The Big Data Debate

Posted by

One of the big assertions that have been made by people working in the social media Analytics space is that social media data can be used to predict individual purchasing behaviour.

At an intuitive level this is appealing. If my friend recommends a product then if I have the same need I will be more likely to consider that product as a solution to my need. Furthermore, I may signal that I have that need by Tweeting that I have (or indicate I am likely to have) such a need.

But from a brand owner’s perspective there is still a fundamental conceptual issue here, as well as a whole host of practical issues. The most obvious practical issue is “Which database do I use to manage the customer?”. The only set of universally-correctly linked set of data is the one that is owned by the individual themselves. Only they know the myriad of on-line identities that they have chosen for themselves.

Brand owners are very aware that whilst systems such as Facebook and Twitter have millions of users, there are still people out there (and we have to remember that on a global basis it is still the majority) who do not, even though they may be economically active in many other ways.

This leads to a second practical issue that is often glossed-over but which is a fundamental prerequisite – how can I link together data from disparate systems and know that I am linking the right data to the right individual? There are many ways it can be done but it adds time, cost and complexity – all of which undermine the very concept of achieving ‘close to real time response’.

What is more, however, these issues are common to every marketer – there are, at the end of the day, only one set of individuals (even if they may well have multiple on-line personalities). So in principle it makes no economic sense for each company to create ‘a single view of the customer’ because at best whatever they create will always only be a partial view. Each company that is trying to sell goods/services to an individual is only interested in information that influences their needs for those particular types of goods/services. The volume of data about other areas is just noise that needs to be simplified – at least in their view, that is. Yet as the social network data experts will tell you – seemingly unrelated data can be connected.

My purchase of an airline ticket may well be of Interest to the coffee shop chain that I use because they will know that at some point I will be going to the airport. So they can trigger a communication to ensure I use their outlet/brand of coffee at the terminal.

Which brings us back to my original assertion – what we are describing is a multi-dimensional view of the individual, not a “single customer view”. But generating that multi-dimensional view can only ever be done by an independent direct marketing (data) company. There are already many out there – and the strongest players will, ultimately, prove to be those like Nectar and Tesco who already have millions of individuals on their databases.

Unfortunately there will be many discussions ahead about data privacy, data linking, data cleaning, meta measures, and many other detailed considerations and that the average marketing director will find extremely tiresome. All they want to do is get their message across to the right people, at the right time, in the right way. Nevertheless these detailed issues will have to solved / handled if the concept of the “connected world” (as trumpeted by companies such as IBM) is to become a reality in any meaningful way. Needless to say, for the best solutions to these problems we should look to companies such as Amazon as the most likely source of the innovation rather than IT manufacturers. Their views on the world are fundamentally different.

Meanwhile, back in the real world of today, what should the marketer do to achieve his/her more immediate objectives?

The first thing to do is to make sure that you partner with a good direct marketing company – and one that spans both online and offline communications. Without this you run the risk of running “silo communications” and that totally undermines the principle of a customer-centric approach.

In addition, make sure the agency has their own consumer database, one where they are bringing together data on consumers from a variety of sources, even if those sources are not necessarily complete. You do not need them to have a full picture of the consumer, just enough to cover the key dimensions. If you need advice on who to choose, please let us know.

Secondly, if you have your own internal CRM customer database then supplement it with appropriate meta-data measures (even if they are only fused or projected) taken from as wide a cross-section of sources as are relevant. For example, in the DIY sector, data on house moves is always useful, as is data on leisure activities.

Better still, although it is harder to achieve in the short term, is to become a trusted adviser. For example, if you manage an airport, do not just focus on information about your own location but become a mine of information about all airports in that region of the world. Establishing a reputation for providing accurate, authoritative, advice about a whole category will make web site attract people back time-and-again and that will, in turn, create deeper customer loyalty and greater revenues.

Thirdly, use econometric modelling at both a strategic and operational level to maximise the returns from your marketing activities.

At the operational level this means creating appropriately-extended customer propensity models from all the measures in your customer database. In particular, calculating an “Effective Net Preference” (ENP) score for each and every customer. This will give you an instant understanding not just of their likelihood to respond to your offers but also what you need to change about your offer (other than a cheaper price) to increase that likelihood and reduce their likelihood to defect.

At the strategic level, it means understanding how much money to allocate to different types of marketing activities by understanding the levels of reach and frequency they generate. This is important because online communities generate lots of seemingly actionable data but often for only very small proportions of the customer base. To know how to optimise the media mix you need to step back and look at the net reach and frequency different types of activities are able to deliver – individually and when working in concert. Econometric modelling both describes the mechanisms involved and also measures and correctly allocates their net sales effects – providing you with the ROI’s from each of the differing parts of your marketing mix.

With this framework in place you can then, overtime, improve the data sources you have, improve your customer and prospect targeting, improve your offer(s), and improve the effectiveness of your marketing communications programmes.

It is not exactly rocket science, more a methodical approach that remains consistent over time but without becoming repetitious. This are the benefits that our clients derive from using our services and if you would like to know about it then please let us know.

The Penny Falls: Marketing Mix Optimisation in a Digital World

Posted by

Look around these days and it seems that many companies – even some of the world’s largest – are finding it hard to know exactly how much to spend on each type of communication. In most cases, however, we find the difficulties are rooted in not being able to “see the wood for the trees”.

Discussions about integrated campaigns, digital media, on-line marketing, multi-channel, multi-layered, multi-faceted, and multi-mixed campaigns get lost in the vast web of competing possibilities for perpetual interconnections.

What is needed is some rational simplicity.

Most people lead “busy” lives and whilst they are, in principle, accessible to every possible means of marketing communication – off-line, on-line, above-the-line, below-the-line, through-the-line and, no doubt, “down-the-line” – to develop the optimum media mix the marketing director must still think in terms of reach and frequency.

Put simply, the role of most communication budgets is to use media (of all kinds) to address three key questions “do they know my brand exists?”, “do they know its features and benefits?”, “do they consider it to be a viable solution for their needs?” If we can achieve ticks in these three boxes then we can move to the higher levels of gaining trial, conversion and loyalty.

The use of differing media, including digital, should be viewed in terms of how effective each will be in helping you achieve those initial 3 key communication objectives.

So the first step is to think about reach.

If all communication channels are worth the same in terms of the connection you could make with your target group then if you were to spend one pound on social media marketing what would be the probability that you would gain an additional buyer? How does that probability compare with having spent that same pound on radio advertising instead? And would the efficiency of both be heightened if you spent a second pound on pay-per-click?

Now add to this the value of repetition. Is one contact via social media worth three contacts via radio? If so, then that changes the cost ratio.

You may be thinking at this point, “but it will vary depending on the strength of the creative”. If you are then you need to take a step back for a moment. All we are comparing here is the relative effectiveness of the channels (singly or, where required, in combination) in delivering the same message. In other words, the equivalent of hearing the communication via the radio versus getting the same message from a friend.

We would picture all this as like using an old-fashioned penny falls machine on the pier in Brighton.  You can choose a number of slots into which you can place your coin – but you can’t be absolutely certain how it will bounce all around on its way to bottom. But what you can be sure of – because you can see it – is which slots are the most likely places to use to get the biggest return – the places where you are most likely to get a payment after inserting the fewest amounts of coins.

And that, in reality, is all you really need to know.

As a real life example, think for a moment about choosing how you might want to travel to the airport to catch your plane. You can drive yourself, go by taxi, by rail, by coach or get a friend or relative to drop you off and pick you up.

In this situation you probably wouldn’t go on-line to work out your personal mathematically-optimum solution as to which of these you would choose. But if you heard a radio ad that highlighted the benefits of driving yourself you might consider that option. And then you might go on-line to compare prices, canvass some opinions, and make a booking. Or you might just think the information provided in the radio ad was enough to convince you.

So in this situation the best marketing strategy is to first get the maximum number of people to consider the mode of transport we wish to advocate. That is likely to favour “display”-type advertising rather than “search”. Both will have an effect, but one is more efficient at it than the other.

Understanding these relative probabilities – and how they work singly and in combination across multiple streams – is at the heart of econometric analysis for marketing. The answers tell you how many coins you need to put into each slot at the top in order to maximise the pay-out you get at the bottom. How they bounce around on the way to the bottom will always be a bit of lottery. What econometrics does is stack the odds in your favour so you can be sure you will get the highest payback from the media budget in your pocket!

Why a Baseline Sales Forecast is more important than ever

Posted by

New technologies have made consumers both better informed and harder to engage. While traditional “one-to-many” mass marketing tools like TV and newspapers remain important, businesses now feel that their reach and influence have declined significantly. This leaves companies with a different and more difficult problem. In a world saturated with media platforms and marketing messages, where attention spans are short and consumers increasingly cynical, what’s the best way to engage consumers and create preference?

Once upon a time life was simple. Choices within the marketing mix were relatively few in number and analysing the past performance and planning the future spend and mix was comparatively easy.

But today the choices facing the marketing executive are numerous and complex. For example, advertising is no longer just a decision between the four major media platforms but now has to take into account the rise of the internet and the explosion of more choice and variety within the old traditional broadcast media platforms.

Now the challenge is managing a mix within the marketing mix – which element of the media mix delivers the best ROI? More importantly, what is the impact of the various elements of the media mix working together?

The old analytics – pencil, graph paper (and then the spreadsheet) don’t provide the right kind of analytics to answer these questions. The reason is that we find it difficult to determine relationships between cause and effect for more than two dimensions.

Just look at the chart above. It shows the variation between just two elements of the marketing mix – price and distribution – and their possible effect on sales. You have admit it is impossible to determine the relationship between these two factors and sales. And it is likely that other factors, not shown in the chart, have had greater impact on sales performance.

How can one tell?

The only safe way of determining the real drivers of sales, and the contribution of each one to sales performance is to undertake Econometric Modelling and determine Baseline Sales – what will be the outcome from maintaining the same marketing mix? Then the results of the modelling, which calculate the relationship between each part of the marketing mix and sales response, can predict how Baseline Sales will change in response to different marketing mix scenarios.

There are four main reasons why this type of data modelling will give you the answers you are looking for:

1. Because the model “thinks” in more than two dimensions
2. Because it identifies relationships that are ‘causal’ rather than just ‘coincidental’
3. Because it can derive the sales effect of each activity, singly and in combination
4. Because it can evaluate and model the impact of changes to the marketing mix and other assumptions on future sales and share performance

Econometric Modelling is at the heart of P&G’s Business Sufficiency program which enables them to look at the outcomes of alternative marketing strategies before they take the decision on which one to pursue.

Now, what’s your excuse for not doing the same?