Ads or Algorithms

This week, while the rest of the industry was in post-Superbowl creative ad fever Wavemaker, the GroupM media agency announced a breakthrough. They’ve used artificial intelligence (big tick) to build a tool (called Maximise, obviously) that can analyse gazillions of media plan options in a nanosecond (I may have got those numbers wrong).

The news broke in ‘Campaign’ on Monday, in a piece by their media and technology editor Omar Oakes who by Thursday was speculating on whether media agency brands were becoming an endangered species as planners migrate to tech businesses and media sales operations.

The media consultant Tom Denford of IDComms wondered aloud if the Wavemaker move signalled the end of media planning.

And the creative agency Mother revealed details of their entry into the media agency business with Media by Mother, which aims to ‘rethink the media agency’.

This paragraph from their ‘manifesto’ caught my attention:

“Media planning needs to be about business and content planning. That means analytical talent who create better inputs for media and understand attribution. It means having content creators who can turn around a meme because they know culture. It means copywriters working within creative management platforms so banners aren’t the shitty posters of the internet. It means experts in media buying platforms who can work like producers, managing activation through the right access point with a deep knowledge of ad tech and data.” (David Gaines, CEO, Media by Mother)

Going back to Wavemaker, media agency networks need tools. They’re a key ingredient in the cement that holds the network together, they’re the answer to the inevitable question – ‘but what differentiates you from the others?’

They also define a way of working. In the eyes of those at the centre, if the local markets all have the same tools they all work consistently. A key requirement for international clients.

It’s never that simple.

Will all local markets understand and use the tools? How will the centre check? How and how often will they train? What will they do if the tools aren’t used?

These aren’t hypothetical questions – we faced the same issues building the Carat network way back when.

There are certain lessons that have stood the test of time.

First, if the local market doesn’t understand the tool they won’t use it. The first time a client asks them to explain the inner workings, and they can’t is the last time that tool will be used.

In another lifetime I once had to present how we estimated cross-media reach to a large client (I was my agency’s head of media research at the time). Being a smart-arse, I put up a long formula to demonstrate 1) how clever we were and 2) the scientific rigour behind our approach.

My line was ‘I’m sure you don’t want me to take you through the formula..’ Unfortunately the client had a deep and abiding interest in statistics and asked me to do exactly that. I couldn’t, looked and felt like a fool and never made that mistake again.

Incentivising offices to use the toolkit can work well. At Carat we used to visit local markets, ask to see examples of the tools in action, check their use with international business, and adjust local bonus payments accordingly.  

Tools are great, but the clue is in the name. You need to have smart humans to work the tools. Smart humans need training in why they should be doing this when local market custom and practice dictate otherwise.

The danger with the Wavemaker approach is that it will stop people thinking about what they’re really trying to do – which is not to maximise the media numbers (a means to an end) but to use commercial messaging to drive business success for the client.

The alternative approach – followed by Mother, and agencies like MediaHub and VCCP is to bring media planning in far closer to the creative process.

It’s a question of whether you focus on the ads, or the algorithms.

  1. Hey Brian, long time no see but I always read your entries with interest!
    As the person who created the Wavemaker platform in question, I feel compelled to bring my two cents to the debate. If I may be direct, the notion that the choice is between maths or creativity is a bit dated in my opinion. The impulse for us to create Maximize was to be able to deal with the vastly increased complexity of planning in a world where clients need us to plan against many smaller “tribes”, rather than the good old HW25-55 of TV days. Cloud computing and algorithms can do optimizations in 6 or 7 dimensions that the human brain just cannot handle. Now what that means is that we are saving huge chunks of time for our planners, time they can dedicate to… creativity in content, or ecommerce strategies for our clients for example. So, it’s really a case of maths “plus” creativity, rather than “or”. I’m sure even my very good friend Dave Gaines at Media by Mother won’t disagree with me, since he called me this very week to discuss the merits of algorithms .

  2. Hi Stephan and thanks for taking the time to comment!
    As you’ll know from reading the post a lot of this is not about your intention in creating the tool but in how it’s used.
    If you can ensure that your network does as you suggest, and use the tool not as an unthinking optimiser but as a way of saving time to be spent on creative endeavours, then great!
    As I said in the piece, tools need intelligent humans to work them.
    I’m sure you have many of them and that you have plans to ensure your tools are well used by all of them!
    All the best to you and yours.

  3. I think it looks great. It only answers a tiny part of the challenges of marketing & advertising, but it still looks very useful…agency tools for optimising reach across multiple audiences are still archaic and slow to use, and if this progresses that forwards, more power to their arm.

    The question remains whether the ‘redeployment of time to creative/strategic tasks’ – without aggressively defending that agenda it is more likely that local operation would use the efficiency generated to recoup the investment, drive lower costs and recapture depressed margin, which is understandable in the circumstances but continues the process of narrowing and commoditising the role of the media agency.

  4. Hi Matthew
    There’s creating it, and then there’s using it.
    I don’t doubt it works well and in the right hands it will be very useful.
    My issue is around training and ensuring it’s used, and properly.
    As without those bits it’s as much theory and fluff.

  5. The worst smart-arse mistake that I ever made was when trying to casually name-drop Bayesian probability into a conversation with a prospective client.
    “I haven’t thought about Bayesian probability since I finished by Maths PHD at Oxford,” he replied. Needless to say we didn’t win this piece of business…

  6. Thanks Richard – really cringe-making!
    It happened to me the other way round a few years ago. Going in to see an (unnamed) media agency techy guy with a colleague at a start-up I was helping, he sat us down and then, looking straight at me as the consultant interloper said: ‘Let me just explain to you how advertising works…’. Never have I been more tempted to use the words ‘do you know who I am?’ or ‘I wrote a book about that once..’. I resisted sinking to his level!
    I am pretty sure he’s no longer around. Must have been the curse I put on him!

  7. ‘… media agency networks need tools. They’re a key ingredient in the cement that holds the network together, they’re the answer to the inevitable question – ‘but what differentiates you from the others?’’
    As ever I think you have hit the nail on the head: the imperative for tools comes from the commercial imperative of agencies much more than the client need for effective solutions.
    I recall Andrew Robertson saying that he’d never met a client with the budget he needed to meet his needs; our job is to use creativity to bridge that gap. Often that creativity comes down to making choices: we can’t afford to do everything could, so let’s do some bits extraordinarily well. ‘Sacrifice and commit’ as Kevin Brown so eloquently puts it.
    The name of this tool, “Maximiser” and Stephen’s comments suggest to me that all this power is being focussed on analysis of reach numbers of small ‘tribes’ rather than communication effects. Even if maximising reach is the most effective media strategy (big question!), the usefulness of this tool will hang on the robustness of media research amongst these ‘tribes’ against the smallest of media channels.
    Maybe focussing creative thinking around what drives effectiveness could deliver greater returns for the client, if not the media agency.

  8. Thanks Jonathan and yes I agree with you 100%.
    I can’t ever imagine a client saying: what we really need is an AI tool to estimate reach amongst ever more tightly defined tribes.
    On the other hand I can imagine a head of product at a media agency reviewing his toolkit, assessing where his agency sits against the competition and deciding to invest in a new tool.
    I’ve done exactly this when I was at Aegis, prior to moving internally to Carat. The issue was always to come up with a way of working that could be applied across our comparatively new network, to bring us together. I recognise Stephan’s challenge, and his actions.
    You also make a very good point about effectiveness. I work with a N.American business who build predictive models of effect. They were amazed that a media agency should appear to be focussing in this day and age on reach as opposed to effect.
    There’s also the GIGO argument. I don’t know what data goes into Maximiser but I do know a fair bit about what’s out there. Much of it is not robust…!

Leave a Reply

Your email address will not be published. Required fields are marked *