The Deep Dive
Media attribution faces turmoil. Hamilton AI has landed, predicting sales before ads run
You need to know this:
- Blackwood Seven is touting AI-powered marketing mix modelling that it claims delivers 20-25 per cent uplift on media spend and can predict sales outcomes before a campaign even launches
- Getting serious traction in Europe, now formalised JV locally with indy shop One Small Step Collective to attack ANZ
- Claims it can determine where money is being misspent
- Says cookie-based attribution is in trouble
- Its primary business is direct to marketers
“On average, we see 20 to 25 per cent uplift. But if the client implements all the insights coming from the models, then it could be even higher”
Swapping agencies for martech
Blackwood Seven started off buying media for clients six years ago while beginning to develop technology to enable smarter decision-making. 18 months ago, the firm decided to double down on its AI-powered media and marketing platform. Now it’s selling off agency operations around the word and concentrating purely on the tech.
“We do not buy media for any clients globally anymore,” says co-founder, Thomas Bertelsen. “We are a pure software and technology company.”
Its promise is to enable marketers to “predict, execute and evaluate sales returns in real-time, across both on- and offline media, from their own laptops”.
And by shining a light on which channels are performing - and which are not - the company says it is “laying the foundations for a new and transparent industry”.
Bold claims. Could it disrupt the future of media planning and buying? Bertelsen is convinced it will.
Showing how spend equals sales
While some agencies plug into Blackwood Seven’s platform, its business is primarily direct to advertisers. Either way, Bertelsen claims it provides a single source of truth for the impact of marketing and media spend on sales.
Marketers, he says, “are living in a world where you have enormous amount of data insights. But in reality, the advertiser is more confused than ever, because what is actually the truth? What is the navigational instrument that actually tells me what to do?”
Disparate systems and platforms, he says, tend to recommend optimising towards their particular specialities.
“So, we see a big need for a platform across all media disciplines, all media verticals and across the whole marketing [sphere] that actually monitors and quantifies the impact your different investments, or different marketing activities, have on sales,” says Bertelsen. “That is why we see a need for Blackwood Seven.”
"Clients are using 40-50 different television stations on an average campaign. We see just as big a difference from the best performing TV stations to the worst performing TV stations … as between TV and search"
A lot of other ad tech and martech providers are promising similar ‘holistic’ insight. Bertelsen says there’s a difference between promise and delivery. He says most ‘multi-touch attribution’ (MTA) systems are cookie-based – and cookies are on their way out.
“MTA is facing a huge issues [due to] the walled gardens of Facebook and Google, with the wall between the big tech companies including Apple, that won’t allow their browsers to register cookies,” says Bertelsen. “The pool of cookie data you can get from the market is extremely limited right now, which is causing a lot of problems for MTA.”
Blackwood’s system, he says, “is not cookie-based at all”, but based on “probabilistic modelling” of impressions, clicks and media spend – turbo charged by AI - and tying it all to sales.
How last week’s TV ad shifted merch
While marketing mix modelling is not a new concept, Bertelsen says the company has made it “more or less real time” via its Hamilton AI platform.
“Most of our clients update data on a weekly basis. So [within one day of updating] they know how their TV campaign affected sales last week. So, it’s much more operational, dynamic and granular,” claims Bertelsen.
Moreover, it means instead of focusing on media channel comparisons, marketers can better understand optimisation within channel subsectors.
“For instance within television, our clients are using 40-50 different television stations on an average campaign. We see just as big a difference from the best performing TV stations to the worst performing TV stations … as between TV and search,” says Bertelsen. “So there’s just as much room for optimisation when you look on a publisher level as when you look between channels.”
"Typically, you will see that 10 -15 per cent has been spent on channels that do not perform. That percentage of budget should be redistributed. Then, we basically see that there might be a tendency to over invest in certain digital channels"
Give more data, get more sales?
By tying media back to sales, advertisers and planners find that some channels are overweight, some under valued. Bertelsen says clients can also “ask the platform to recommend the next media plan by [inputting] data about specific sales [targets] you want to reach or a specific budget you want to use - and then it will optimise the media plan on insertion level.”
He claims the results are significant.
“On average, we see 20 to 25 per cent uplift. But if the client implements all the insights coming from the models, then it could be even higher,” says Bertelsen.
After getting the media mix right, “all the other data variables in our models come into play”, he says. That can be everything from weather data, to macroeconomic data such as unemployment rates, to how competitor activity is affecting sales, to product pricing - and any other data marketers are willing to share.
Bertelsen suggests it is in their interest to share as much data as possible.
“You have to be very structured and very disciplined when data onboarding because we need a couple of years of historic data and that is naturally also a journey for the client, giving us access to that data,” he says.
But he says going through a “disciplined” data process with marketers can shine a light on problems they may not have known about.
“That is beneficial, because if there are quality issues with data that the client is unaware of, then we will definitely nail those quality issues and solve them with the client.”
Predicting sales based on media spend
The ability to predict sales based on media spend gives marketers a powerful tool to take to the CFO.
“Three months before a campaign launch, you have the possibility to simulate the campaign and [determine] the ideal way of buying and allocating media. So upfront the client, together with the media agency, has a very good understanding of how the media allocation should be done and the sales that will be generated,” says Bertelsen.
“That means being able to approach the CEO or CFO saying ‘The $2.5m we are going to spend in August will sell 9,000 subscriptions or 100,000 products if we spend it this way,’” he claims. “Then if the CFO says, ‘no, you can only spend $2m’, you can quantify that as well. That is a very, very powerful insight that naturally makes the marketing department work a bit differently - because you can actually monitor a campaign’s effect [on sales] before the campaign is launched with a very, very high certainty.”
“It won’t wipe out channel planners”
Bertelsen asserts that the Blackwood platform will not replace media planners and buyers. He suggests it frees them up to add value, to be “more strategic and creative” instead of wasting time buried in spreadsheets.
“[It moves] the media agency away from doing a lot of number crunching to being more strategic as buying partner saying, ‘Okay, this is what the platform says - based on historic insights. What has changed and what therefore do we recommend to change in the platform’s recommendation?’”
Armed with a better view of how niches within channels are performing, Bertelsen says the platform’s insight also enables planners and buyers to negotiate in different ways with publishers. Instead of having conversations about reach, frequency, eyeballs or clicks, “you can negotiate on [the channel’s] ability to deliver sales”.
“So it will change the way you work but it basically gives the channel planner very, very good information about how to optimise campaigns going forward,” he says.
“Our platform does not physically do the media insertions, you still have to book the thing. It’s an analytical tool that gives you a lot of recommendation and insights, but you still have to implement the learning,” adds Bertelsen. “Therefore [agencies’] work will change - but it will not wipe out the channel planners. The platform can just inform their decisions a lot.”
“If we get data on creative, then we can separate the effect of different creative - and that is really, really important because it matters so much”
Outsource functionality, not responsibility
After two decades working within media agencies, Bertelsen says the ongoing conversation about in-housing media misses the point if people still don’t fully understand how their spend will drive business outcomes.
“I’ve been the CEO of a media agency and I never understood why a company spending $50m just outsources their whole decision. They do not have a clue what that generated for the business, or if it generated [anything] at all. Media is a significant share of the marketing budget, and the lack of knowledge about how it [generates ROI] … I cannot understand it,” says Bertelsen.
“I just see it as cost control: If we spend something, what happens? If you ask the sales guy, ‘if I give you 10 more sales people in this market, what would the outcome be?’ The sales director will come up with a very, very precise estimate of what the outcome will be. Marketing needs to deliver the same.”
Highlighting what's missing
Yet when the platform is deployed, it doesn’t necessarily recommend radical changes in channel mix, says Bertelsen. But it does pinpoint where money is being wasted within channels.
“Typically, you will see that 10-15 per cent has been spent on channels that do not perform, that you need to skip. That percentage of budget should be redistributed. Then, we basically see that there might be a tendency to over invest in certain digital channels,” says Bertelsen.
“So, it’s not [always a case of] actually moving money from the offline to online media but basically at least redistributing the money within digital channels.”
Good creative nails social
One of the platform’s consistent findings is that creative makes a disproportionate difference in social media versus other channels.
Social media is “very important” for many clients, but its efficacy is “extremely variable … because we see that the creative element is much more important,” says Bertelsen.
“Two advertisers can spend the exact same amount doing almost the same thing but if their creative is very different then one can have a phenomenal social program and the other find it’s not worth it.
“If we get data on creative, then we can separate the effect of different creative - and that is really, really important because it matters so much,” says Bertelsen.
“As long as we can get data on it, and historically which creative has been used, we can separate it in the model and we can quantify the effect of each creative.”
"Some people can become concerned about what it does to their business model, their agency. But if they look at this as the opportunity to actually get better and actually improve what they’re doing, it plays a really strong role in housing that whole media-client relationship as well"
The local angle
Blackwood Seven has been working with indy agency One Small Step in Australia for the last year. Now the two have formalised a joint venture.
MD John Williams says the next step is to roll out MMM and Hamilton to verticals that will quickly see ROI. “We see that’s really strong within banking, finance, insurance and telcos,” says Williams. “Automotive has been a strong growth area for Blackwood in Germany and we see some big opportunities for efficiency and optimisation within the retail sector,” he says.
So how does all this visibility and predictability – and planning – go down with local media agencies?
“We’ve seen different reactions,” says Williams. “But ultimately, it allows them the space and opportunity to drive media a lot harder. Blackwood Seven allows you to plan channels and get down to a publisher level. So it takes away a lot of the grunt work – but it doesn’t take away the need to look at how we deploy and when we deploy,” he suggests. “So, ‘How can we actually incrementally add value to this media plan? How can we drive this beyond what the actual the Blackwood Seven platform is actually pushing out?’”
Nevertheless, says Williams, “You’re right. Some people can become concerned about what it does to their business model, their agency. But if they look at this as the opportunity to actually get better and actually improve what they’re doing, it plays a really strong role in housing that whole media-client relationship as well”.
Cool story, where's the growth?
Using AI for complex modelling that delivers 20 per cent uplift and predicts sales results off the back of media spend sounds like marketing nirvana. But if the Blackwood Seven platform is as good as its proponents would have us believe, why isn’t it bigger than Facebook and Google and why are media agencies still the go-to guys to manage planning and spend?
“It’s still very early days for platform-driven media mix modelling. Clients are currently getting part of it in an non-dynamic way, as part of a media review or some modelling from different agencies and getting basic results,” says Williams. “The natural next step is moving into this dynamic space. They are getting into it a lot more quickly now, but we are probably 12-24 months away from [Blackwood’s traction] in Europe.
“As we build case studies, I think we will see more of those clients coming on board and start to utilise this and integrate it into their brand management and media management," he says, “and also integrating really strongly with their media agencies.”
Want more? Here's what Blackwood's AI feeds on
“Naturally, we need a lot of media data and that can come from the client or the media agency ... Most of it, we can pull [via] the APIs for all the digital buying platforms,” says Thomas Bertelsen.
“What we also need is marketing data. We need to know when the clients send out newsletters or text messages or push notifications - all that stuff. Basically, all the marketing and media activities that they are doing in order to try to increase their sales, we have to have data upon.”
It also requires product data.
“If suddenly, you innovate your product and then [sales] double, that should be taken into consideration because otherwise marketing might see that as a result of marketing and it’s basically just product innovation. So we need information on products and pricing - because you can also stimulate sales by reducing your prices. We need to take that into account.”
Some models take in more than 200 data points, says Bertelsen. But input is theoretically unlimited. “All the data sources you could imagine,” he says. “We put that into a model and then basically the model finds out what has had an impact - and how much.”