为高保真度的解决方案。在这个过程中，我们巩固了一些概念，如在每张幻灯片上传达目标绩效的色标，以及将数据主动翻译成人类可消化的语言的小工具。我们的目的是使软件包的视觉处理更加精简和易于理解。
低保真模拟，将想法和概念变为现实
The Result
Copilot Insights was launched in March, 2021. It provides an automated, in-depth analysis of every Copilot campaign. The package is available across all major DSPs to compare and contrast campaign performance in a unified view.
The Copilot Insight package contains ten bespoke views split across three sections:
目录, Copilot Insights 
Machine learning is embedded in our analytics to predict potential uplift generated by individual optimization components. These models actively deconstruct Copilot optimization to estimate uplift on a given goal type. Uplift models are embedded in Copilot Insight visualizations and highlighted across the package.
估计CPCV的提升，副驾驶的启动 
After reviewing insights available in the industry, we concluded that similar analyses often lacked simple human interpretability. To improve the experience, we embedded widgets that translate data into simplified language and offer key takeaways in every view. The additional takeaways succinctly communicate salient campaign information and make it easier for our clients to digest and understand campaign management.
小工具、网站和应用实例 
It was important for our design specialists to provide analyses in divergent, engaging, and aesthetic formats. By pulling together various analytics streams, we aimed to simplify the complexity of applying AI to media buying and audience targeting.
Copilot定位，分析Copilot活动的特点
The story continues!
Head over to the GroupM website for the latest updates on advanced programmatic solutions.
You will be redirected in in
seconds
