The future of retail marketing and common mistakes people make trying to get there

12 Jul ‘17

Competition is growing in all industries, creating new challenges for retail marketing. How can we use new technology effectively to transform marketing strategy?

A digital transformation is taking place in traditional sectors of the economy. US retailers plan to close more than three and a half thousand shops in 2017 and e-commerce, led by Amazon, is dominating established businesses. It is not difficult to predict: next in line are the remaining retail markets, including medicine, HoReCa, tourism, traditional banking and insurance. In other words, any business that works with large numbers of private customers.

From the marketing perspective, the root cause of the transformation is a change in customer behaviour. A study of B2B marketing trends by Boston Consulting Group (BCG) notes the consumerisation of business purchases. The new generation of procurement decision-makers prefers to spend less time talking with salespeople, and more time choosing for themselves based on information gathered online using all kinds of devices.

BCG has identified the following trends that are applicable and observable in retail:

  • from uninformed, vendor-driven buyers to informed, marketing-driven buyers
  • from intrusive mass marketing to personalised choices
  • from offline, individual channels to integrated communication built around on-the-go consumption, i.e., from smartphones
  • from manual work and disparate systems to automation using a single data set
  • from one-time measurements of scale and «last-click» attributes to constantly measuring the results and impact of all channels
  • from divided, hierarchical organisations to integrated, flexible structures
  • from creative to tech and analysis

Marketing is moving from brand positioning and coverage to focused, automated interaction: what the company wants to say to which segment, and when. This is also happening in FMCG, where the emphasis on coverage is inevitable because purchases are largely spontaneous. For example, Unilever identifies its most loyal customers by conducting interactive promos and uses the data to optimise its digital ad spend by targeting similar audiences with «look alike» algorithms.

Our goal

An example of good marketing is, a global leader in its segment and the world’s third largest e-commerce platform. The company pays minimal attention to design but runs thousands of A/B tests at a time and pays particular attention to providing a single user experience on mobile and desktop, with a focus on personalising its messages even before the customer has placed their first order. The CEO of emphasises that most of their marketing decisions are based on data analysis. The company makes sure that the number orders lost due to errors in its experiments does not fall below a specified level because it views this as a dangerous sign of slowing innovation.

From a technical standpoint, this is a matter of good old CRM implementation, in the original sense of the term: an ideology of building a business around a single customer profile.

Customer Relationship Management is poorly defined: it can be a system for collecting complaints; a way to manage the business processes for ordering; or sometimes a combined system for campaign management, data storage, outbound marketing and all analytical systems. The reason for this is that holistic CRM implementation approaches have evolved into frightening technical schematics from architectural committees that have taken years to draw up and have been available only to companies awash with cash.

This is not unjustified! Even average retail businesses are dealing with gigantic amounts of data: hundreds of thousands of customers, millions of transactions, and tens of millions facts about online behaviour. Bear in mind that for effective marketing, a lot of behavioural data need to be processed in nearly real-time. For example, the effectiveness of the standard e-commerce mechanics – «abandoned cart» and «abandoned view» – falls by an order of magnitude just 30 minutes after a buyer’s visit ends. This means that tens and hundreds of thousands of user interactions need to be processed in minutes. Not to mention peak loads during sales.

This is why large businesses still rely on complex, costly solutions from industry titans like Oracle, IBM, SAP, and Microsoft, despite the fact that many of them are rooted in the ’80s and are clearly behind the times. Everyone else, who can’t afford to spend millions of dollars on these implementations, believes that systematic approaches are inaccessible to them and remain content with half-measures.

B2B marketing technology leapt into the cloud more than five years ago. Recognised market leaders like HubSpot, Marketo, SalesForce, and Pipedrive have raked in billions from their IPOs. The B2C revolution is just beginning: it was made possible a few years ago thanks to growing computing power and, most of all, exponential growth in storage system speeds due to the transition to the solid-state disk (SSD).

Things that only big business used to be able to afford are becoming available to medium-sized and even small businesses. With modern equipment, standard solutions can cope with retail data processing, no longer requiring such complex architecture, millions of dollars, massive technical requirements and years of implementation. Gartner’s «multi-channel campaign management» category, which is dominated by titans of industry, has begun to include several young cloud companies in the past year: Sailthru, AgilOne, Listrack, etc.

Common mistakes on the road to agile marketing

Big data

Amid all the hype, it’s easy to give in to the temptation of simple solutions. But no, a simple green «boost revenue» button has yet to be invented. Technology is another tool that opens up new opportunities in workflow management. Implementing even the most modern software will not solve any problems in itself; in fact, it usually does the opposite – creates new problems.

This is particularly evident in the case of big data and machine learning, the most-hyped solutions promising the victory of machines over humans yesterday. In practice, machine learning tends to be reduced to automating the search for the optimal solution from a large number of options, subject to several conditions:

  • measurable criteria for finding the correct solution
  • having a large set of historical decisions (learning curve)
  • well-defined human variables that affect the result

This is very powerful, cutting-edge technology, but the entry barrier is quite high: you need to accumulate a huge set of clean data. Ninety percent of successful machine learning is collecting, cleaning and classifying data. As analysts say: garbage in, garbage out.

This is why automatic product recommendations work so well at Amazon and not so well at a new shop that has bought a service promising triple-digit sales growth after installing a widget on your page. Additionally, in order to automate a choice, you need to have something to choose from. Before you buy a service that can select a personal offer for each customer, it helps to make sure that you have at least a couple dozen such offers.

The chaos of excessive decentralisation

The market for cloud-based marketing technology is booming: a marketing-centric US blog has counted 5,381 (+40% in 2016) online services offering various marketing solutions. Moving away from the classic large implementation approach creates the risk of going to the other extreme: choosing a bunch of solutions with confusing integrations and winding up without a data set that you can trust.

For a long-term, client-focused strategy, it is critically important to have one stable infrastructure element when gathering data for machine learning and quality multi-channel communication: a centralised database with unique customer IDs that are used across all services and channels. Depending on the size of your business and your technical skills, there are a number of ways to build this kind of database: from Google Sheets and 1C to cloud-based marketing CRM. The important thing is to set up this database from the start and to make sure it can integrate the services you need to test with reasonable efforts.

Lack of metrics

Having a flexible approach with multiple experiments does not mean lacking control. Without clear metrics, faster speeds lead to faster Brownian motion. It’s amazing how many marketing departments function without clear, high-level KPIs.

Good metrics should be marketing-based and comparable. Most importantly, each metric should have an engaged owner who understands how it is calculated, regularly looks at trends, demands new hypotheses and asks for the results from past tests. The purpose of reporting is not to monitor, but to generate next steps. To change, not measure.

There aren’t many such metrics, and they differ from business to business and depend on development stage. As a starting point, here are a few best practices: conversion to first purchase, conversion to second purchase, and rebuy rate. You should also pay attention to the number of A/B tests and automated campaigns you have running. These are good indicators of marketing agility.

Statistical ignorance

Marketing and business are largely based on gut feeling and taking risks. Everything can be measured reliably, but the amount of time it takes to do so often makes it meaningless.

Still, statistical application and evidence-based marketing are growing rapidly, especially in digital advertising channels and e-commerce. You need to have a basic understanding of statistics to be able to simply ignore your intuition when necessary and listen to specialists showing confidence intervals, control groups and A/B test results.

The owner of a traditional business told us how a competent billboard test led him to abandon his vision of using a design with elements of corporate branding and instead use a simpler version from the marketing department that incremented sales by 30%. This business is far ahead of its competitors, and I understand why. You have to learn to trust numbers and the scientific approach, even though it is especially difficult for successful managers and entrepreneurs to do so. It is equally useful to be able to look past the pretty numbers that fill the presentations of unsuccessful marketers.

Even without reliable measurement, it helps to simply formulate your next steps like a hypothesis: taking <action> will increase <metric> by <forecast change>.

Using traditional mass-marketing methods

An important difference with digital channels is that customers are usually free to unsubscribe from ads. They can unsubscribe, send messages to spam, block banners or facebook ads, or file an official complaint about SMS spam.

Research conducted by companies in the West confirm our own [9] studies, which indicate that after 6-7 consecutive uninteresting communications, people begin to stop reacting to any communication from a brand.

It’s a classic sad story: a business spends years collecting the contact info of hundreds of thousands of customers, but never uses them. When they start building a direct marketing presence, they send frequent messages to the whole list. They experience a sharp jump in sales, brief euphoria, and then an equally sharp, disappointing drop: they burn out their customer base, with the exception of the most persistent buyers, which our experience shows are no more than 25% of the database. It’s difficult and takes a very long time to rebuilt customers’ trust. Don’t repeat this mistake. You need to start carefully, making sure you have a clear segmentation policy and control over message frequency, and that you’ve set up a feedback loop to avoid customer dissatisfaction.

People come first

Behind all the dangers, the fundamental task remains unchanged: how to use innovation to speed up your work. Technology alone is insufficient: you need to change people’s behaviour, i.e., their culture. Larman’s laws state that the most effective way to change culture is to change organisational structure.

Modern marketing functions are focused on email marketing, brand marketing, media buying, or even CRM implementation, and look more like an IT team. These are not the kind of structures that can foster continuous improvement.

If you want to speed up your work, you need to strive to transform your marketing function into a product development team with the authority to make change at all points of contact with the customer. In an ideal world, with a team of 4-12 people sitting in the same room, you should have enough knowledge and authority to test hypotheses: from generating ideas, to launching campaigns and analysing results.

Modern marketing directors think in terms of the product, and how customers will view their interactions with the company. They look at communication via ads, the website, mobile apps, and the call centre: they view transactional messages and advertisements holistically, flowing from one to the other.

Possible transformational starting points

A sign that your business has been successfully digitised is the absence of an IT function altogether, except perhaps for office equipment support specialists. Technology is no longer implemented separately for business use: tech is business. It changes so rapidly that implementation should take place in parallel with work, meaning that IT specialists should be part of the functional teams.

Out of the same need for agility, few people use internal IT resources to build websites these days. Likewise, testing marketing hypotheses should not require going outside the marketing director’s area of responsibility. From a business perspective, mobile apps, loyalty programmes, automatic communications, analysis and other automation shouldn’t be any different from website development. The technology has already been standardised to the point that the marketing team should be able to take ownership of their own tech.

Gartner forecasts that CMOs will spend more time on IT than CIOs in 2017. In the US, some people even believe that these two positions should be combined. Large marketing teams are often structured by channel: email, SMS, loyalty and media. It would be a mistake to think that such a structure leads to an optimisation of the parts that harms their sum: cannibalised traffic, conflicts and uncoordinated communication with customers. A possible alternative is to set up groups based on high-level goals like customer acquisition and retention.

Naturally, some channels might require unique skills and have little in common with others. For example, teams launching brick and mortar stores; but even in this case, these teams should adopt common marketing solutions for their channel and be responsible for common metrics as part of the larger group, rather than optimising their own KPIs.

Product development teams need marketers with tech skills who understand integration and data flow. They also need at least one analyst to ensure their data are clean and reliable. Analysts are particularly hard to hire, but it’s doable if you avoid machine learning specialists who are more interested in working for search engine giants like Google or Yandex. Again, it helps to start with the basics: SQL, basic knowledge of econometrics and statistics, an interest in the subject matter, and attention to detail.

It’s great when you are able to add content production to your team: a designer and an editor. In some cases, it’s also possible to integrate a developer and customer support specialist. The more you can do without leaving the room, the better.


Increasing competition means you need to take retail marketing to the next level:

  • ensure a single user experience across all channels
  • work with people, not traffic
  • iterate fast and release often, rather than making big changes that require lengthy preparation
  • base decisions on data, not intuition

We can see a common thread in marketing: market needs are forcing companies to flatten their organisational structures, moving from functional departments to product and project teams.

There are no out-of-the-box solutions, but you can begin your own transformation by setting high-level KPIs and creating the conditions to convert your hierarchy into a more flexible organisational structure. It’s great if you have an understanding of strategy, but even if you don’t, it will crystallise in the process. The workflow itself is relatively simple:

  • choose the most promising hypotheses
  • make sure you can trust your data and metrics
  • test hypotheses and watch your metrics

Repeat over and over until you iteratively arrive at a workable strategy based on your own experience and data, gradually transforming your marketing and company in the process.

Alexander Gornik, Mindbox CEO
Source: Forbes

Tell us a little about yourself

We’ll respond within 24 hours

Partnership request

Typically we’re answering within 24 hours