Select Page

Driving sales with predictive analytics


A homegrown software program sat unused on a shelf of a major ad agency. Its secret power: calculating social media marketing ROI. My team and I dusted it off, productized it, wired up some predictive AI code found on GitHub, and brought in more than 30 new enterprise customers.




Type of engagement



Project length










I had started working with a new team at a global ad agency. My remit: to provide the oversight, tools and training to empower local agency teams in 15 different markets, who would in turn provide competitive, unique social media marketing services to agency clients worldwide. Unfortunately, the concept of cross-market Centers of Excellence has a lot of challenges, and each of these teams felt they knew their market best. Few of them wanted to hear from “corporate” — unless we had some usable tools to gift them.

Corporate social media teams are generally considered pure cost, and agency clients generally had no idea how to justify the cost of social media campaigns and resourcing. One of my teammates mentioned a software package created as a project, now languishing on a shelf. The software calculated social media return on investment, using an approach called Earned Media Value.

Social media ROI is a bit of a holy grail, so finding this gem was like finding a rainbow-colored unicorn on a shelf that laid golden eggs. Surely this was a tool that we could productize and distribute to our local agency teams!


We first looked at the code, which was 90% prototype, not customer-friendly and definitely not package-worthy. We introduced it to our internal software engineers who were excited about expanding a product that we could take to market.

My core team began working on the value proposition, messaging, and product identity. We did a little market research with a few potential customers, and by the time we’d helped them understand the Earned Media Value concept, they were intrigued.

We decided to polish the user interface, fill in any functionality holes, give it a good visual identity, and offer it to our existing clients.

One day, during an agency-wide hackathon, a young developer in our Kuala Lumpur team said, “since we’re calculating social media ROI, we should use predictive analytics. There’s code out there on GitHub that will read in two years of social posts and predict what the user will post two years into the future. It’s free. I could wire it up and we could likely predict what people will post on social media in the future. Interested?” We gave her the go-ahead and in less than a day, the code was integrated into our platform.


We gave the new platform a name and visual identity. In the meantime we unleashed front-end Javascript developers to overhaul the old Tableau code. Looking for internal buy-in, my business partner and I began to pitch it within the agency network. Most of the agency was heavily focused on broadcast media and paid search buys, so it was tough getting anyone to care, but we had a few scattered champions.

Once we worked out a pricepoint, we put together a pitch deck and took it on a roadshow to client events in France, the Netherlands, Germany, Ireland, Italy, Japan, Spain, the UK and the US.


As we educated clients on Earned Media Value, more and more global companies signed up, including a major automaker who let me know the product was the primary reason they signed. All in all, our AI-powered ROI platform landed over 30 new agency clients and a considerable spike in revenue.


Want to chat?

8 + 1 =