Challenge

Twitter is a unique platform to discover what’s happening, anywhere in the world, in real-time. Every day, news stories break, gifs go viral, and someone, somewhere, breaks the internet. Twitter covers all of these, with the latest news, reactions and conversations. The fast-paced nature of the platform poses an issue for our marketing strategy: how do we predict the big news stories of any given day and serve these to users, showing Twitter to be the best source for live content, no matter your interest? 

To do this, we needed to work with Twitter to build a solution to make their marketing as real-time as Twitter itself. We needed to reach new users and drive up active users on the platform. 

Insight & Strategy

In 2018, we developed a proof of concept within Paid Search to tackle this problem. Using Twitter’s API and machine learning to create search ads in real-time based on Twitter trends. It performed extremely well, so we began developing the prototype into a full, multi-channel campaign management tool, using AI to drive personalised, coherent user experiences across the user journey, based on high-quality conversations happening on Twitter. We call this tool Magpie. 

Magpie identifies the best trending content on Twitter and uses machine learning to analyse it and build creative and write messaging on the fly to serve to users interested in the content on other channels. For each user, it creates a curated experience by serving a collection of ads, across channels on topics that are of interest to them, based on their unique interests, behaviours and demographics. 

All of this happens in near real-time, meaning Magpie is running an always-on, multi-channel campaign that creates and serves thousands of personalised ads every day, creating personalised user experiences at scale, with minimal human intervention. This is the first time AI and user-generated content has been used to run a real-time digital campaign before. At any given moment, the conversations happening on Twitter are powering a campaign that’s reaching new users off-platform and engaging with them. 

The Solution

Every hour, Magpie pulls the UK’s top trends and fetches the top Tweets. It analyses these conversations and uses machine learning, trained on historical Tweet data. For every conversation that happens, Magpie identifies the sentiment being expressed, the topic being discussed, and uses this to turn user-generated data into structured data that can be manipulated at scale. 

From here, Magpie then assesses whether a topic is brand safe, based on the signals it’s amassed. We’ve worked to fine-tune Magpie’s abilities over time, training our models to cope with thigs like Retweets, hashtags in comments, and even sarcasm. 

Once a trend is identified as brand-safe, Magpie again uses historical data to identify which channels the trend will perform best, which audience to reach, and how to speak to them. It assembles channel specific creative, crafting natural sounding ad copy based signals like sentiment and category, and the audience being reached. Ads are automatically trafficked to the media buying platforms and mapped to appropriate audience targeting to reach relevant users. For each ad, Magpie identifies the optimum landing page that it predicts will drive the best experience for the user. All of this happens in real-time, 24/7.  

Results

This is the first use of bespoke AI to fully automate a multi-channel digital campaign end-to-end. Pulling together user-generated content with machine learning and custom integrations with multiple media partners has led to a far more reactive, real-time strategy. When a new trend or story breaks, Twitter has relevant ads across multiple search and display networks within 15 minutes. 

Magpie also aligns Twitter’s marketing strategy with their core brand values of being an open, real-time source of news where anyone can find content relevant to them. No matter a user’s interest, Magpie’s scalability means we can serve niche topics to the right users in a personalised way. 

From a direct response perspective, Magpie has surpassed all expectations, looking at our performance marketing campaigns year on year: 

• 33% increase in conversion rate 
• 212% percent increase in conversions 
• 51% decrease in cost per conversion 

Overall, we hugely exceeded our target and drove significant year on year improvements on active user volumes for the client. 

We truly believe this is a global first that represents a leap forward in terms of what it means to run a digital marketing campaign, by creating a unique and industry-leading piece of ad tech.