In an era of uncertainty, marketers must utilise all of the tools at their disposal to make budgets stretch further.
A thorough understanding of marketing activity’s impact on consumer behaviour is fundamental to success. Failure to grasp the complex dynamics between ad spend and response could leave businesses in a vulnerable position.
Econometrics is the answer to this conundrum. It is the language of attribution, the statistical underpinning of marketing mix modelling.
It is also the most scientific and reliable way to discover which levers of the marketing machine can be pulled to achieve a desired outcome – whether that be maximising sales, increasing conversion or attracting a specific audience.
Done right, econometrics gives marketers the answers to crucial questions:
- Which channels are driving sales?
- Which are virtuously impacting each other?
- Where does diminishing returns set in for each media channel?
- How does brand marketing affect sales?
- What is the optimal mix of channels to use and when should they be deployed?
For business, the changes in consumer behaviour brought about by the coronavirus pandemic has infused these questions with added salience.
With consumers spending longer indoors and consuming more content, there are undoubtedly opportunities to be seized upon by using the appropriate analytics. Procter & Gamble (P&G) and Unilever have adopted a counter-intuitive approach by increasing marketing budgets to help them understand the changes COVID-19 has wrought.
As market leaders, P&G and Unilever can afford to go against the grain. For most brands, the pandemic means a more uncertain business environment and budget cuts across the board. With marketing resources becoming scarcer there is a greater incentive to make every pound count. Companies cannot afford to waste money on ineffective channels, but nor can they spend beyond what is appropriate on fully functioning channels.
Applying winning data science techniques in the shape of econometrics is business’ best bet for navigating uncertain terrain and ensuring that budgets are optimally allocated to give the best returns.
How COVID-19 changed consumer behaviour
Covid-19 has significantly accelerated an already burgeoning ecommerce channel. During lockdown, huge numbers of consumers flocked to online shopping including those who had previously been reluctant to do so, such as the over-65s.
Even when social restrictions were gradually eased, online shopping stayed strong. A survey conducted by the World Federation of Advertisers showed marketing spend for the first three quarters of 2020 was down in all areas apart from online display (+6%) and online video (+9%).
Yet digital budgets are not immune: according to IAB UK/PwC online spend fell 5% overall in the first half of 2020, largely due to a drop-off in search and classified ad budgets.
Organisations will need to understand the behaviours of the nascent online shopper to give context to this instability. The new lifestyle imposed on many people by the pandemic will invariably lead to digital having a bigger presence at the marketing attribution table. Measuring the correct impact of digital media, taking into account the activity of other channels as well as external market-related factors, requires the scientific tools of attribution.
Another key lesson the pandemic has taught us is that brands must pay attention to the destruction of loyalty.
This has clear ramifications for brand advertising and marketers’ ability to account for changing consumer dynamics. Resting on the laurels of a loyal customer base, even for the biggest brands, is no longer an option.
To understand where businesses should place their bets, econometrics is key.
Good versus bad economics
When it comes to econometrics the devil is in the detail. Only by sifting through large amounts of data from different sources and applying the correct analytical techniques will it be a worthy investment.
In other words, it is not for the half-hearted. When econometrics is juxtaposed with arbitrary attribution models that go off first click, last click or everything in- between, there really is no comparison.
Good marketing econometrics starts from the assumption that the omnichannel consumer is constantly being subjected to advertising. They will sometimes see and hear hundreds of messages before committing to a purchasing decision.
Given this, the question becomes: what is the best way to understand how the aggregate forces of advertising and purchasing ebb and flow together?
The answer is to build a model with the complexity to recognise this multi- dimension marketing environment yet simple enough to produce intelligible and actionable results.
The features of this model and its specification are what separates ‘good’ marketing econometrics from ‘bad’ . Good marketing attribution models have:
Appropriate lag structure – i.e. will test for the existence of lagged independent variables. This feature ensures that the statistician is not restricting themselves to contemporaneous time. In the advertising world there is often a time lag between channel deployment and an impact on an individual’s behaviour. This is especially true with more traditional media like direct mail and inserts.
Brand analysis – marketing attribution is incomplete without considering how brand advertising is behaving in the mix. Metrix Data Science has devised an analytical approach called Brand Ecosystem ModellingTM, which evaluates the impact of marketing activity on key brand metrics and further investigates how this feeds into sales.
Diminishing returns – marketers deal with scarcity of attention. There is a limit to how much channel activity has a positive impact. It is important to use the appropriate methodologies to investigate the spend threshold that attains optimal response results. Diminishing returns requires testing to see if the data fits an upside-U shape i.e. non-linear pattern.
Interactions – the interplay between channels needs to be measured because none works in complete isolation in an omnichannel world. For example, Metrix Data Science has observed the virtuous relationship that exists between TV and door-drops play out for many brands. Accounting for this in the analysis is crucial, otherwise marketing channels are treated as siloed entities.
The case for scientific attribution
Without a scientific approach to marketing attribution, organisations will not have the strategic vision they need to combat the tricky economic times ahead or maximise opportunity during the next growth cycle.
Simply knowing the latest consumer trends isn’t enough. Businesses must understand what changes to shoppers’ behaviour means for them.
They require the tools to determine the correct media mix, to know how to engage prospects and when to deploy those channels. Econometrics, the scientific approach to attribution, is the answer brands need.
Metrix Data Science has helped many clients to navigate the complicated world of marketing attribution. Applying econometric analysis that recognises the nuanced dynamics between marketing activity and sales allows our clients’ marketing resources to go further.
The key reason long-standing client The Salvation Army (TSA) came to Metrix Data Science for attribution analysis was the charity’s shifting marketing focus.
Due to the new regulatory framework of GDPR, TSA had decided to drop cold mail as a marketing channel. It wanted an impartial view of the best media allocation available to compensate for the loss of a key channel.
The novelty of our approach started with data collection. Since TSA only runs campaigns around Christmas, data availability is limited to three months in the year.
Metrix Data Science therefore collected data going back three years to maximise the number of data points. This would not only enhance the robustness of estimation, but also facilitate a longer-term view of media performance; a vital element of the project, since previous marketing analysis was based on short-term ROI.
To understand a multi-faceted environment Metrix Data Science built a suite of econometric models, paying specific attention to model specification for each.
Our analysts produced models to predict the response for each marketing channel. This approach ensured the complex relationship between media spend and channel response was factored in.
Metrix Data Science incorporated the insights gleaned from the modelling into a tool that enabled TSA to predict the outcome of different media budget allocations in terms of donation income, response volume and ROI. The tool was complemented by a plethora of recommendations to optimise the potency of TSA’s media allocation strategy.
Key results –
- Increased donation income by £ 680k for lower spend
- ROI increased from 1.04 to 1.24
- Increased average donation value by 7%