In this blog post, we are going to explore Next Best Actions / Next Best Offers use cases, and the reasons why these use cases are a hot topic right now.
Introduction
Next Best Action (also known as best Next Action, Next Best Activity or Recommended Action), is a customer-centric approach to marketing that considers multiple different actions that could be taken for a specific customer and then decides on the best one to provide. Moreover, Next Best Action commonly requires the need for automated self-learning decisions, using machine learning, to make a decision in real-time that will improve the chances of a conversion at the very moment that a customer is making a purchase decision.
In this article, we will see successively:
- Context
- Limitations of mass marketing
- Presentation of the Next Best Action (NBA) / Next Best Offer (NBO)
- Description of marketing methods
- Next Best Action use cases in pharma
- Presentation of our approach
- Added value of Prevision.io
- Examples of similar issues in different economic sectors:
- Telecom
- Energy
- Bank
Context
For a marketing team, each customer journey is unique.
Its challenge is therefore to find for each customer the ideal action that will take him to the next step, to achieve a final goal (purchasing process for example).
The “Next Best Action” (NBA) or “Next Best Offer” (NBO) is an ideal marketing case for Machine Learning and Artificial Intelligence.
Machine Learning is, by definition, used to find specific trends in customer data, based on their behavior, intentions, attitudes, beliefs: browsing, past purchases, demographics, household data, etc. And above all, Machine Learning makes it possible to extract the interactions between these multiple criteria, and in real time. For example, at some point in their journey, some customers will prefer an email to a phone call.
A well-designed machine learning model can reveal the NBA with a high level of truthfulness, and above all, explain the reasons why this very NBA might be chosen.
Client side challenges
Have you ever received an advertisement for products you are not interested in ?
Worse yet, have you ever received an advertisement for products that are not relevant for you ?
→ Companies have a wide range of product offerings and a large customer base.
What are the traditional methods of marketing products ?
Approach 0: mass marketing
Historically, organizations used to send the same messages to all customers, using the same media channels.
Currently, this is viewed as spam.
The diagram below shows the traditional approach in action: every customer sees the same sequence of touchpoints.
Nowadays, marketing favors personalizing / individualizing the relationship with its customers, rather than using the same channel, the same message and the same offer for all customers.
Approach 1: Segmentation
A second approach consists in segmenting customers according to certain data (for example socio-demographic data): these segments group together customers according to this data, in order to create different customer journeys within each segment. The same messages, with the same media, are used for all customers in each segment.
This is an improvement over the previous approach, but all customers in the same segment who share the same data (socio-demographic) may not have the same behavior when faced with a purchasing process, for example.
Points for improvement:
- Same communication per segment
- What message ?
- Which channel?
Favor an individual approach
Approach 2: Next Best Actions
The most effective solution is to identify at the customer level what is the next best action.
This method is currently recognized as the best practice in modern marketing.
This approach takes into account all historical data for each customer.
Available data:
- Demographic data: age, sex, geographic area, income, marital status, …
- Behavior data: purchase of products, options, services & subscriptions, means of payment, etc.
- Usage data: where, when and for how long
- Interaction data: proactive / reactive, store, website, email, sms, telemarketing
- External data: purchase of files, public data, …
Using this data, it is possible to find patterns between the characteristics of customers, the sequence of their customer journeys and the purchasing decision.
For example, customers under the age of 30, who received an email followed by a reminder text the next day, may be much more likely to buy online. On the other hand, 55-year-old customers living in the north will prefer contact with a customer advisor in the store.
The best way to find these patterns is through machine learning – where algorithms learn by looking through all the data found on a customer journey, common relationships in order to predict a sale, a type of product, etc.
Once the algorithm has learned, you can use it to select which marketing content to serve. When a point of contact enters a marketing process, the machine learning algorithm will predict a probability that this customer will end up with a specific purpose (purchase or not, registration for an event, download of a brochure, etc.). If the point of contact is different, the probability will change.
A machine learning algorithm specific to a “best next action” proposes the highest probability for each action in a marketing journey.
In the example below, for customer A, you select the third option for your next communication to this customer, because adding this touchpoint brings the greatest probability of sale (75%) and is therefore the one that would most likely bring that particular customer closer to a buying decision.
List of next actions and associated probabilities
With the Next Best Action methodology, each customer has a personalized journey.
These action models can be complex and time consuming to create, but with Prevision.io’s automated platform, the next best action models can be created with just a few clicks. Prevision.io helps you quickly generate accurate predictions to find the most suitable NBA for each customer.
Simply put, the next best action refers to deploying models that use predictive analytics and machine learning to recommend, in real time, what actions a customer might take, based on their profile, previous actions and needs. Among a universe of potential actions, the best action is the one that benefits the consumer the most (by providing him with what he needs and what he wants) and the brand (by giving him access to the best customers, in real time).
Next best offer is a subset of the next best action.
“Deliver the right messages, at the right time, through the right channel.”
Next best action use cases
Use case | Description |
Anticipate the needs of each client | Use a holistic view of each customer and all available information to understand their current situation. Anticipate what each customer needs next to stay satisfied and loyal. |
Increase conversions and purchases | When a marketing offer is the next best action, you improve conversion rates and purchase volumes. |
Create omnichannel experiences | Choose not only the right action and the right time for each customer, but also the right channel through which to act. Avoid channel conflicts and create a better customer experience. |
Develop affinity with the brand | By making sure every interaction is relevant and includes more than just offers (service actions, content, etc.), you’ll make customers feel like they’re being served by your brand. |
Improve the customer relationship | Personalize contacts with your customers using history, historical behaviors, and preferred communication channels. The customer experience improves as the processing time gradually decreases. Customer retention increases as discussions become more focused. |
Build a customer engagement index | When properly calibrated, a single measure of customer engagement combining engagement across all channels, can be strongly correlated with sales. This makes it possible to quantify the value of the “right engagement” and to follow it in real time. |
Next best action approach
To complete a next best action identification project, you need to:
- Get your data (historical and if possible a 360-degree customer view)
- Train machine learning algorithms easily and quickly
- Operationally test your marketing (message / channel / recommendation)
- Analyze the performance
- Continuously and always improve your models very easily
Example of a similar approach with a Telecom operator
- Development of many scores:
- Appetency by product,
- Appetite by service,
- Positioning at the risk of churn,
- Cross-selling arrangements,
- Probable characteristics (composition of the outbreak, presence in the outbreak, etc.),
→ For each customer’s behavior, we can calculate a score and propose an NBA …
- French Telecom Operator recently perfected its probability tool by moving from an individualized approach to a customer-ego-centered logic
- Until now, each score placed each customer on the set of available scores but without a relationship being established between them
- The novelty of this recommendation tool lies in our ability to identify for each customer what their priority for purchasing products and services might be
- Customer satisfaction is approached through the “fragility” score. Telecom is thus able to provide customer advisers according to a personalized summary for each customer
The KPIs below illustrates what the information made available could be:
- The 3 products / services for which the customer would be the most appetizing, ranked by priority
- The color of the smiley shows the degree of appetite for these elements
- The smileys express the customer’s sensitivity as well as an alert in relation to his overall disposition towards the Telecom operator (during work, based on the traces left by the customer over the last 15 days on the Internet)
Example of a similar approach with an Energy operator
Use case:
- Appetite for a B2B offer
- Predict the optimal contact channel (the one that will maximize the probability of conversion)
- Taking into account the cost of action
- Actions combining appetite for the offer, addressing channel (mail, mail, sms, phone call) and cost of addressing
- Measurement of the gain in value / life expectancy (churn measurement)
Example of a similar approach in Banks and Insurance companies
- Use case:
- Predict collections
- On the entries into recovery predict the best action to put in place for them to come out
- Contact the client and resolve the situation,
- Predict whether to freeze the loan for a few months,
- Make a clearance plan…
- Actions combining the prediction of entry into debt collection + the action to be implemented
- Measurement of the gain in value / life expectancy (churn measurement)
Conclusion
We have introduced the next best actions use cases and the different approaches to address them. We have also seen how certain sectors of activity wish to meet this need.
If you want to know more, because we think this subject deserves a discussion beforehand, we invite you to contact us to describe your use case.
L’article Introduction to Next Best Actions est apparu en premier sur Prevision.io.
