Spotting signals and taking action

Back Published: Friday, 20th July 2018 by Vince Jeffs, Sr. Director of AI Product Strategy, Pegasystems & Matthew Tod, Chief Data Officer, Celebrus

All the talk about personalised marketing, one-to-one marketing, targeted marketing etc. relies on an organisation’s ability to know consumers, build a meaningful relationship with them, and sense and respond in the right channels at the right time. To put that more practically, to effectively serve them, an organisation has to spot the signal that individual customers are sending out and then take an appropriate personalised action, preferably in real-time. This activity is at the core of delivering the better customer experience that many consumers crave.

Can your organisation really spot behavioural signals? Can it actually take an action in real-time? Honestly, most aren’t very good at this.

A Signal is defined in the Merriam Webster dictionary as “something that incites to action” which is an interesting definition; signals either generate, or at least highlight the need for action! Historically signals were generated from aggregated data and were applied to broad groups and segments, but if modern marketers are to improve customer experience, signals must be detected in real-time and actioned at the individual level.

There are broadly three ways to create an individual customer signal:

  • Singular actions: A single page view, phone call, or tweet could be used as a signal. It’s straightforward, provided the data is collected and all the actions have been reliably allocated to an individual. The content of the action can be analysed to discover intent or the context used to create a basic signal.
  • Compound event: An aggregation or pattern of actions that when processed generate a signal, for example repeated calls to a contact centre, visits to a help section on the website, or repeated journeys through the buying or registration process could indicate a problem. This is a higher quality signal as it is based on multiple data sources, channels, and perhaps multiple devices and it usually covers a period of time, though more work is needed to extract the signal.
  • Real-time scoring: Application of a model to the data, generating a score (or keeping track of state) that is calculated after every interaction - this is arguably the most interesting area, and it’s proven to deliver superior results. Advances in AI, data capture and streaming analytics make this one of the most popular methods of choice today.

Can your organisation spot a signal? Matthew’s bank can’t spot signals!

This is a real life example of a customer experience he recently had:

  1. He went to his bank website by typing in the URL
    • This is unusual as normally he uses the bank’s app
  2. He searched for “Child account”
    • Never done that before
  3. He got lost in the search results, and searched for “Children’s bank account”
    • Clearly, he was determined to find something!
  4. He found the right product page
    • Never been to the page before
  5. He scrolled up and down, but did not click through to any detailed info
    • Engaged but not convinced
  6. He saw no offers, promos or reasons to act right now
  7. He left the site after about 3 minutes
  8. He has not been back to the information since, though he has used the app

So, the signals he clearly gave off were:

  • He is now interested in finding a child account
  • He is definitely in the market to buy
  • He wasn’t sold any type of offer
  • He is cooling rapidly

OK, Matthew was not explicit in giving off those signals; the raw data needed some processing.

Signal processing is a collection of “techniques that are used to improve signal transmission fidelity, storage efficiency, and subjective quality, and to emphasize or detect components of interest or intent in a measured signal.” It’s not a new area, but it has to be done to convert raw data into something of value that can be acted upon. In this case though it would have been a very simple processing task to work out with a high degree of certainty what he was probably trying to achieve!

So, Matthew’s bank failed; it had a clear shot to acquire a new customer (Matthew’s child) and they missed it.

What should his bank have done?

Well, the answer is simple. They should have fed that signal into a decision-making system. And that decision-making system should have been equipped with an always-on customer brain that could quickly recall the necessary historical picture of Matthew and key relationship insights such as:

  • Matthew has been a loyal customer for 20 years, with savings, checking, CDs, and various other accounts
  • Matthew has children
  • Matthew has the bank app and uses it on a regular basis

That same brain should have ingested the key contextual data fed from the real-time signal, and combined it with the historical profile:

  • Matthew entered the website directly via plugging in our URL
  • Matthew used site search, entering keywords giving amazing insight into his intent during that session
  • Matthew got to a specific product page, and had specific telling behaviors on it

Finally, the brain should have been able to provide helpful recommendations to Matthew such as:

  • Populating a web container on the product page he was viewing with great offers on getting help applying for and opening a child account
  • Fired actions to follow-up with Matthew, via email or agent calls, to help him as required with all the necessary support for completing his child account application process

Simply put, Matthew’s bank needed real-time signal processing and a decision engine running that could pick up on his current behavior and journey, arbitrate between different possible recommendations, make a decision, and drive immediate action (and orchestrate follow-on actions as appropriate)! This lack of an integrated signal detection, a brain, and linked execution capability created a poor customer experience and cost this bank a new customer. This scenario is being replayed in many banks, for many products, thousands of times per day around the world.

If improving customer experience is at the heart of your competitive strategy, then your organisation’s ability to listen, spot signals, and then act appropriately will be critical. So back to the original questions posed at the start: can your organisation really spot behavioural signals? Can it actually take an action in real-time?

×

Privacy settings

At D4t4 we are all the about the data. We are serious about data protection and your privacy so we will only collect your personal data and use it with your permission. We use cookies to collect statistics to optimise website functionality and deliver content tailored to your interests.

Our three categories of collection are detailed below.

Essential browsing only We will only collect the essential data required to enable core site functionality. We will not collect any personally identifiable information or behavioural data.
Browse anonymously We will only collect your browsing behaviour on the website to help understand our customers' needs and improve the experience for everyone. We will not collect any personally identifiable information so we won't know who you are.
Personalised experience We will only collect information that allows us to identify you and make your browsing experience as smooth as possible by remembering your log in details and saved items. In the course of dealing with you, we may need to pass your personal data on to third-party service providers contracted to D4t4 Solutions.

You will be able to change your options at any time by clicking the Privacy settings link and our full Privacy Statement can be viewed by clicking the relevant link.

v20180607