Is marketing science or art? It seems that the only correct answer to this good old question is “a bit of both”. That being said, its scientific dimension has gained some serious ground over recent years, a fact which the following article explores further…
My first job, when I first started working in marketing some 20 years ago, involved conducting market research through quantitative and qualitative studies, and the statistical analysis of Contact Centre activity. Collaborating with some big firms made it clear that the two factors that determined studies’ reliability and level of consideration were the method used and the quality of the results analysis team (experience, seniority, profiles…). A poor statistical sample and a wrong question wording could indeed ruin a whole study.
How the digital era affected marketing: before and after
Based on this, we could then carry out market research before launching a new product, measure the impact of a campaign and optimise our communication budgets for the next campaign. To do so, teams combined mathematical, psychology and marketing skills…with already a few relatively powerful statistical analysis and geomarketing IT tools.
So, what has changed today? Not much, at least in terms of substance. Nonetheless, digital technology has massively amplified two elements: the amount of data available and the ease with which tests can be carried out (skipping the web 2.0 step here for the sake of simplification…). And any Marketing department that succeeds in harnessing these elements can be said to have jumped on the scientific marketing bandwagon.
From testing to continuous integration
Testing to validate prior assumptions constitutes the very basis of the scientific method, especially since tests must be reproducible and, hopefully, always generate the same result (unless they have helped identify an additional variable).
Testing keyword purchases, title wording, colour scheme or the position of a button on a screen… Anything can be tested at increasingly lower costs, as long as sample size and structure are adequate and consistent.
On a large scale, a really small conversion gain on an acquisition path or in a conversion/sales tunnel in e-commerce can eventually yield a considerable return… This is in fact the key to many technological solutions’ ROI studies (chat window display, screen customisation and presented products, slightly varying prices, deadline to send a reminder email…).
However, increased testing means an increased need to reduce the cost of error or failure in order to make them acceptable, and consequently a growing demand for tests that are easy to set up and measure.
Implying, that when faced with a selection of technological solutions, the one allowing you to perform tests in the easiest way should stand out as the best choice. How much time does it take you to build a new campaign? How much time does it take to deploy a new feature in your CRM tool or e-commerce platform?
The consensus today is that, in terms of technological solutions, anything that does not allow continuous integration, multiple deployments and measurable tests, must be updated or replaced without delay as it prevents you to adopt a scientific approach to marketing.
Understanding causal relationships
We also feel that it is essential that Marketing departments understand their campaigns’ causal relationships in order to be able to optimise them. Data analysis and testing should also help in this regard.
Of course, many relationships are already known from experience: they are what could be named the invariants, to borrow a term from Idriss Aberkane’s lecture on knowledge geopolitics. Just as in geopolitics, energy (resources and deployment ability) and water are invariants that both explain motivation and/or conquest, in the business world, location is an invariant (hotel, restaurant…).
Identifying an additional factor
But the idea has to be further investigated to gain precious insight. Is this one street location as appropriate for a clothing shop as a restaurant? In fact, more than the location, it is the type of trafficpassing by in front of the business that will most likely determine success or failure: which population, on what days and at what time, why? (visit MyTraffic and Flux Vision to know more). This simple example shows that an additional, strongly explanatory (strong causal relationship) factor can be identified using external data sources.
In conclusion external contextual data also helps, to a large extent, better understand phenomena and improve decision-making in marketing.
This is why, when working on customers’ projects, we always enrich their studies, data science initiatives, and even data catalogue with contextual data such as attendance statistics, movement patterns and segmentation information or a list of all events and marches since 2015.
Who knows what this data cross-referencing could reveal?