Ai Editorial: Gearing up for the era of automated personalisation

4th December, 2019

Ai Editorial: Personalisation models in e-commerce are not only looking at aspects like predicting a visitor's likelihood of conversion, but machine learning deployed to match different offer variations to each visitor also adapt to changes in visitor behaviour, writes Ai's Ritesh Gupta

 

There is greater clarity on the journey personalisation today.

Be it for being adept with data (it has to be captured, pooled, processed and analyzed) or personalisation being an ongoing journey, an organization has to prepare and remain diligent.

For instance, in case of delivering targeted content to digital visitors, there is a need to create a personalisation rule (meeting conditions such as mapped behaviour, a certain score etc.), personalize the content (content based on the conditions in the personalisation rule), the layout etc. For all of this to work, the quality of data counts. As it turns out in case of personalized content, it is served on the basis of all the behavioural data about the visitor, plus machine learning deployed to match different offer variations to each visitor also adapt to changes in visitor behaviour.

The much-talked about individualized/ hyper-personalized offer is all about an individual offer based on one's preferences, activities and behaviour. In the ideal conditions, each individualized offer is unique, as shared by Comarch in a recent blog post. The objective of personalisation is to interact with a visitor in ways that optimize for a desired result featuring that individual, such as higher spend, continued loyalty, etc. It is about influencing a customer’s behaviour.

As for the progress being made in e-commerce, specialists already talk about automated personalisation. It not only looks at areas like predicting a visitor's likelihood of conversion, but the model also is trained to focus on a section of traffic to sustain learning throughout the life of the activity, and evaluate evolving preferences of the visitors.

There is provision for using offline data, propensity scores or other custom data to build personalisation models.

No magic bullet

It is fascinating to assess how machine learning helps in assessing the most effective content when targeting diverse visitors. It happens over a period of time as the algorithm learns to predict.

Algorithmic optimization isn’t an overnight phenomena; it entails a layered development of ever-increasing complexity. Plus, machine learning isn’t a magic bullet, capitalizing on its prowess demands diligence and a methodical approach. It one of the many building blocks that lays foundation for an astute loyalty initiative. In the initial stages, don’t target BMW of machine learning, recommends airline loyalty and big data expert, Mark Ross-Smith. Rather than thinking of AI, focus on simple things. For example, first target simple areas like identifying and addressing a traveller. “How to address someone in the right way at the right time,” he points out. A connection between a brand and its customers will build over time, deepening with each gratifying interaction.

Few key takeaways:

  • Personalisation requires clear vision, resources and investment. From the outset, the roadmap needs to be clear. So how the graph of personalisation is moving along, the entire organisation needs to be aware of.
  • For any airline to achieve personalisation at scale – the success lies in being able to unite all of their data. Data-driven organisations ensure their customer data collection fits in with constantly evolving behaviour based on their context. So if an airline doesn’t end up connecting the dots and act on the context of a situation, then it would end up missing out on optimizing the experience.
  • Move towards “operationalising” actionable data. It is imperative to interact with passengers across channels from one core hub and having an ability to answer swiftly and aptly in context.
  • Airlines also need to plan for how to set up rules and decisions for all channels, so even if starting with one channel.
  • Delivery of content: Explore the pros and cons of coupled versus decoupled web content management architecture (for instance, in case of coupled architecture while the initial setting up is simple, scaling up is one issue. As for headless CMS can result in freedom while finalizing on a front-end user interface technology say for an app, on the flip side the CX or customer experience ends up being decoupled as well.  This would limit the ability to personalize the overall experience).  
  • As Comarch points out, as a general rule, higher personalisation levels require more complex technological solutions and carry a higher degree of risk, but also may result in a much more significant set of business benefits.
  • Don’t rush to eradicate the problem of silos. Be careful about pushing for a sudden change. People aren’t going to be responsive to “new jobs, new skills, and new people to work with” resulting from drastic steps. Don’t risk the continuity of teams.

 

Keen on exploring topics related to stepping up the conversion rate? Ai has planned AncillaryRevenue Conferences in Berlin, Bangkok and San Antonio in 2020:

https://lnkd.in/fGSVujK

 

Editorials

  • Ai Video: Curbing loyalty fraud with machine learning +

    12th December, 2019 Airlines are looking at ways to curb illegitimate access to a member's loyalty account and resulting imbalance in a member’s spending or reward activity owing to hacking. Considering Read More
  • Ai Video: Nurturing loyalty via positive emotion of being rewarded +

    6th December, 2019 Airlines are evaluating ways to sharpen their respective loyalty cycles, i. e. letting a member earn and burn faster. What makes the game interesting today is how Read More
  • Ai Editorial: Gearing up for the era of automated personalisation +

    4th December, 2019 Ai Editorial: Personalisation models in e-commerce are not only looking at aspects like predicting a visitor's likelihood of conversion, but machine learning deployed to match different offer Read More
  • 1
  • 2
  • 3
  • 4
  • 5