Skip to main content
News Analysis 26 Apr 2023 - 5 min read

‘Byron Sharp should like this’: Big datasets behind Dentsu Media ANZ's deal with AI market mix modelling startup Mutinex; links core business KPIs to media, brand metrics; CX next

By Paul McIntyre - Executive Editor

L-R: Mutinex CEO Henry Innis and Dentsu Media CEO Danny Bass want to change media buying with AI-led ROI "industrial analytics".

Dentsu Media ANZ just inked a deal certain to set tyres smoking among media companies and rival agency groups; the local unit of the Japanese-owned global communications holding company just went outside the camp to ensure it’s place in the lead pack deploying AI in automated econometrics and market mix modelling. The most unsexiest of capabilities in marketing is about to beat slower humans doing bespoke projects and go mainstream with faster, cheaper and more predictive automation. The new alliance, with an undisclosed exclusivity period, pledges better results for blue chip clients by planning, buying and even changing media and publishers mid-flight based on pre-determined real-time tracking of business KPis, including revenue or profit - instead of just traditional media and brand metrics.  

 

“Our ultimate goal here is to connect marketing metrics with business metrics,”

Danny Bass, CEO, Dentsu Media

 

Media and marketing metrics connect

It's going sci-fi. Or maybe just AI. The eternal quest to prove how media and advertising activity contributes to current and future enterprise revenues [demand] and the P&L is about to break open. 

Econometrics is not a traditionally sexy theme for most in the marketing supply chain, agencies particularly, but it just got there for global agency holdcos. AI and automation is about to throttle slow, bespoke and human-led econometric modelling which in media, for instance, identifies the channels or publishers delivering the best ROI outcomes for a specific brand. The problem with conventional market mix modelling is that recommendations are sometimes six months or more in the baking.

Most of the banks, many insurers and a handful of consumer goods companies have deployed the Mutinex SaaS platform internally; the Dentsu Media deal is the first for an agency group and for both firms is borrowing from the “partnership programs” designed by marketing tech players like Adobe, Sitecore and Salesforce.     

“Our ultimate goal here is to connect marketing metrics with business metrics,” Dentsu Media CEO Danny Bass told Mi3. “How do we square the circle on that and make that a reality and make it happen? This partnership will be a very strong component of that as we move forward.”

 

Professor Byron Sharp Byron [Ehrenberg-Bass Institute] will like this. It’s about big, robust datasets using AI to derive real world business and marketing rules. When we're looking at the trends in the industry, that's really quite a pioneering change

Henry Innis, CEO, Mutinex

White flag

Bass joined Dentsu Media in August last year and is preparing to relaunch the business in coming weeks with a new suite of paddock-to-plate designed products, platforms and managed services, including a move to link CX and media via Dentsu’s Merkle division. The Mutinex deal is the first glimpse of what a re-engineered Dentsu Media will look like. Most agency groups have built internal econometrics teams, including Dentsu Media, but it’s the first global agency network to acknowledge that partnering externally with the tech smarts, processing power and an “industrial analytics” of a platform like Mutinex in econometrics is better than attempting to build something itself.  

‘Seismic’ new rules for allocating media

So, will this deal change how and what media Dentsu buys?

“The short answer is yes,” Bass said. “To what degree? We will see. Fundamentally, and this is an industry thing, not a Dentsu thing, have media schedules moved quickly enough to reflect the changes in media consumption? The easy answer to that is no, they haven't. But there needs to be seismic shifts in every area of this industry in order to do that. If you look at a pitch now of any great size, what an agency is asked to do in terms of answering buying templates are very similar to what they were 20 years ago. So we can see the challenges that are coming. We can see what's happening with audiences and we need to be able to make quick decisions on behalf of clients that we've never done before. The ability for Henry's team (Henry Innis, CEO Mutinex) to work closer with my team in both industrial analytics, but also importantly client partners to better inform and make quick decisions, particularly given the way the market may go in the next few years is where we have to go.”

Bass cited the rapid audience declines in free-to-air TV as a case in point – while the headlines might suggest a fast reshaping of media schedules out of terrestrial TV to alternative channels, the Mutinex GrowthOS platform could flag the opposite because revenue or profit ROI scenarios were still delivering. “I don't want this to be a TV bashing thing, just to be clear,” Bass said. “It still accounts for a lot of spend but if we're seeing audience fluctuations in the way that we are, then how can we best use that information, working with Mutinex, to best inform clients? It might well be that if the headlines are saying TV audiences are declining and the automatic reaction is to move away from TV, Mutinex data might say don't do that because what we're seeing is you still should be leading with TV.”

Media automation done in five years

For Bass, the Mutinex alliance is part of the media group’s preparation for a media system that will be all but automated and programmatically traded within five years. 

“This is that first step into what I think everyone now agrees that within, I would say five years tops, the very vast majority of all media will be traded programmatically and through some form of automation. I don't think there's any question of that not happening. That's not to say certain media channels will not exist but there is an absolute inevitability that at least 90 per cent is going to be traded programmatically or through automation in one way shape or form.  To have a relationship with Mutinex that makes sure we are making those right decisions and not being overly influenced by headlines, gut instinct or just a feel is core to where we need to be.”

What does Dentsu get?  

Mutinex's Innis said the Dentsu alliance gives it aggregate ROI data on a media spending pool of $1.5bn in Australia, derived from GrowthOS’ current client base in the platform. It does not give visibility at a category or sector level but Dentsu Media can overlay it’s own data and clients to derive media recommendations. “We have the largest aggregate data set on ROI trends in the market,” he said. “That ROI data is from a combination of data sources -  economic factors, seasonality, pricing what those components will do to influence business metrics. Dentsu will have access to those feeds, which is quite unique. There’s never been a scaled ROI data set out in the market that's then used in this way. And it's just the top line - let's say a top line channel ROI on out of home, a top line ROI on TV and so-on. What it will give is a very strong sense of how channels are performing from a return on investment and P&L perspective.”

AI ‘foundation models’ – wtf?

Innis argues ‘foundation models’ in AI will revolutionise the marketing and media sectors – ChatGPT is the most obvious example. It’s a term coined by the Stanford Institute for Human Centred AI in 2018 which refers to a model that ingests broad datasets that can be trained for specific tasks – like ROI media modelling for profitability or revenues. “There is going to be a whole suite of foundational model businesses in different domains like economics, logistics and images,” Innis said. “What we’re doing with Dentsu is one of the first truly scaled partnerships between foundational model technology and the people who are going to actually enable and make that technology really valuable and used and well adopted in media."

Innis said "Professor Byron [Ehrenberg-Bass Institute] will like this", referencing a blog post Sharp published on how market mix modelling "muddles marketers". "It was an excellent post on market mix which highlighted the core issue," Innis said. "Bespoke modelling basically means a model has to hold up for a specific dataset, not under a range of conditions. Foundational models are trying to challenge that way of working in data science. It’s about big, robust datasets using AI to derive real world business and marketing rules. When we're looking at the trends in the industry, that's really quite a pioneering change and I think it's emblematic of where I personally think foundation model businesses will go in the marketing sector.”

Super exclusive

As revolutionary as both companies claim this deal will be, Innis and Bass would not be drawn on the deal terms or exclusivity window. “Henry is not part of Dentsu so I'm fully aware that there will be other relationships,” Bass said. “But what we've worked through is I believe a great period that gives us first mover advantage. We've got to galvanise a team behind this, train them up, get them going, work together. So we've agreed a period of time which can make this work really well.”

Equally, Innis tapdanced around terms. “Our business has been built some 70 per cent through referrals and it will tank if we do a bad job. I really want to get this right and not necessarily explode the market on this one. We’ve found the right partner to work through and get this right and to get it working at scale. The exclusivity period is not going to be forever, obviously, but I would say Danny's obviously got a head start and he's got a head start for a reason. We're trying to build something here that's scalable for us and that works for Dentsu.”

What do you think?

Search Mi3 Articles