Live Webinar - Optimize LLM performance through Human-AI Collaboration

EmPushy Leverages Appen Data Annotation Services to Improve User Experience

NLP Analysis Reduces Noise in an Oversaturated Mobile Environment
December 21, 2022
Share

The Project

EmPushy, an organisation providing an added layer of audience intelligence to help brands and marketers improve audience understanding and subsequent engagement KPIs, is founded by Kieran Fraser, a former PhD student and researcher in the ADAPT Centre for Digital Content Technology based in Trinity College Dublin. Dr. Fraser completed his PhD in the domain of Artificial Intelligence and his thesis, An Approach Toward Empathetic Orchestration of Scrutable Push, addressed the lack of empathy in pushed communications.

EmPushy is now helping brands and marketing agencies improve their marketing campaigns by optimizing performance and reclaiming marketer’s campaign creation time. Kieran, and team members Shridhar Kulkarni and Beatriz Esteves, are achieving this through intelligent inference and categorization of campaign text content.

The Challenge

In a noisy content climate, marketers and sales leaders are focused on unique ways to capture their audience’s attention and stand out from competing messages. One of those avenues is push-notifications.

Push notifications are highly targeted clickable pop-up messages generated by apps and displayed to users of those applications. These notifications generally serve a purpose for the creator of the app and can be delivered to users at any time, day, week, or year. Push-notifications increase website returning traffic and app usage, improve user engagement, and help get more subscribers. Push-notifications are more likely to be seen than online ads and emails and cut through the noise to place a brand’s message directly in front of its target audience. Notifications associated with a piece of time-sensitive information, however, are only applicable for a limited time causing the notification to lose accuracy and usefulness once expired. Push-notifications also have no associated cost to the user and a far lower cost to the business than comparable SMS messages or even email.

EmPushy was looking for a tool that could provide multi-label annotation and classify data with audience targeting and GDPR compliance, empathizing quality, speed, and scale. It was a tall order and has never been done before in this domain. More than a vendor, EmPushy was looking for a partner to deliver against such novel criteria.

The Solution

Their project involved the identification of redundant notifications. In each item, the contributor is presented with the text content of a push-notification and tasked with categorizing the Call-to-Action (CTA) of the push notification. The CTA is the desired goal or action of the notification. For example, an online store may send a notification alerting users to a sale and the deadline for when the sale ends. In this example, the CTA encouraged by this notification would be for a user to purchase something on the sale before the deadline.

The company used a combination of Appen’s Open Crowd to ensure annotators were a representative sample set of the end user in addition to a Dedicated Crowd. Appen’s on-site dedicated team of contributors provided edge case data and quality improvements to ensure diverse data was collected.

The Result

Using Appen’s diverse pool of Open Crowd and Dedicated Crowd for their tasks, EmPushy was able to raise their F1 score, a measure of model accuracy, from 60% to 85%+. Through several projects, they were able to capture more than 200k campaigns, over 35 diff domains, 500 brands, and collect 223k judgments.

The success of EmPushy’s models relies on high-quality accurate data and Appen was able to deliver with the speed and scale to help them acquire more clients.

At EmPushy, we enlisted help from Appen to annotate push notifications so we could help brands be more empathetic about the messages they push at subscribers. The support and expertise provided to us by the team at Appen facilitated the creation of novel marketing NLP models, research paper publications, and new client partners. The Appen platform provided speed, quality, and customer service second to none and their ethos for empathy and equality sets them far above the competition. A joy to work with, highly recommended!”

- Kieran Fraser, CEO & Co-Founder at EmPushy