AppD Hackathon Project Shows APM Power in Digital Marketing
Back in the days when my AppDynamics colleague, Piyush Mathur, and I worked for a social media analytics company, we often heard digital marketers complain about their IT systems not scaling due to the additional load generated by big budget campaigns. We also saw enterprises lose a lot of money due to application performance issues, such as when users couldn’t apply a coupon code during checkout.
One particularly memorable eye-opener was when a large Indian e-commerce giant, launching its first mega-sale, invested a significant sum on advertisements and social media campaigns. While the effort brought in significant traffic, poor infrastructure management caused the retailer’s site to crash, resulting in a loss of revenue and brand image.
For AppDynamics’ January 2018 Hackathon, Piyush and I decided to take this real-life problem and solve it with the help of APM. That’s how our hackathon project was born.
The project is designed to provide real-time tracking of every dollar spent on online campaigns on digital marketing platforms—think Google, Facebook, Instagram, and so on—and the traffic each platform brings to the campaign.
Our dashboard provides a big-picture overview of every marketing campaign underway, including how well the company’s infrastructure and services are handling the traffic load. Data is pulled in from various channel APIs, such Google Adwords and Facebook Marketing. The UI displays key metrics such as dollars invested; the number of campaigns running, as well as their health and growth; overall revenue generated; and digital footfall (the number of people entering a store or shopping area in a given time). Our project is unique in that it combines social media analytics and APM on a single dashboard. In our research, we were unable to find a software tool on the market that offers this distinct combination of features.
To succeed in this effort, we needed expertise from different domains. So we reached out to AppD software engineer Siddharth Dutta who, during his tenure at Perspica, gained experience in machine learning algorithms. We also sought help from Perspica founding member Arpita Agarwal, who has a deep understanding of APM.
Where APM Fits In
You may be wondering: How can APM improve digital marketing? Global advertising spend on social media is growing 20% annually, and by 2019 will be worth $50.2 billion, according to ad agency Zenith. This is a great market to nail down, obviously, and we believe our APM-based project can provide an innovative way for marketers and DevOps to scale their campaigns and take their organizations to new heights.
In addition to helping marketers estimate the revenue generated by a particular campaign, our project delivers key metrics to DevOps teams for determining how many extra instances a service—say, apply coupon or checkout—will require to handle the additional load. And while tracking the flow, it also provides real-time insights to the campaign manager—helpful for budgeting across various marketing platforms—and to the DevOps team for optimizing infrastructure cost.
For tracking, we used a GUID (globally unique identifier), which is passed to each application service as a header. This creates a path for understanding user behavior; machine learning algorithms including an exponential model for capacity planning and a multivariate adaptive statistical filtering (MASF) model for baselining (both planned for future AppD releases) are executed to generate real-time insights, enabling marketers to optimize their budget spend strategy across myriad digital marketing platforms.
To prepare for extra traffic, we provided a nascent version of capacity planning, which can be used to plan the number of instances per service required over a particular interval of time. Our load management widget is an effective way to give the DevOps team a heads-up for infrastructure planning. For enhanced capability, the widget can be integrated with various cloud APIs to scale the system automatically in the future.
While our hackathon project is a working prototype, it implements several of AppDynamics’ core features and demonstrates the potential of APM in digital marketing. Without much effort, AppD customers will soon be able to build similar custom applications, particularly as we integrate Perspica’s machine learning technology. Over the two days of the Hackathon, we learned a lot about the market horizon, challenges, and future scope of digital campaign planning—and how a determined and talented team can showcase the power of APM across a dynamic business ecosystem.