Digital Attribution Isn’t Enough: Q&A with Paul Dyson

by Lindsay Rowntree, ExchangeWire


Attribution modelling is a powerful tool for measuring marketing channel effectiveness and ensuring channels further up the purchase funnel are given credit for their contribution. However, argues PAUL DYSON (pictured below), founder, Data2Decisions, marketing goes beyond online; and the holy grail is to achieve holistic, connected measurement and optimisation of your entire media investment. Dyson explains to ExchangeWire how holistic marketing can successfully link online to offline.


ExchangeWire: Many advertisers claim to have comprehensive attribution models in place that inform their overall marketing strategy. What’s the reality, in your opinion?


Paul Dyson: Yes, many clients are using attribution models; but this can only ever inform the online/digital part of their marketing strategy. Many advertisers also use offline media (TV, OOH, radio, print) and the majority sell offline (very few advertisers sell 100% online at the moment). So, for these advertisers, attribution modelling fails to account for the impact offline media has in driving people online. This means it cannot measure a true sales effect if sales happen offline. More and more offline media is becoming addressable; and marketing tools, such as loyalty schemes, can help us link offline sales with online marketing. It will, however, be some time – if ever – before we have single-source data that measures all online and offline parts (media and sales) of the consumer journey at an individual level. Measurement and modelling techniques, therefore, need to take both online and offline into account to accurately inform the overall marketing strategy.


What’s your take on the effectiveness of digital attribution modelling?


There are things that attribution modelling can do very well, and at speed. For instance, assessing the effectiveness of different keywords for paid search, identifying which parts of the online journey are inefficient, or optimising the creative message via live testing; but these are focussed on improving the online part of the journey. Where many of today’s attribution techniques fall short is in linking together online and offline worlds. Also, some of the attribution techniques are fairly basic and seem to have bypassed the analytical rigour to which other modelling techniques, such as econometrics/MMM, adhere. In fact, many attribution techniques aren’t really modelling – they are simply counts or impressions with arbitrary weights applied to the different touch points. In this sense, the output needs to be treated carefully. As long as the user is aware of this, they can still add significant value to the digital part of marketing effectiveness


How does holistic marketing effectiveness differ from digital attribution?


Digital attribution is a subset of holistic marketing. More accurately, digital attribution evaluates only part of the marketing function – the bit that happens online. The issue is that if, for example, our TV campaign creates significant traffic online (and it is not unusual to see literally hundreds of thousands of additional searches in the five minutes after a TV ad is shown) then digital attribution methods can fail to recognise this. This gives all the “value” to one or more of the online touch points it measures when in fact this online traffic would never have happened without the TV ad. Traditional econometrics does a much better job of evaluating holistic marketing activity – online, offline, short and long term, and external influences, for example. However, it cannot usually report to the same level of detail that attribution can (e.g. search vs social vs display vs VOD) and it takes time (weeks) to collect data, build models, and analyse results. The ideal is a combination of traditional econometrics and digital attribution to give us the best of both worlds. This has indeed been the focus of Data2Decisions for the past two years, resulting in our Total Attribution approach.


It’s often said that we have all the answers, but aren’t asking the right questions. What questions should advertisers be asking ahead of embarking upon a holistic view of marketing investment?


Paul DysonIf we want a holistic view of marketing investment, the key question to ask is, “Are we accounting for all the marketing and non-marketing influences?” So, what is the role of offline and online media; but also the role other marketing activity, such as pricing, promotions, and competitor analysis? And what is the role of external factors, such as weather and economy? Attribution modelling, for example, might show a much increased sales effect this week compared to last week, but we know we have a new online ad this week, so we assume it is the new ad that is much stronger. However, if the client was also running a 20% off promotion this week (and not last week) then this could explain the increased sales. Marketing effectiveness is a complex process; and the evaluation techniques need to reflect this.


Which media channels/business functions are typically most impacted after implementing a holistic view of marketing investment?

It depends on how effective the different functions have been. Historically, it is the advertising functions for a number of reasons. Firstly, advertising effectiveness is difficult to measure. It has usually a small effect on sales, compared to pricing and promotions, and often the impact can be a few weeks after the advertising, meaning it is difficult to measure with confidence. Secondly, it is a complex area – which media channels to use, when to advertise, how much to spend, which audience to target etc.; and it can be difficult to know how to optimise advertising effectiveness. Finally, one of the biggest drivers of advertising effectiveness is creative quality. It is not unusual for a new creative to be 10 times more effective (or less effective!) than the previous creative and so overnight, advertising effectiveness can change dramatically when a new creative is used. This has knock-on effects for optimisation – we need to spend less on a more effective creative to achieve the same uplift in sales; we should move money from media channels with less effective creatives to the channel with the more effective ad and so forth. All this uncertainty about advertising effectiveness usually means it is the one area of marketing offering the biggest opportunity for improvement.


Are businesses prepared to listen and respond to what the model is telling them?


In general, yes. Who wouldn’t pass up the opportunity to improve effectiveness and returns? However, it is rare that recommendations are implemented to the letter because there are often other (strategic) issues to bear in mind. For instance, lack of budget may prevent us spending at the optimum level; optimisation might suggest not advertising at a particular time of year, but we may have to support a new launch; long-term analysis might suggest significant opportunity for growth, and therefore value, in investing above recommended budgets in the short term. At the end of the day, marketing always has and always will be a mix of skills – we are, after all, trying to influence human behaviour. The best marketers are those who realise this and understand that they have to bring together a range of inputs (modelling, attribution, research, customer opinions, gut feel, available budgets or economic factors) to make the right decision, rather than depending on just one of those factors.


This article first appeared on ExchangeWire