The New Magic in Retail

Where is the Magic in retail? It used to be that having a presence online was Magic in and of itself. Now, with the advent of AI and Machine Learning making an impact and the widespread adoption of flexible architectures, we are on the precipice of the New Magic in Retail.

Dom Selvon

Thu Jan 31 2019

It’s an important time for retail in an ever more connected world.  Brand consistency across all the touch points and a suite of APIs and common business domain services ensure that the customer is comfortable with the retailer regardless of which channel they are transacting on. Brand alignment and a single view of data in all directions – from the customer and from the retailer, provide the customer with a brand aligned view into the retailers world leading to increased loyalty.

Achieving omni-channel excellence is a challenge currently facing every retailer in the space. Historically, presence on the internet with a webshop of any kind was seen as a kind of magic. The ability to sell online and fulfil, deal with returns with less manual interaction and be able to keep a storefront fresh for just one channel was difficult. That is no longer the case. It's been commoditised and seen as par for the course. It's even viewed as simple. We're now in the domain of Artificial Intelligence, Machine Learning and continuous delivery of targeted content to each and every one of the millions of individual viewing a retailers online offerings on all of the internet connected devices you can think of. This is where the new magic lies. The ability to present a seamless and frictionless experience regardless of the individual, what medium they are using and indeed the channel through which the brand is being presented.

The challenges that this new magic present to the retailer mean that a dedicated and focused, static monolith concentrated on a particular domain (i.e. commerce, or PoS) is now unable to cope with the horizons being presented to it. There needs to be more flexibility in the architecture and more flexibility in the software realising that architecture. This gives the people running the architecture more flexibility in realising a brands vision.

Flexibility in Architecture

Conways law states that every IT system built by a business is constrained by the communication systems of that business. I’m paraphrasing slightly, but this leads to inevitable business silos caused by new technologies becoming available. There’s no point trying to break Conways law. Instead retailers must embrace it. Having flexibility in their architecture means that the software works for the business rather than the business needing to change to align with the software. With a service oriented architecture, spinning up a team to focus on a proof of concept or to trial an new technology is easy. Want to try leveraging a data set fronted by bootstrapped ML algorithms to present automated recommendations to customers? Spin up a new service and present the results in a single channel. How about providing your business to delegate ownership of specific areas of retail operation to empowered groups? With a monolith, that's nigh on impossible and some form of centralised operational team needs to police that. With an automated service paradigm, you can spread you metaphorical wings.  

To align the software to the business does require thoughtful enterprise road mapping and a thorough review of enterprise components. Retailers should Take T-I-M-E to understand legacy applications before decoupling them to put them back together again. 

An involved and business lead enterprise strategy that embraces change will ensure customers can transact at any touch point with the retailer and allow the business to grow and experiment with what makes them unique.

In order to realise a truly omni-channel architecture, a retailer of sufficient size is now compelled to look at decoupling their services and re-coupling them as either micro-services, or in a service oriented architecture. Unless a retailer has a high propensity for risk, a transition is necessary to achieve this. This decoupling and atomising a retailers functions are imperative to  allow their business scope to grow and scale. 

Migration of key services can be managed in different ways to realise this transition. This is determined by heritage and system dependency, data model, business roll out timing, requirements and risk appetite.

Further to this, identifying an abstraction layer to put in place, even if the source systems remain the same, is imperative to embarking on the omni-channel journey. This layer enables the concept of strangulation to become a reality. Without it, you are constrained by the source and destination systems and you produce a very tightly coupled enterprise suite of software.

Think security first. Reducing the surface area for attack and reducing the number of potential vectors for attack are the primary areas to concentrate on when it comes to securing a distributed architecture this flexibility requires.

Tokenisation provides a business and an individual with a Common view of payment credentials. If we have a common view of the customer so data is not duplicated, and these customer specific tokens are stored in a secure location, then the surface area for attack is greatly reduced.

When it comes to vectors for attack, introducing a split or service based architecture means that there is a natural filter between business functionality and the ux. This provides an excellent filter point and allows for greater validation and security to be performed.

Flexibility in Software

The current trend in DevOps as a culture means that there is a wide variety of software that is marketed to solve "The DevOps Challenge". Consul, Helm, Kubernetes, Docker, VMWare, AWS tools, GCP tools, Chef, Puppet, Kubespray and many others. All of these could potentially have their place in a modern retail IT operation, and because of the sheer volume of options, information overload and blocks due to decision paralysis is a very real problem. Modern software is built in such a way that spinning up proofs of concepts and trying them out coupled with developer tier pricing models means that investment is minimal and knowledge is very easy and quick to come by for a motivated development and infrastructure team.

The software that will provide the domain specifics (such as web shop functionality, or store PoS functionality) to the new retail enterprise is also manifold. You have the likes of commercetools and Moltin offering modern cloud based engines of commerce embracing a micro-services paradigm, but hidden behind a rich API to implement and leverage as the retailer wishes. GraphQL is rapidly gaining traction as a mechanism to draw data from multiple different domains without requiring the domains be co-located. GraphQL comes with its own challenges, but that's another discussion entirely. We are seeing a raft of alternative payment providers offering innovative solutions and software that allows for unique experiences in store that haven't be readily possible with the PoS software of yesteryear.

Big data is also becoming increasingly more consumable with the likes of Oracle and the Oracle Data Cloud, and Sqreem offering API based access to large lakes of data that is consumable through mature machine learning algorithms to provide relevant and targeted data to your consumers. Adoption of GraphQL, and microservices will allow you to seamlessly integrate this big data with the engine of commerce without being constrained by the platforms.

Content management is undergoing a paradigm shift also with the advent of the Digital Experience Platform and headless content platforms being provided by the likes of Adobe, Bloomreach, Contentful, Content Stack and Amplience. The customer, as an entity can reside in a variety of platforms now from heavyweight enterprise grade Customer Engagement platforms from the likes of Oracle and SAP to apps like Magentrix that plug into heavier CRM software such as SalesForce. 

Point of Service software is rapidly becoming much more lightweight and even the dedicated software such as Oracle's Xstore is pliable enough to be relevant in a flexible omni-channel retail world.

Simply put, the software must not dictate the business processes, but must be sufficiently flexible in itself to be able to plug into an agile business.

Flexibility in People

Whilst I've said that the business shouldn't change to meet the requirements of a monolithic piece of software, the business should be flexible enough to adapt to change. Furthermore, legacy ways of working, particularly in IT, will have to change to cope with the capabilities of a truly omni-channel retail operation that embraces all the new technologies that are coming to light right now.

The ability to proactively monitor a live swarm of services all operating together to provide the enterprise experience for customers is something that must be baked into the architecture, the software and the people. Ensuring that the business operations do not get affected by failure of any service, all services need to be designed with failure in mind and the ability to gracefully recover are concepts that need to be embedded in the DNA of personnel working on building and managing these services. Security becomes a first class citizen in the enterprise with a policy of automation and the path of least privilege ensuring that the software can largely look after itself and the people involved can do no damage.

Managing a single server is relatively easy. When adopting a service oriented or micro-services future, people will find themselves managing 10s if not 100s of services all needing containers and service registries and the ability to recover from failure. Achieving this manually is an exercise in futility and so automation is the only viable option.

Delivering the New "Magic"

eCommerce was Magic and it is still difficult to do well. Omni-channel commerce across the customer, order, inventory and payments coupled with a truly personalised experience for the customer that leverages machine learning to reason about the reams and reams of data collected is where the new Magic sits. This is achievable, but only through serious thought and intricate care taken into transitioning from the old way of monoliths controlling all aspects of an enterprise to dedicated services performing small discrete features extremely fast, with true resilience at internet scale.

Like what you've seen?

Get in touch