Using artificial intelligence (AI) to identify and target specific consumer groups via social media insights and loyalty data has seen Australian chocolateria, San Churro, chalk up 6.6 per cent return on investment and a predicted $500,000 in additional in-store spend.
The brand teamed up with agency partner, Online Circle Digital, last year on its ‚Äė1000s of reasons to celebrate‚Äô campaign, designed to encourage consumers to recognise small things in everyday life as reasons to celebrate, all while positioning chocolate as a key element in the mix. The program included product giveaways and instant prize wins, capped off with a trip to Spain.
Objectives included increasing in-store transactions and driving reach and awareness, as well as signups to the brand‚Äôs loyalty program, El Social. The campaign ran during August and September 2017 and was promoted via social media and in-store advertising. San Churro operates 54 dessert cafes nationally.
‚ÄúSan Churro is known by its customers for being the destination for celebrating with friends and family,‚ÄĚ the group‚Äôs marketing director, Mark Attard, said. ‚ÄúThis campaign was developed to bring this to life and reinforce this perception while also showing our customer the value of being part of our El Social loyalty program.‚ÄĚ¬†
Online Circle Digital head of media, Brendan Sanders, told attendees at the recent IBM Think conference in Sydney that the campaign took advantage of two technology platforms. The first was the agency‚Äôs Social Pulse data aggregation tool, which pulls in social data from a broad range of industries to provide clients with benchmarks on industry sentiment, as well as how they compare and compete in their industry marketplace.
The second was IBM Watson, Big Blue‚Äôs AI engine.
Using Watson and Social Pulse, the agency devised four personas for San Churro – young female students, young mothers, young female professionals and the brand‚Äôs current loyalty program database – and used these to craft and target messaging. ¬†¬†
‚ÄúSan Churro had never done any segmentation before. The team had some data that gave us an indication towards female oriented, younger demographic, but that‚Äôs all we were given,‚ÄĚ Sanders explained.
San Churro‚Äôs best performing channel was social media, and Social Pulse benchmarks showed the brand had a far more engaged audience than its competitors. From there, the agency began analysing and collecting data from its platforms, including comments and statistics.
All this data was pulled into Watson, allowing the team to start segmenting who would be the most engaged audiences on social media platforms, with insights into which demographics were most valuable in receiving advertising messages.
The team wrote 500 pieces of ad copy, then used Watson to process the data and gain sentiment analysis in order to rank which copy would provide the best cut through. In the end, 48 pieces were pushed through to market.
Off just a $40,000 investment, San Churro gained a 6.6 per cent return on investment, with $6.60 generated off every $1 spent. The brand reached 2.2 million consumers, drove 22,000 new members to social platforms, which will attributed to $500,000 in additional in-store spend, Sanders said. Over the course of the campaign, it also had 19,000 website visitors and shifted brand percent by 4.5 per cent. ¬†¬†
‚ÄúThe campaign success was highly determined by Watson, giving us real insights before campaign started to tell us which customers we could reach that would resonate, then make those media dollars work far better,‚ÄĚ he said.
Sanders said the biggest initial hurdle was gaining client trust in using the technology for the campaigns without any A/B testing in market.
‚ÄúSan Churro was the first to give us the go-ahead to do it and rely on AI for media planning. This was San Churro‚Äôs first campaign run with us as well,‚ÄĚ he said. ‚ÄúWe give them a lot of credit to do innovation an embrace what we think is a big change in the marketing space.
‚ÄúUsually we‚Äôd write ad copy, push to the marketplace then find out what resonates on the fly. With Watson, we were able to determine which pieces of creative would resonate before campaign started. We still did more A/B testing, but it was far more refined. This gave us a big sprint and acceleration to the start of the campaign.‚ÄĚ
To do the sentiment analysis, Watson tapped into San Churro‚Äôs large portfolio of social followers. Sanders explained the software had 10,000 unique data sets within the month leading to campaign that were analysed.
‚ÄúIt was a huge data set to work with. You need a strong basis of data to make those predictions to ensure they‚Äôre realistic and not skewed,‚ÄĚ he advised. ‚ÄúSo it was important to have client with great social following to begin with. With less data sets, you‚Äôre inferring more. I‚Äôd say you‚Äôd need at least 1000 followers to provide reasonable analysis and insertion.‚ÄĚ ¬†
Attard said AI buying media and creative advertising can lead to issues for brands, noting programmatic-based brand advertising on YouTube serving ads against undesirable content as a case in point.
‚ÄúThe technology was simply optimising to cost-efficient reach, without understanding the full implications of where it was placing advertising,‚ÄĚ he continued. ‚ÄúAt this¬†stage,¬†we still believe there is a level of human control required to be involved in making marketing decisions and the parameters for its decisions are fully understood, we embrace artificial intelligence to find insights within large quantities of data which would otherwise be impossible to draw correlation from and use this to drive additional marketing efficiencies for our campaigns.‚ÄĚ
It was the initial exposure to AI provided by Online Circle Digital via analysis of San Churro‚Äôs Facebook community that helped the brand buy into AI‚Äôs potential. ¬†
‚ÄúThe analysis reinforced some of the perceptions we had – that our community was vibrant and engaged and built upon what we had suspected,‚ÄĚ Attard said. ‚ÄúIts claims were sound and justified by observations. This helped to build our confidence for AI to then assist with segmentation of audiences and recommendation for campaign messages. ¬†
‚ÄúAI didn’t control the campaigns, but deep insights to help us make smarter decisions.‚ÄĚ¬†
Online Circle Digital also benefitted immensely as a business. ‚ÄúOften when running ad campaigns through media, we do a lot of A/B testing on what type of message will resonate best with which types of audience. And we test and learn as a campaign is running,‚ÄĚ Sanders said.
‚ÄúIn this case we didn‚Äôt need to ‚Äď we already knew which creative messages would resonate and the results speak for themselves. It cut down our campaign management workload by 50 per cent.‚ÄĚ
Thanks to the proven success of the AI-led campaign, Sanders said he‚Äôll be quick to run with the results. ‚ÄúNext time, when something surprising does come back with another client, we won‚Äôt shy away from it, but embrace it,‚ÄĚ he said.
Sanders also wanted to now go back to Watson to run through the results of the campaign to see what else it can learn from the results.
‚ÄúThis will allow us to do benchmarking for campaigns moving forward,‚ÄĚ he added. ¬†
Off the back of this first success with Watson, Online Circle Digital is looking to embrace AI further across its client work, and Sanders flagged a benchmarking tool to provide consumer sentiment benchmarks for brands before then after a campaign, was on the way.
‚ÄúThis will show what a campaign has done to shift perception of a brand after a campaign is finished,‚ÄĚ he said. ‚ÄúIt‚Äôs not just measuring awareness, but the effect it had on consumers based on real comments and feedback they‚Äôre giving to those brands. ¬†
‚ÄúIt‚Äôs early days‚Ä¶ we‚Äôre excited about what else we can develop to utilise for clients for media planning and creative development moving forwards to make dollars work as hard as possible.‚ÄĚ ¬†
Follow CMO on Twitter:¬†@CMOAustralia, take part in the CMO conversation on LinkedIn:¬†CMO ANZ, join us on Facebook:¬†https://www.facebook.com/CMOAustralia, or check us out on Google+:google.com/+CmoAu¬†¬†