The data hygiene steps you can take today to accelerate machine learning tomorrow
By Jason St-Cyr, Technical Evangelist, Sitecore
Digital marketers today are caught in the middle of a huge digital transition.
Behind them lie the old, slow, and manual methods of creating content, connecting with customers, tracking engagement, and measuring the results.
Ahead of them is the promise of the brave new intelligence-enabled world, teeming with possibilities—task automation, massive-scale analytics, and real-time, data-driven decision making powered by Artificial Intelligence, Machine Learning, and Deep Learning solutions.
For the most part, these promises have been made in good faith. Connecting clean, structured, and tagged data with powerful machine learning algorithms is an exciting prospect brimming with interesting use cases like:
- Automated content tagging
- Self-assembling web pages
- Dynamic audience discovery
- Predictive scenarios and best next steps
However, for many, this bold new future of seemingly magical efficiencies might be further away than it seems. The industry is currently packed with pundits and vendors getting excited about what might be possible with AI without thinking about the steps needed to get there. While the benefits of becoming an AI-enabled marketer are clear, the journey a marketer and their team needs to take is still hazy.
So what is holding us back?
The industry right now is obsessed with going faster and beating the competition. It is poised to leap into an AI future that hasn’t fully arrived yet, and there are a mix of obstacles slowing things down:
- The underlying technologies are still maturing and can be prohibitively expensive
- New disciplines and processes are still being refined, and we don’t yet know what good looks like
- There’s a major shortage of individuals with the skills to do this work
This is partly a question of time. There’s not much a marketer can do to address these challenges other than wait. At the same time, if everybody is waiting, then nobody is on the leading edge pushing the technologies to make them mature faster. The industry needs a mix of innovators and second-generation adopters to help grow and refine the solutions over time.
In the meantime, if you are not interested in being on that leading edge, there is something a forward-thinking marketer can do to prepare. In fact, it is vital preparation that directly dictates the success of your future AI initiatives.
Getting your data ready for the AI future
For marketers to effectively leverage AI, they need to use tools that connect structured, tagged, and organized data with machine learning algorithms. After all, self-improving algorithms are only as valuable as the data and content they process. To find opportunities for automation and trigger intelligent responses, you need data that fulfills certain criteria.
For instance, you need to use quality data. The data needs to be clean, accurate, and free of duplicates. Crucially, in the new GDPR world, you need to have data which you have permission to use.
At a more operational level, you also need to make sure you have a clear and consistent vision of the organization’s data landscape. Practically, that means working to break down data silos by supporting cross-department standardization and integration. No more excel files of contacts on everybody’s laptops!
You should also be looking at whether you are tracking the right data to be making automated decisions. Demographic data can be helpful in some ways, but can also lead to issues such as unconscious bias, or unfair outcomes. Therefore, I usually recommend valuing the tracking of customer behavior and context over basic demographics like gender or age.
Clean, structured, and tagged data is a critical prerequisite for leveraging AI and machine learning. But there are three key additional requirements that are easy to forget:
- Choosing an AI-ready CMS
It’s not enough to simply point machine-learning algorithms at vast data sets and wait for the insights to pour in. The value of AI-enabled analytics is in the ability to look for patterns in data across multiple sources.
To support that, you need a flexible and extensible CMS that integrates, collects, and connects relevant customer experience data from all your external systems—things like your CRM and e-commerce systems.
- Investing in scalability
It takes a lot of processing power to run massive-scale analytics. Once upon a time, this would have priced out most organizations. Fortunately, cloud service providers present a flexible and highly cost-effective option for even small teams to access the vast computing resources needed.
Don’t immediately gravitate to the larger providers. If you’re not running business-critical processes, the high-availability, multi-region resources that enterprise-scale cloud providers offer might be overkill.
- The right architecture
To optimize the content you deliver to different audiences, you need an object-based approach to content storage that breaks down content into discrete elements that can be identified, analyzed, tagged, cross-referenced, and rearranged. When your CMS can store content in this more atomic manner, you have a lot more flexibility for using existing content across channels rather than having to rewrite entire pages for specific interfaces. This also means that when you start looking at automated content optimization, your system will be able to more readily analyze the impact of smaller segments of your content and provide you with insights on what is working best.
AI capabilities may be one criteria on which you might be evaluating content management systems. For others, check out “The definitive guide to choosing a content management system” today. Please, let me know what you think about it – you can reach me on Twitter at @AgileStCyr.
Jason St-Cyr is a technical evangelist at Sitecore. Find him on LinkedIn or follow him on Twitter @AgileStCyr
Get the ebook: The definitive guide to choosing a content management system / Read it online.
Blog series, post 1:The right and wrong reasons to invest in a new CMS
Blog series, post 2:Three key pillars of building the business case for a new CMS
Blog series, post 3:10 editorial features your content creators need in a CMS
Blog series, post 4:How to mitigate risk during a CMS implementation
Blog series, post 5: Preparing your CMS to handle your company’s first surprise spike in page views
Blog series, post 6: Preparing for AI-driven customer experiences