AI driven modelling is both present and future – and media agencies should give up the fight to own it

A number of years ago, I wrote an article about a new organisation called Blackwood Seven, an AI-driven SaaS media modelling, optimisation and media trading platform. Fast forward to 2021, and Blackwood Seven, along with various competitors, is still growing. In Blackwood Seven’s case, it has divested the inventory purchase part of the original service offering and is now focused specifically on AI-driven marketing modelling.

Smart move. Back then, the only question mark I had was around the ability of a new player to gain real access to the breadth and depth of media inventory at the scale enjoyed by the established players.

Tech or no tech, the historical relationships and ways of doing business between publisher and agency still exist, particularly further up the seniority chain. It would have been difficult to penetrate and I’m guessing that this, in part, is why Blackwood Seven’s decision to divest media trading was taken.

The models have been around for ages – it’s the tech that has changed the game.

The use of AI in models of this nature is now both present and future. But media agencies need to give up the competitive fight to self-develop similar service offerings of their own.

The first econometric model I worked with, in a media agency, was in the early 2000s. Seems a long time ago, and it was. Regression analysis was used to calculate the best media flight and weight scenarios and predict the effect. The inputs were limited pretty much to above-the-line, paid media, and the weather.

As a media planner/buyer/manager, I’d spend large chunks of time fiddling with weekly TARP levels (or TVRs as they were called in the UK) in the excel-based model. And then it was off to the client, with a ‘hey presto! The model says that this is the best mix’ presentation.

The tech has changed the game, but not the fears

Since that time I’ve been involved with different networks in the development of, selling in of, operating of market mix models, last-touch attribution models, multi-touch attribution models. The agencies were always very fearful of outside players in this space coming in, for six primary reasons.

  • Loss of additional and profitable revenue gained from model building/data and tech services
  • Loss of ability to build stronger ties/increase client retention by virtue of having the model built within the agency
  • Loss of traction and influence with clients
  • Loss of ability to charge out chunks of a traditional agency team, such as media planners
  • Potential exposure to inefficient or ineffective media buying to date
  • Loss of control over media budget allocation

This last was the most potent concern. And to be fair, in some circumstances, the fear is justified. If, for example, a client insisted on incentivising based on ever-cheaper advertising, how could agency-publisher deals that achieve cheap advertising be managed if the allocation of dollars to one channel or supplier over another could not be controlled?

Agencies have strived for years and years against these threats by trying to build their own modelling services. With, it has to be said, and at best, ‘varying’ degree of success.

Agencies still grapple with the same challenges

When working with TrinityP3 clients, questions around modelling solutions are common – they arise in pitches, in commercial and output assessment work, in projects we run to help organisations take the right path forward with data-driven marketing solutions.

The central challenges of agencies trying to do it themselves, as expressed to us by clients today and as seen in my own personal agency experience from the early 2000s onwards, have never really changed.

  • The struggle to find the right talent truly capable of building out such models, with knock-on effects across data validation, limits to model inputs, sensitivity and big-picture accuracy
  • Lack of trust from clients given the potential for conflict of interest – the agency driving both the media dollars and the model that tells everyone where to place those dollars to their best advantage
  • Cynicism from marketers fearing the ‘lock in’ – the agency builds the model and becomes stickier, meaning that removing the agency, if it becomes necessary, is harder
  • Clunky process, expensive build-outs, long lead times, lack of ability or capacity to properly refresh or iterate models with enough regularity to maintain relevance
  • Complete absence, or a lack of, true AI capability inherent in the models. Experience in more than one network suggests over-stretched data teams trying to cope with media planners wanting to run high multiple scenarios to try and find ‘the best approach’. Clearly, on any media plan there are virtually infinite combinations and so without AI running the scenarios, the ‘best scenario’ remains a human guessing game.
  • Lack of ability in agency buyers to clearly articulate model findings or recommendations as part of a media plan with enough gravitas to convince clients

I’m not suggesting that agencies have not made strides forward with these challenges. In isolation, I’m sure there are a few good case studies around. But none of them, as far as I can see, come close to the specialist providers.

Embrace the change – and the advantages

Agencies should instead be embracing the objectivity of an external, specialist model provider. At least, those agencies who desire true objectivity should. As someone said to me recently, the response of an agency to the entry of an external modelling supplier can actually be a good indicator of integrity and a true desire to spend the media dollars in the best possible way for outcomes, rather than outputs.

The adoption of a competent independent modelling solution is going to provide a single source of truth for agency and advertiser – across the full range of marketing and commercial inputs (not just media or advertising). A smart media agency would be embracing the technology and the potential it offers, and also using this as a lever to re-negotiate commercial terms in such a way that they get paid against value and outcomes, not against cheap media inventory supply.

The source of truth extends out to advertising agencies – a strong AI model is able to properly ascertain the right flighting mix of creative messaging type and execution.

We still need the humans

None of this means the death knell of media strategy and planning. It means automation of choices made in day to day placement, freeing media strategists and planners to look upwards and ahead.

Human intuition is always going to be valuable in driving new ideas and approaches, determining the right cultural fit of a sponsorship opportunity or a media first, or generating bespoke ideas to build out a strategy. A model, in this context, generates only one output, and ultimately, provides the ability to test, learn and evolve.

The same goes for advertising creativity. Creative execution and iteration should not cede to a predictive model – the model should provide the best route in and iteration of how (in which channels) any given piece of messaging connects over time with consumers.

The tech is the enabler, not the originator; as such, the decision about creative strategy and approach remains very human and always should.

Build for the Future

Whatever happens, holistic AI-driven modelling is here, has been here for a while, and is surely going to play an ever-increasing role in marketing over time.

Agencies and marketers can licence and manage the front-end platforms, use the solutions and drive better results. But in terms of the actual model build and iteration, the best way to apply the technology is surely via a solution built by independent experts, rather than by the existing media agency.

Media continues to be the single largest budget item for most advertisers. But media has changed significantly. Find out about our media solutions here