Why Marketers Should Be Wary Of ‘High ROIs’

By Jerry Daykin, Senior Media Director, EMEA, GSK Consumer Healthcare

It seems pretty obvious that marketers should want a high return on investment (ROI). Knowing that they earned more for their business from every dollar they spent is exactly the sort of serious performance target marketers rightly like to achieve. Over recent years ROI has become an increasingly popular measure across the industry for this very reason, so it may surprise you that high ROIs can actually be a real warning sign, especially for businesses looking to grow.

The term is used to broadly describe marketing success but is essentially a simple equation: what was the value that my marketing activity drove for my business, divided by the costs of running it. Simple to say but generally much harder to calculate, not least because some of the return can playback over a very long period of time. It gets even trickier because we inevitably want to understand the separate ROIs of different channels and tactics, many of which will have run at the same time that. For simplicity today I’m considering ROI to be a combination of both short- and long-term revenue impacts, as it is typically calculated.

One of the biggest factors affecting your ROI is how much you spend on media, because this is likely by far your biggest cost factor. It’s almost always true that spending less and buying cheaper forms of media will improve your score, though it certainly won’t lead to a more impactful campaign. Digital channels in particular have benefited from high ROIs in their early days because advertisers invest relatively little in each channel, and these ROIs can be further ‘massaged’ by shifting some of your inventory into cheaper placements within the same channels,

Given that it seems fairly possible to game ROI calculations, or for them to show only a slice of the whole picture, I’ve always had a heavy scepticism towards them. It was interesting to hear the marketing scientists from the Ehrenberg-Bass Institute for Marketing Science (home to ‘How Brands Grow’) raise doubts too at a recent presentation. Interesting but perhaps not surprising, Byron Sharp—Professor of Marketing Science & Director at Ehrenberg-Bass Institute—has previously described ROI as a ‘silly metric that can send you broke’.

Their primary issue was with the methodology that is used to calculate ROI numbers, which in the world of big business tends to mean MMM, or Media Mix Modelling. This econometric undertaking analyses all your media and marketing activity, alongside your sales data, and begins to deduce the effects of it all. Over time, especially when you use channels in different relative weightings or at different times, such models build up a view of how effective any campaign was as well as how well specific channels performed.

Some elements of this however become a self-fulfilling prophecy; the system has an idea of how well certain channels will perform, and that tends to reflect on how well they come out in the next model. As the primary inputs into the system are spend, reach and pricing, it’s also hard for it to take into account other factors such as the quality of the creative used on specific channels as it tries to work out their weighting. There are ‘single source’ approaches to calculating ROIs using direct sales data (e.g. from loyalty cards) and hold out groups to normalise other factors which arguably do a more accurate job, the trouble is their outputs are then incomparable with the numbers from other systems, and highly short term focussed.

There’s a question about whether these models truly work, and if for instance they can then be used to predict future results, Professor Sharp for instance commented previously that “there’s widespread ignorance of the failure of multivariate models to produce estimates with any predictive validity. When I write “widespread” I include many people with advanced degrees in statistics”.

The biggest caution the Ehrenberg-Bass Institute had however was that the models don’t do enough to account for the distribution of light weight users who buy a brand. They end up showing great results for small campaigns which reached easy to activate heavy buyers, whilst showing lower numbers for campaigns which push out to a tougher lighter audience. As anyone familiar with their work would know this is a real red flag as it’s these lighter users that ultimately allow a brand to grow and become more successful, penalising campaigns aimed at reaching them starts to look like shooting yourself in the foot.

Not such a clear-cut measure after all is it? My interest in ROI numbers then goes one step further, to the impact that spending more has on a campaign’s rating and therefore to the ‘marginal ROI’. For any campaign you can draw a simple chart with potential spend along the horizontal axis and predicted revenue up the vertical access. In reality this graph is never a straight line, but a curve that starts steeply and begins to flatten out. The first few dollars you spend are always the most effective, reaching the lowest hanging fruit and cheapest to reach consumers, as you push your campaign out to those light buyers you start paying more to reach them and getting less direct return, flattening out the curve.

The marginal ROI measures at any point on that graph what additional revenue you’ll get for the next $1 you spend, effectively it is the ever-shifting gradient of the graph. As mentioned, it starts very high, as does your overall ROI, but decreases the more you spend. This isn’t a bad thing; it’s simply reality.

What this points to is that having a high ROI, and in particularly having a high marginal ROI, most likely means you are not spending enough money to maximise an opportunity. Rather than celebrating your efficiency you should really be working to unlock more spend within your business so you can invest and grow. Only when the marginal ROI flat lines and your additional investment starts losing you money should a growth focussed business really put the brakes on, and even then you can sometimes justify pushing onwards if establishing the brand rapidly is critical to longer term success.

Of course most businesses simply don’t have enough money to max out every potential marketing opportunity so ROI comes back into play as an interesting tool to identify which are most worth investing limited resources into. ROI data is perhaps most useful at the macro level in helping marketers understand which markets, which brands and at what times to spend their money, though even here a layer of human knowledge always helps unpack raw figures. Truthfully it’s also still more useful to be analysing the marginal ROIs and shape of the ROI graph in each case, rather than just pure numbers.

Overall improving the effectiveness of your campaign becomes less about hitting a certain ROI score and more about changing the shape of that curve. An effective advertising strategy will steeply rise to drive big results from initial investments and then flatten out very gradually allowing you to push your investment out considerably further along the curve. So don’t fool yourself into thinking that by spending less and buying cheaper media types you’re doing your business a favour, drive real value by shaking marketing campaigns which allow you to push out across that curve.

Despite what people sometimes say about the difficulties in proving digital media effectiveness, it’s no different to measuring traditional channels, except perhaps for the fact many advertisers invest too little in it for it to show up in models. As digital channels are starting from a point of heavy under investment they are early on the curve and typically have higher ROIs and scope to increase spend without dramatically dropping those. As spends have increased notably however digital ROIs have started to observably drop across the industry which of course raises questions about whether digital channels are losing their magic. I would argue this is a fairly predictable trend as we move further along their curve, and indeed marginal ROIs may be one way of starting to fairly ‘right size’ digital media spends.

Of course if you want to go further and stop focussing on ROI altogether what are we left with measuring instead? Reverting to their theoretical roots the Ehrenberg-Bass Institute’s team suggested using measures of penetration, mental and physical availability, and strength of distinctive brand assets. Some of these are easier to measure than others, but they give a clear focus to marketers trying to measure success regardless of their tactic or channel. In digital in particular we’d do well to stop filling out endless dashboards with largely meaningless digital actions, and instead find metrics which map to these concepts, even if that means adopting more traditional research approaches once more.

It’s important that marketers are accountable for every dollar they spend, but once again it requires their knowledge and understanding to interpret the various measures.