Augmented Decision Making: Taking AI to the next level
April 2022
In partnership with
Contents
1. Mobile Growth Opportunities 2. Augmented Decision Making for Customer Retention 3. Benefits of Loyal Customers 4. Beyond Telecommunications
5. Evolved Decision Making
Through multiple applications across the value chain, Artificial Intelligence is fast becoming an invaluable tool for telecoms to streamline operations, reduce costs and improve revenues. However, it must be approached with caution, maintaining the right balance of human touch and automation. Whilst we have massive amounts of data at our disposal, the challenge is how to account for it all when making decisions. Decisions are often still based on gut feeling: more of an art more than a science. Augmented Decision Making (ADM) allows for the best of AI while keeping humans in the driving seat. In this paper, we explore the applications of ADM for customer acquisition, retention and monetisation. Focused on the telecoms industry, we highlight why such a solution is needed, how it works as well as best practices and pitfalls to avoid.
1
Mobile Growth Opportunities
The opportunity for mobile growth… AI is at the heart of a personalised approach Mobile operators recognise the need to combat value erosion from falling airtime, high churn rates and competition from over-the-top (OTT) players to stimulate growth in customer lifetime value. Successful initiatives will require a deeper understanding of customers than ever before. Operators have a huge amount of data at their disposal, and have an opportunity to harness it to increase customer value whilst managing the need for rapid response times. A new wave of AI is emerging as a solution: Augmented Decision Making (ADM), which unifies human and artificial intelligence. Figure 1: Initiatives to combat falling customer lifetime value
Falling airtime revenue
Cost reduction
High Churn
VAS
CUSTOMER LIFETIME VALUE
OTT Threats
CUSTOMER LIFETIME VALUE
Loyalty Programs
Segmentation
Segmentation Segmenting the consumer base is no longer enough to create a competitive advantage. Instead, breaking down consumers into micro-cohorts will deliver value through deeper personalisation. Where in the past operators may have catered to the ‘Youth’ segment, a more specialised approach would sub-group ‘Youth’ into students, millennials, gen-z, etc. In the near future, wholesale will facilitate more hyperconnected services that cater to new micro-cohorts with evermore specific requirements, shown below. Figure 2: The growth of niche wholesale services 5%, Youth/Media
4%, Bundled
9% Int. Roam 12%, Business
23%, Discount
2021 18%, Other
12%, Ethnic 17%, Retail MVNOs + Sub-brands: 1,409
1
7% 1%
4%
14%
2040
21% 6%
10%
MVNOs + Sub-brands: 4,960
37%, Other emerging hyperconnected services Example micro-cohorts: Elderly Enterprise Health Banking
EdTech Sports Automotive
…relies on greater customer insights Loyalty Programs
Figure 3. Churn evolution from a selection of global operators
Currently, loyalty programs rely on customer engagement platforms and manual rules to offer generic rewards at unspecific times to try and reduce churn. In the APAC region, operators like Circles.Life, Foxtel, and Optus lack a cohesive loyalty program, attempting to woo customers with untailored offerings. This has limited value with churn rates remaining relatively stable year on year, as shown in figure 3. Better tailored loyalty programs would provide huge scope for improvement.
4%
2%
0%
Customers exit due to anything from poor network quality to competitor marketing. Understanding why customers leave and the stages they go through in the run up enables personalized intervention. Engaging at the right time with the right action is a more efficient and effective means of retention.
2016 2017 2016 20172017 Blended
2018 2018
2019 2020 20192020 2020
Post-paid
Pre-paid
Loyalty schemes in the APAC region
Warehouse MobileSuper Gold Foxtel First Program
My Digi
My OPTUS
Value added services Personalising experiences with niche value added services such as m-commerce, HealthTech, and more cater to smaller customer cohorts. Tailoring services creates opportunity for cross-selling and up-selling to individuals that will increase retention rates and bolster customer lifetime value.
M-commerce
Security
Health Tech
Reduce cost to serve The consumer lifecycle has many touchpoints that come with an associated cost to operators, some examples of which are shown to the right. These are an unavoidable aspects of gaining and retaining customers. To increase lifetime values in a highly developed market, reducing the cost for serving customers is vital.
Location Services
Data Mgmt.
Customer touchpoints that can leverage ADM to reduce cost to serve
• Fraud detection • Churn prediction • Revenue assurance • Promotions
• Billing • Service queries • Sales support • Onboarding
2
ADM for customer retention
AI can help operators improve customer management… It’s time to overhaul customer retention strategies Historically, operators have taken a knee-jerk reaction to prevent customers from leaving with last-minute offers and discounts. This approach is both expensive and ineffective in the long term. A scramble to provide services at lower cost than competitors simply to prevent churn devalues the brand and fails to build a strong long-lasting customer relationship. An AI-led approach allows operators to analyse data on a scale that humans simply cannot, to find patterns and indicators that can help prevent customers from leaving.
AUGMENTED DECISION MAKING (ADM) A system that recommends a decision to humans using diagnostics, prescriptive predictive analytics, resulting in the synergy of human knowledge and AI capabilities
Rules Based Customer Engagement Pre-defined outcomes based on a set of if/then rules that must be coded by humans. • Limited by the size of its underlying rule base • Not always possible to define rules in a programmatic or declarative way
• Difficult to add rules to an already large knowledge base without introducing contradicting rules • Requires more manual reviews and is not scalable
The key to building and maintaining loyalty lies in anticipating and reducing dissatisfaction before the customer attempts to leave
The value of AI is making better decisions than what humans alone can do. Traditional customer management tools work with limited rule-based models on a reactive basis, while ADM with machine learning can perform large scale analysis and see patterns that humans alone cannot (see figure 6).
Augmented Decision Making Has the ability to learn new rules and improve on its own by working on smaller cohort samples and analysing results to refine recommendations • Must go through a training process, but is ultimately more useful • As it does not use pre-defined structures, results can be less predictable • When a ML AI system is incorrectly taught something, it takes a while before it “unlearns” and can determine the correct behaviour
…with an Augmented Decision Making approach AI allows for insights that go beyond what humans can do A successful AI system is a self-feeding loop that gathers inputs, performs analysis and presents recommendations that are subsequently converted into inputs to begin the cycle again, as shown in Figure 4. This ensures the system keeps learning and adapts to new situations. An AI approach to customer retention is a two-pronged strategy. On the one hand, it aids in churn reduction by anticipating dissatisfaction with services, allowing the operator to rectify the situation. On the other hand, it gathers insights that help build customer loyalty by highlighting upselling opportunity patterns and hyper personalised micro-cohorts for marketing efforts. ADM for customer retention is a continuous process Figure 4. Example of ADM inputs, analysis and outputs
AI platforms combine data from multiple sources of the customer journey: • • • • • • • • •
Location Internet usage Device type Demographics Customer care interactions Purchase and billing history Interests Quality of service issues Success of marketing campaigns • Results from previous recommendations • Competitor analysis
Taking the inputs, AI systems perform different types of analysis: • Logistic regression models use historical data to predict the likelihood of certain events
• Natural language processing measures customer sentiment from chat history, emails, social media mentions… • Scenario analysis of proposed offers or campaigns
AI recommends the optimal action, providing answers to questions such as: • Which services and products are the most successful with this specific cohort? • When is the best time to make contact with clients? • Which customers have a high probability of churning but a high potential of being retained? What is the best offer I can give them? • Are customers happy with our services? What are they saying on social media?
3
Benefits of loyal customers
Loyal customers have a higher value than new comers… The benefits of customer retention: Australia MVNO case study Benefits of loyal customers • Up-selling and cross-selling is easier • Organic promotions through brand ambassadors • Loyal customers are resilient: discounts and offers by competitors have a negligible impact
Figure 5. Case study of ADM impact
Augmented decision making
Traditional customer management
By improving customer retention, ADM can have a double impact on operators’ bottom line. Through churn reduction, ADM increases the average lifetime of customers, reducing acquisition costs and through loyalty benefits, increases ARPU. Together, these have a compounding effect on overall revenues. ADM’s impact can be seen through a lifetime value analysis, that is, the average revenue that a customer will generate throughout their time using the service. The following case study shows how ADM resulted in a significant increase in lifetime yield for an Australian MVNO.
Lifetime (years)
x
2.5 years
x
x
Revenue ($AUD/year)
=
Lifetime Yield ($AUD)
$480 ($40*12)
=
$1,200
$480
=
$2,160
2 years increase
47% churn reduction
4.5 years
$60 increase per year
$5 increase in monthly ARPU
4.5 years
x
$540*4.
=
$2,430
…demonstrating the impact of Augmented Decision Making on operators Augmented Decision Making in action The value of ADM lies in an improved customer experience through the generation of insights by collecting and analysing thousands of customer data points far faster and more efficiently than humans alone. These benefits fall into two categories: Figure 6. Benefits of ADM
Monetary
Operational
✓ Analysing sales data (who, when, where) for cross-selling and up-selling opportunities, coming up with personalised offers or packages ✓ Price optimization with specific offers and real-time quote optimization that considers multiple indicators like location, customer transaction history, seasonality, prices of similar offerings
✓ Hyper-personalization for offers, ads, and marketing campaign optimization. By understanding customers better, brands can create a one-to-one connection leading to deeper relationships ✓ Measure and analyse customer sentiment and take proactive measures to rectify adverse situations. Sentiment analysis can mine social media, chats and email conversations to measure customer’s feelings towards services and products
ADM closes the gap between data and decision making Figure 7. Simplified overview of ADM process
Rapid Data Ingestion Forecasting & scenario planning
Real time competitive insights & analysis
AI models for acquisition, retention & yield
Automated decisions
4
Beyond Telecommunications
Loyalty AI solutions are versatile… Augmented Decision Making Framework Media companies
Telcos
Financial services
Utilities
SMS | Portal | Web | App | Chatbots | Contact centre
Decision Augmentation
Delivery
Acquisition
Retention
Monetisation
Offer engine
Churn management
Loyalty
AI TOOLS
Analysis
Scenario modelling
Planning
Prediction
Data inputs Close rate and deal size
Internal sources
Churn rate
Average time to resolve tickets
Customer surveys
Customer service interactions Upsell rate
Location
Customer acquisition costs Behavioural data
Billing history
Customer Lifetime Value
External sources
Data usage Sales cycle duration OTT Usage
Technology innovations Social media
Industry surveys Competitor insights
…and can be used across industries AI data analysis has virtually unlimited applications Any industry that depends on close customer relationships can strongly benefit from AI-based solutions, as seen in the figures below, to increase customer satisfaction and anticipate migration to competitors. Beyond the customer loyalty benefits, the same AI-led approach of machine learning data analytics can be applied to fulfil industry specific needs.
Banking & Financial Services
Healthcare
Personalised financial planning on a larger scale but at lower costs, as well as richer real-time analysis and implementation of financial inputs Logistics
Better patient management and streamlined operations, including scheduling, bed occupancy and recurring prescriptions
Travel Optimisation of supply chains through micro segmentation of consumer habits to ensure the right product in the right place at the right time
Internet of Things
Personalisation of offers, dynamic real-time price adjustment and recommendations that take into account multiple inputs
Public transport & infrastructure
Increased capacity to manage millions of devices due to rich real-time analytics
Retail
Maintenance predictions, more accurate timetables that reflect peak usage and real-time price adjustments Utilities
Increasing shopper engagement and conversion with personalised offers, inventory management and optimised, real-time pricing
Predictive models to anticipate usage and prevent outages and network maintenance on a more personalised scale, plus improved operations
5
Evolved Decision Making
The pitfalls of augmented decision making… Misgivings about AI can be eased with established protocols and transparency As AI systems are capable of training and teaching themselves, faulty data can lead to incorrect conclusions, replicating and amplifying pre-existing human biases. At the same time, AI can be seen as undermining human responsibility in favour of “black box” decision making that is difficult to understand or rectify. Well-thought out AI systems need multiple check-points to ensure constant monitoring of inputs and outputs, whilst measuring recommendations against a set of established parameters.
In decision augmentation, AI is part of the tool box, but not the only tool available. It keeps humans in the loop, ensuring sign off before final decisions and task automation, also allowing for quick intervention when errors are flagged
In terms of privacy, as AI becomes more sophisticated, it can use personal data in more intrusive ways. Data collection and privacy concerns can limit the uptake of AI by companies, consumers and regulators. Sensitive and confidential data must be anonymised to protect customers’ privacy, unless clearly opted in. There are a number of best practices to ensure AI is used in a way that benefits both companies and consumers, but as the technology it still in early stages, it must be improved upon: Work with regulators and law makers to ensure rules and legislation safeguard the interests of users whilst enabling AI innovation Give customers full ownership of their data and options on how it is used
Anonymise sensitive data to ensure privacy while allowing its use in AI-based analytics Poor security may leave companies open to cyber attacks. Secure systems must be at the forefront of any system that deals with private information, regardless of AI use
Ensure input data is of quality and of relevance to the objectives, as well as running regular evaluations to maintain quality.
…are not outweighed by the opportunity it provides The next evolution in decision making is combining human and machine understanding Augmented Decision Making is taking traditional AI to the next level by enhancing existing marketing and campaign capabilities to gain strategic advantage and to rapidly improve business operations.
The benefits of ADM lie in the syn ergy b et ween h u man knowledge and AI capabilities in rapid analysis of high volumes of complex data
ADM is a system that recommends a decision, or multiple decision alternatives, to human actors using prescriptive or predictive analytics. The value of ADM lies in bringing decision-making to a more intelligent level -- a level where important business decisions are made based on all of the available data, including real-time data, with the minimum possibility of human-made errors and bias.
The application of ADM is the ideal evolution for the telecommunication industry. MNO, MVNOs, and IoT providers process large amounts of rich, fertile customer data which are susceptible to mistakes, corruption and or human bias. ADM eliminates these problems, directing more accurate decisions for more successful campaigns and revenue driving tactics. ADM is providing companies with optimal methods to utilise their data, driving revenue and streamline operational overheads. However, the benefits of ADM go far beyond traditional customer strategies to drive revenue. ADM is a more transparent methodology for decision making, so when key decision points are met, auditors and investors can know the ideal option has been taken. Companies should leverage variants of AI for faster, better decision making that would appease both investors and customers.
The Isoton platform ingests data from every user interaction; network usage, geo-location changes, and billing without relying on third-party vendors. ADM is an exclusive feature of the Isoton Platform through a single integration and has successfully taken the retention strategy of some of Australia and other countries’ most well-known telecommunication brands to the next level. What will your next move be?
Isoton Pty Ltd Greg Steer gsteer@isoton.com Level 1, 206 Greenhill Road Eastwood 5063 South Australia +61 8 8372 9000
PT Isoton Insight Indonesia sales@isoton.com Noble House 30th Floor, Unit 30-107, Jalan Dr. Ide Anak Agung Gde Agung Kav, E No.4.2, RW.2, Kuningan Kecamatan Setiabudi, Kota Jakarta Selatan, 12950