Creative Production

Improving campaign effectiveness for lower cost – the power of creative scoring with AI

Procurement Conversations

“AI shouldn’t only be of interest to marketers, it should be at the centre of procurement conversations around creative effectiveness”

Measuring creative effectiveness isn’t new, even in the early days of advertising, creatives would have two versions of ad and then debate which they thought was going to be more effective. With the move to online campaigns, this evolved into AB testing with data showing which assets got the best engagement, making creative choices less subjective. However, it was still a binary choice between A or B. 

Then the media world came up with Dynamic Creative Optimization (DCO), for campaigns that are already live. It involves putting a selection of assets in front of a specific audience and using machine learning to understand which of the five, six, seven, or more assets gets the best engagement. If that’s done well, it can be powerful. But it’s also quite tactical because it’s only operating in that moment.

Understanding which advert is working well for a specific audience, at a specific time, doesn’t give much strategic insight into knowing what should have been in that ad in the first place. There is obviously market research, consumer testing and vox pops that an agency can try, but the goal is to bring some objective data into the process. 

You can hear more from Lindsay Hong speaking about Creative Scoring and AI at the Marketing Procurement iQ Conference (Europe) in London on 16th April. Register Now!

At SmartAssets we tag content, so it’s no longer at the asset level but rather at the creative-component level within each asset. For example, is it a man or a woman? Indoor or outdoor? What animals are in it? What are they doing? What emotions are present? What feelings can we pick up? What colours are there? What brand logos are shown? All of the crucial elements that creative strategists will be considering when they’re putting together their master campaigns, as well as the variations that post producers will be looking to amend in subsequent assets.

We’re able to look at all the assets a brand has ever run and start to see, based on their own data, what has performed well historically, and more importantly we can understand why, and correlate this back to granular creative components. For example, if you’re trying to sell a laptop to students about to go up to university, should the creative show an academic setting or do you want an outdoor or home setting? These are the kinds of considerations we can now understand based on all the available brand data and ad history.

The key aspect is that, with AI, the future of creative scoring is going to be predictive. DCO generally works better with long-tail and longer standing campaigns because it needs to be running for a while before there’s enough data to really understand what’s happening. With DCO you’re only looking at data you’re accumulating on the placement of that asset. But with AI, we’re looking at a long history of available data. 

The future is going to focus on leveraging a brand’s internal intelligence – proprietary creative content and associated performance data – to better understand their consumers. Leveraging this data at scale will allow brands to become predictive, rather than reactive. 

Three types of scoring

AI can go beyond performance, and there are at least three ways we use it to score content – brand, platform and human psychology. 

1. The brand side is simple. It’s asking whether the right version of the logo is in the right position on the asset and for the right amount of time? Is the partner logo displayed correctly? Does the brand voice come across properly, for example: is it natural and outdoorsy, or perhaps more corporate or formal?      

This is where a lot of the problems come from, getting the right brand consistency, particularly when CMOs need so much content and it’s being sent off to low-cost production houses all around the world.      

2. Then you’ve got platform adherence and best practice. So many assets get created but then never go live or never get seen by anyone because they fail at the final point of activation, because they don’t meet the platform requirements. Again, sometimes they meet the requirements to go live, but they fail on best practices, such as TikTok videos including eye contact. With SmartAssets, you can automate platform compliance, ensuring that every Facebook ad you ever create is in the correct dimensions, which alone you think should be pretty straightforward, but it’s amazing how sometimes those things get lost through the production process.
3. There is also the human element around emotion, memorability, saliency, and behavioural science. We build algorithms off the back of the latest research papers in these areas. For example, there is usually an emotional journey in a video, which is very important to keep the audience engaged. If you have too much variation in emotion, it becomes stressful, and people drop off. If you don’t have enough variation in emotion, it becomes boring, and people drop off. So, there’s a sweet spot for the amount of emotional turbulence that you have in an ad, and AI will help you reach it.

AI content scoring in use

A practical example of where AI scoring would be very useful is global brands, with central brand guidelines, allowing markets to operate their own strategy so they can act quickly to follow market trends on fast-moving channels. The problem is, you give them the central guidelines but inevitably, without some central control, those markets might interpret those guidelines slightly differently. 

For example, the central brand objective may be to target the brand towards a more luxury consumer, with a more elevated positioning. But your people in-market may be just trying to make sales and keep their margin by getting assets produced as quickly as possible and out on to social media. The quality of those assets might not be of the quality that reflects the luxury experience and supports the global brand ambition.

This is why it is a good idea to have a centralized system to put all assets through in a very quick automated fashion and say ‘yes’, this is the brand objective or ‘no’ this does not adhere to the objective. With AI scoring, you can quickly identify when assets are not up to scratch and therefore will not achieve campaign objectives.

Procurement gains

AI shouldn’t only be of interest to marketers, it should be at the centre of procurement conversations around creative effectiveness.

First of all, it empowers companies to reduce their spending on DCO. You don’t need to produce as many assets because AI can predict how content is going to perform. On top of that, you don’t have to put a big media budget behind all of those versions. When you start to learn and predict what really works, you avoid the time and cost of creating a plethora of unnecessary content, which is just so wasteful. 

At its base level, AI tags content so you know which assets you have so there may not be a need to reshoot a picture or video since you may well already have the creative you need. AI also allows you to harness existing and new data to look at how well the content has performed.

You may find a series of shots of lemons in a bowl in front of a gorgeous backdrop and then AI will tell you which one has, or will, perform best. That can save creative teams so much budget not having to reshoot what they already have or relearn what they already know.

We have to be clear that there is a human labour cost saving here, too. But it doesn’t necessarily mean cutting your creative or marketing team, it just means people can get more done more quickly and they’re not reliant on hard-to-manage outsourced suppliers      making terrible creative that isn’t on-brand and is draining budgets.

Technology is doing the job, but it’s being led by talented people. That’s the crucial aspect – it’s helping your best people, not replacing them.

Interested in this article? We have much more on this topic and more at our next Marketing Procurement iQ Conference (Europe) on 17th April, Kia Oval, London. Book Now!