The programmatic debate: From Mad Men to Maths Men
There are valuable lessons to be learned from the financial services industry when it comes to automated trading, writes IMGROUP’s Steve Sydee.
There was a great debate about programmatic at last week’s Automated Trading debate, and it is clear that it is here to stay.
For example, Magna Global, the Interpublic Group media investment unit, recently predicted that the amount of global ad inventory bought and sold using programmatic technology will rise this year by an astonishing 49% compared to 2014, reaching a value of $14.2 billion worldwide in 2015.
What lessons can be learnt from organisations that have already been down this path?
Trade(r) secrets
As the media market shifts towards this new tech-driven emphasis, the sales mentality of advertisers looking to survive and capitalise must necessarily also alter.
A World Federation of Advertisers (WFA) report concerning programmatic argues that a “new approach to digital media trading based on a financial investor psychology provides a foundation from which increased competitive advantage can be achieved”.
When it comes to dealing with numbers, financial services firms know their stuff. Technologically, too, they’re ahead of the curve. Algorithmic trading (AT), which is carried out by computers pre-programmed to follow a specific set of instructions for placing trades, uses extremely similar technology to programmatic, and has been used by the banks for several decades already.
Georgios Antikatzidis, a consultant at IMGROUP and a financial services and AT expert, has identified the pros and cons of the technology in that market. Here, we explain his views on the pros and cons and how they might be anticipated to translate into the new, lucrative but hazardous sphere of programmatic.
Bigger, faster, stronger
First, a resounding entry in the ‘pro’ column.
On the whole, AT has spectacularly increased the frequency, efficiency and accuracy of trading. The robotic traders outrank the human traders in almost every respect. High Frequency Trading (HFT) has enabled trading to take place at a pace commensurate to the dizzying speed of the market itself.
Computers are highly disciplined – not to mention highly informed – decision makers, unaffected by the unpredictable emotions that distort human judgements and behaviour.
[advert position=”left”]
All this considered, the attraction for advertisers of putting AT technology in charge of bidding for relevant ad-space hardly needs spelling out.
This technology can find the audience your product or brand needs and reach them via web-space in a matter of milliseconds. And with new bi-directional TV platforms such as Sky’s AdSmart becoming available and regularly improving the ability to focus at a more granular level, the trackable audience needed for targeted advertising is growing all the time.
Cutting out the middle-Mad-Men
‘Pro’ number two – AT is much more transparent than the models it has superseded.
While the bulk of trades used to happen in the relative secrecy of over-the-phone and face-to-face negotiations, as Georgios says, “We now know what deals are being done, being closed, when, where and between who.”
Now that most trades are a matter of record, there is less chance of fraudulent and unfair practises going unnoticed, to say nothing of the obvious benefits associated with having ongoing access to a huge store of historical data about prior business transactions.
Some observers would argue that the advertising industry could use more transparency. The process of ad-buying has historically been obscured by its dispersal along a chain of ‘middle-men’, all taking their own commission along the way.
In fact, this is still (currently) the case with the programmatic process, too, to such an extent that the WFA estimates that in most programmatic campaigns, “still more than half of the advertiser spend goes to middle man fees.”
Clearly, the advertising industry is a long way off attaining an ideal level of transparency – but many believe programmatic is a step in the right direction.
Risky business
It’s been good news for champions of programmatic so far, but the use of AT technology certainly comes with a few ‘cons’.
Firstly, although the financial markets are always volatile to some degree, volatility is heightened by AT due to the expanded volume and frequency of trades. The resulting, profuse level of reflexive market activity increases instability and risk. When things go wrong with HFT, they go very wrong – and fast.
The second, related ‘con’ that comes with AT is the con-artistry of that familiar villain of the stock exchange: the rogue trader. Markets are certainly open to massive manipulation via HFT technology, and such manipulation (often deliberately) increases market volatility, too.
For example, 2010’s trillion-dollar stock market crash, popularly known as the ‘Flash Crash’, has been at least partly attributed to the (arguably) unscrupulous tactics of HFT traders, including ‘spoofing’, ‘layering’ and ‘front-running’.
In the wake of the ‘Flash Crash’, such tactics were cracked down on by financial regulators. Regulation protecting both advertisers and customers against exploitation is likely to be increasingly required in the industry as programmatic grows in importance.
In fact, increased regulation is arguably already desperately needed, with or without the influence of programmatic; as IMGROUP’s Steve Sydee has written: “Fraud in digital advertising is rife – and there are myriad ways it is carried out.”
Clever robocops
From this standpoint, programmatic could look like just another headache for advertisers to deal with in an already migrainous marketplace. However, it may be automated computer technology that offers the solution to the very problems it has created. New, improved, and self-improving automated technology: Machine Learning.
Machine Learning is a branch of Artificial Intelligence based on pattern recognition. The technology gathers all the data available at that time, in real-time, but also analyses trends over time, and learns from these. The more data gathered, the more refined the technology’s approach becomes.
For the banks, this technology offers a viable tool for applying risk controls to HFT activity, with computers now capable of learning, detecting and even predicting the advent of patterns associated with market manipulation and fraud.
A.I. = R.O.I!
Machine Learning offers a safeguard against another ‘con’ that banks have had to contend with when it comes to AT: over-optimisation.
Over-optimisation occurs when a trading algorithm is too slavishly attuned to back-tested data; this can mean that the algorithm is not sufficiently flexible to respond to market changes and is not malleable enough to be shaped to accord to evolving business objectives.
By contrast, Machine Learning continuously improves its own operation through the constant analysis of data and recognition of patterns. The benefits of using such technology to run your advertising campaigns are obvious: increased, and perpetually increasing, accuracy, with the flexibility to alter campaigns in real-time according to changes in market conditions.
From the point of view of advertisers, Machine Learning doesn’t only offer the above advantages enjoyed by banks that utilise it, significant and welcome though these are; it will enable them – through a process known as ‘True Prospecting’ – to reach new audiences for their brands.
As Richard Foster, writing for IAB UK, explains: “Programmatic buying powered by Machine Learning can operate at ‘pre-awareness levels’ in the sales funnel, finding new customers and driving new engagement and conversions.
“It can exclude targeting users already aware of the brand, and instead model their behaviour and attributes and then find new audiences displaying similar signals outside of the sales funnel.”
This technology will not only enable you to better target those who are already aware your brand, but to target those who have never heard of it, but can be confidently expected to be intrigued by it when they have.
Making sure the numbers add up
Both Machine Learning and programmatic depend upon being fed good quality, accurate data. Barbara Manley, writing in The Huffington Post, having spoken to over 30 different agencies and marketers about programmatic, writes that “Data ownership is critical to most marketers that we talked to…[And] strong interest in any opportunity to enhance data with external sources.”
Georgios says, “Banking is a matter of striking the right balance between risk and reward. Machine Learning and Big Data analytics enables you to take more calculated risks, and to make a better prediction of your ROI.”
The same principle applies to advertising, except that it isn’t only your risks that will become better calculated through data analysis: your targeting will also become more precise.
It’s hardly surprising that, in the Information Age, information really is power; but advertisers need to make sure they’re getting the best information they can and connecting data in a reliable way – and they can’t do this alone.