Why retailers need to embrace cloud computing for growing sales

Data science can spot trends that the human eye (even with very impressive statistical support) cannot match. Finding nuggets of insight and value is easy when you do your first optimization run on a line of products – often anomalies can become apparent from calculating correlations in the customer journey data using statistical models. But broadly, you – the analyst – define the scope of the analysis. Machine learning will take you further – making possible the mining of data patterns amongst millions of interactions in your customers’ cross-stack journeys beyond what the human can consider and at a speed that cannot be matched using traditional statistical methods.

If machine learning can add brains to your team, then the cloud can add muscle… Finding the next significant optimization may require you to go through a test and learn process on very diverse sets of data and huge volumes. It should be said that making sense of consumer behaviour data and building the right strategy to select and test these data clusters still requires both a good business knowledge and data science skills to manage test impacts on the revenue. But cloud computing provides you with the flexibility and power to summon different technologies and processing capabilities in hours not weeks.

Automation

Which leads us on the next benefit of the cloud – the ability to turn analytical insights into action…

The mainstream digital marketing platforms – such as Google marketing platform and Facebook – are built on a cloud-based infrastructures, making it increasingly easy for marketing activities to leverage cloud. Many SaaS (software as a service) products run on one of the three major clouds – AWS, Azure and GCP. This makes the promise of mass personalization a practical reality for the advanced retailer. By “mass personalization” I mean homepages configured to each individual customer’s profile; ads and emails that are based on previous decisions that a specific customer has recently made; and offers that anticipate that customer’s most likely next purchase with an offer or nudge to close the deal.

This is the process by which we use cloud computing to turn the Machine Learning into triggered responses and tests, to use the digital tools to actively sell to customers rather than let the customer do all the work. For example, Google Marketing Platform and GoogleAds and Facebook have TargetCPA and TargetRoAS algorithms which process the data in near real time. Google Analytics has “session quality” which lets you create audiences with high likelihood to purchase or complete an action. And there is plenty of evidence of this approach delivering results for retailers: a recent Salesforce study on 500 million consumers, for example, showed a 14% increase in average order value when product recommendations are based on AI.

And it’s worth noting that this carefully curated and tailored marketing process is what our customers are demanding from retailers. Ultimately, you can argue that this is how we can deliver on the promises we make when we ask for permission to market to consumers under GDPR – we need to offer something of value to the consumer in return for being given access to their personal data.

Centralisation of the audience insights

I’m avoiding using the word omnichannel here, because I think we are all tired of that word – it promised something to do with intelligent targeting but tends to mean bombarding consumers with messages in every channel in a wasteful way. With budgets tight and requiring careful optimization, what retailers want to do is create a central repository of audience insights in one channel that can be analysed and deployed in other channels, so that the next ad or offer is intelligently targeted.

As digital marketers, we put the greatest emphasis on connecting data with the activation tools for our marketing. Cloud computing today is not only tools for system integrators and software engineers – it’s transforming marketing as we know it, so that the findings from machine learning can be automated and deployed directly in the digital marketing tools in ways clunky and expensive when using traditional data warehouses.

An example of this might be the enhancement of your CRM programme. A retailer may be happy with a 20% open rate for an email campaign, but that still leaves 80% of your loyal opted-in customers who have not seen your offer. What do you do with this 80%? CRM teams hold this data, and will be tempted to send another email, but that may not be the smartest approach. A central repository of audience insights in a cloud platform will allow you to target the missing 80% with a range of different tactics. Maybe show them 5-8 banner ads over the next 5 days. Or change your bidding strategy on their search terms. Or reconfigure their next landing page experience …

You may argue that this is already feasible with existing CRM systems you have on premise. But tying components of your martech stack together to experiment with new activations to your customers is one of the key goals of any centralisation project. And this is considerably cheaper in the cloud because the initial setup costs are close to zero.

The important thing is to be continually learning the tactics that most efficiently close the sale in the widest range of circumstances, and this requires insights and an on-going testing framework.

The options for identifying audience traits and deploying the fullest range of digital marketing tactics is endless when connected to the cloud.