After a day of mountain biking in the Kosiosko Royal National Park in Australia’s Thredbo Valley, I found myself in the depths of a serious conversation around Artificial Intelligence with a fellow cyclist, whilst hurtling down a mountain trying to avoid falling off the side of a cliff.
Laughing as I realised he knew far more about data than I ever will know, one of his first comments was, “All this talk about Artificial Intelligence is way off – Machine Learning will come first. Those talking about AI are literally doing just that, talking. AI will follow long behind.”
It is not often I find a kindred spirit, especially one who rides a bike like that, but there you go, that’s what they call serendipity.
This conversation got me thinking, and doing so has lead to this 4-part blog series, where we will explore what successful organisations who work with Tealium in Australia are doing to prepare for Machine Learning and Artificial Intelligence beyond.
We are currently having a lot of conversations with customers and prospects who want to know about Artificial Intelligence and whether Tealium is going to solve the challenges of scale and automation through AI. The answer to this is yes … and no. Do we believe, Machine Learning and Artificial Intelligence will eventually enable much of the long sought-after nirvana where our electronic relations with people interested in our brands will be as fluid, relevant, and consistent as those we have created in the face to face world? Yes. Is Tealium embracing that as part of what we offer? Most definitely yes.
However, we would be negligent if we simply told you to go out and buy another tool (what is another one in an average stack of 90) that you can solve the problems of scale and automation with. Unfortunately, it doesn’t work like that.
If you are puzzled over how to bring together the many growing sources of data you have, to unify that data, understand it and act on it, all in the same time it takes your customer to click the purchase button, then read on, Macduff.
At Tealium we have spent ten years understanding what makes data fast and useful to all, and whilst we are backing the capabilities of ML and AI to ultimately solve these issues, we don’t foresee it will happen by some magical shake of a wand. Anyone seriously embarking along the path to the world of ML and subsequent AI is preparing now by consciously building a neutral, cross-channel dataset. One that is ‘Machine Learning Ready’ so that when they are ready to start utilising the power of Machine Learning, their data will be ready, too.
When developing something truly unique that maps unchartered territory, it helps to have some guiding principles upon which to make decisions.
We find our most successful ML pioneers are building a neutral dataset which is anchored in the following principles:
- The customer at the center
- Real time
- Data readiness/access ownership
- Privacy, governance, compliance and security by design
- AI/ML Ready
These principles form the basis upon which brands are consciously designing a unified, customer at the center, first-party dataset. Not just giving lip service to the concept of data as an asset, but actually treating it as such. The principles and their outcomes can be summarised as follows:
Customer at the Center: This concept has many facets, but when it comes to communications, the underlying dataset must be one primed for delivering what the customer needs next, not what an organisation wants them to receive. All decisions on how, when and what to communicate should be based on the state of the customer (or potential customer) with respect to the organisation at hand. That can only be achieved if there is a centralised understanding that can be activated as it is generated. To be effective this must be:
- Responsive: Data must be responsive and fluid to provide that edge we call ‘experience’
- Data-Centric: As we have moved into a world of distributed trust, data is needed to reliably bridge the digital relationship gap
- Invest in Content: Data is part of it, content is also a key and needs investment
- Continuous Testing, Learning and Adjusting: Continuous learning and adjustment will become the new business as usual
Real time: The moment matters in the electronic world just as it does in face to face relationships. Ask any shop assistant how they make the most sales – it is by watching and listening to their customer’s needs and responding appropriately. There is no campaign, channel or device. Data is the language of relationships, so to capture the moments that matter most, data must be activated in real-time by design.
Access and Ownership: As the volume of data grows, organisations must enable data democracy. For many it is also about ownership of first-party data in a consolidated manner:
- Organisational Data Freedom: To scale its use, businesses must ensure data is accessible, understandable, and usable by all.
- Omnichannel: The key to this starts with a data layer. This concept is born out of digital which has become omnichannel. It consists of standardising and unifying all data, from all places at the point of collection, in the language of the business, thus creating a fluid dataset in a language that everyone understands.
- First-Party Data Focused: First-party data is a brand’s greatest competitive asset as no other organization has that dataset. Control, ownership, and activation will be key to fully realising its value. It is not only the history of previous transactions and loyalty but also the behavioural data that provides the insight to intent with first-party data.
- Internal Alignment: This is not just a technical requirement but will satisfy business strategy as well. It is important that all internal groups participate in both the definition and implementation of internal alignment.
Governance By Design: Whether you call it governance, privacy or simply respectful, sustainable relationships and the complexity of communications, combined with the growing understanding of the value of data, means organisations are now architecting this into their datasets and tools.
- Consent: Ensuring consent is stated before any data is captured is key as is the ability to review or apply this as required.
- Transparency of Use: To maintain a sustainable relationship where data continues to be freely given will require its use be transparent.
- Consumer Rights: With this transparency, the consumer will increasingly expect the ability to review, delete and be notified of the data a brand has about them and organisations will need to provide real-time capabilities to do this.
- Control and Audit: Finally, to facilitate all these capabilities requires fundamental data control and audit. It will be imperative to concisely articulate the data you collect and how and when it is distributed.
AI/ML Ready: As the volume of data grows, so will the need to automate it using the power of Machine Learning and beyond. Having built a dataset that is customer-centric, real time, accessible, and properly governed means organisations will be ready to utilise the data. The core processes to enabling this mandate are:
- Valid and Consistent Streams of Data: First up is the development of a framework which enables the efficient and consistent onboarding of new data streams which are ready for consumption by Machine Learning and the subsequent activation. This includes engineering the needs of Machine Learning into your organisation’s omnichannel data layer.
- Listen and Respond to That Data: Once the framework is in place, working with what the data tells you is important. Interpretation will be as important as generation.
- Hire, Train, Enable, and Prioritise: This is new, not all the capabilities will exist in your current team – you will need to hire, train and enable employees to achieve these goals. As this happens, prioritise your projects and use these to form the basis of your training and recruitment.
- Brand: Marketing is an art and a science, ultimately data and Machine Learning are still tools to enable the scalable delivery of your organisations brand and the messaging and engagement is the art of this piece.
Tealium helps our customers build a strong data foundation first, one firmly anchored in the bedrock of an owned, accessible-to-all, neutral, governed dataset that will be fluid in nature. We believe this will enable capabilities such as ML to flourish, but without the principles as fundamentals, there will be a constant need to rebuild the underlying architecture. A process that is costly both in time and resource.
The first step towards ML readiness is building a strategy and roadmap for uniformly onboarding data from any source in a state that is immediately ready for use by all.
This is a concept we will explore in part 2 of the ML/AI Data Readiness blog post series.