Automation is coming; that much is without doubt. I’m sure you’ve already read three articles this morning about AI, served to you through a social media algorithm. Performance marketers are desperate for some of the action, but living up to the hype has so far proven difficult.
The most successful companies in the ad tech space are becoming more efficient by effectively combining their automation tech with the inventory exchange process.
Within performance, and on mobile, automation is found towards the ad-serving stage, and I estimate over 90% of campaigns are operating this way. They particularly focus on optimising CTRs to drive a more efficient spend, but that’s the easy bit.
At the other end of the scale, this needs to be matched with advertiser requirements. If we look at true end-to-end automation, very few people, if any, have managed to pull it off.
With that in mind, let’s take a look at what the sticking points are, and how we can overcome them.
The ultimate goal for performance marketers is an effective ad-serving algorithm that buys CPM traffic, but delivers CPA campaigns.
This system will learn, very quickly, which inventory is proving profitable for the network as well as the advertiser, and buy more of it. Since it is the growth area for this sort of campaign, we’ll focus here on mobile.
Currently, inputting data still requires a human being to adapt creative campaigns, and set objectives and KPIs. An ideal scenario would see KPIs set by a machine. This machine would analyse results, adjust certain goals and turn them into measurables, before syncing this with an ad campaign.
It takes a brave marketer to hand over the reins at this level. I think the capabilities are already there, but do we trust machines enough yet?
The difficulty faced by those advocating automation is that mobile programmatic inventory is almost always bought on a CPM basis.
Performance campaigns, however, are run with a CPA metric in mind. The joining of those two different ways of measuring presents a real challenge.
It’s risky, as the efficacy of an automated campaign can only be tested in the real marketplace, with real clients. Inevitably, the first attempt with any campaign contains a lot of misses. When you’re on the hunt for sales, low conversion rates on inventory bought at a CPM basis hurt your bank balance.
But with a machine-learning technology at play, the buy can be improved quickly and effortlessly, leading to more conversions. This might mean that the inventory itself is better suited to conversion, or different creative is optimised.
For sophisticated machines, having both is definitely possible.
To achieve better automation, publishers can use a DMP to make their data available to buyers; the further they can push their data, the better.
SSPs and publishers need to work together to normalise data. Increased transparency with the buy side will also help us learn more about what targeting is working, and who can be targeted.
On the buy side, the main demand is for scalability, and, of course, good pricing.
As the programmatic lumascape consolidates, some of these demands will be met. Fewer, larger intermediaries will enable scalability, and more potential for automation.
But it will be those who can create the strongest software who win in the end. In the current situation there are hundreds of companies developing their own tech.
Those with the best data and the most sophisticated ability to use it will conquer all.
As we tie ourselves up in knots debating metrics, technology and auction types, advertisers remain the most interested in using their marketing budget more efficiently and effectively.
Once companies make demonstrable savings on their spend, while upping their final conversion rates, we’ll see larger budgets invested into customer acquisition.
To do this, they’ll need to share more data than they’re currently comfortable with. We’ll need to convince them that it’s also in their best interests to do so. Again, transparency is key.
We’re talking real-time data, accessible by a machine. This would mean advertisers (brands and their agencies) sharing data further down the funnel, to allow the machine to see if it was being successful. If they really want to optimise their campaigns, this could mean purchase data.
Of course, we’ll need brands to be ready and willing to have a pool of cash available for a machine to test with, which is no small ask.
The creative world is currently in a bit of a tizz over its future. ‘Will artificial intelligence replace the creative director?’ is the frequent question from agencies.
While there will always be a human touch in creative campaigns, once individual campaign assets are created, the ability for machines to assemble them isn’t hard to achieve.
Automated ad creation through retargeting is already a reality. Retailers have huge asset banks to draw from when creating personalised ads. Using shopping cart data to automatically build new ads is saving swathes of time in the iterating of creative. Most of these will never be signed off by a human.
The impact of automation
If we see more companies adhering to the above, we’ll start to see some knock-on effects.
As the Harvard Business Review notes, an increase in the economic viability of machine learning algorithms will transform the value of other inputs. Thus, data associated with biddable inventory will become even more important, and first party or unique data will be a valuable resource indeed.
A combination of strong first and third-party data gives you the ability to operate lookalike modelling. This predictive aspect – probabilistic data – is improving all the time, aided by machine learning.
The current dialogue seems to be that automation allows us to operate at a larger scale, to a better degree, and frees up capacity for strategy, ideas and creativity.
In short, I believe our jobs are safe.
Looking ahead we can expect to see a continued push towards consolidation, as the market adjusts itself. Publishers and brands will push for direct relationships, so for those in-between to win out they’ll need the best tech, with automation at the core.
As brands realise that downstream metrics are increasingly easy to measure, they’ll brief-in more performance campaigns. If we can add automation to the mix, we’ll be able to make them super-efficient.
This will only be achieved through an increase in transparency on all sides. Trust will be the key word for 2017 – if advertisers, agencies and programmatic partners can align and work on their mutual interests rather than individual gain, automated performance campaigns will become the norm.