Chief Architect at Pelatro. Proud to help 40+ Telcos/BFSIs offer the finest contextual marketing experience to their 1B+ subscribers.
getty
CMOs across the world are optimistic about transformational outcomes from AIML. The meteoric rise in popularity of GenAI has created a new problem—an overnight supply of GenAI experts and software, albeit with zero track record in native business cases. Companies and consultants who possibly had not a single mention of generative models in their websites a year back now claim to have turnkey solutions backed by GenAI!
AIML, Generative Or Otherwise Are Just Means—Goals Come First!
Enterprises need to carefully unscramble the puzzle and walk out of the fluff. Advanced AI is already employed to solve business problems in core marketing functions such as content creation, social media listening, product description synthesis, etc.
Telcos already use a lot of ML for What-If analysis, segment discovery and NBA. Marketing analysts are prudent at running hypotheses to arrive at the following.
• Right threshold for spending CTAs ($X on prod Y by time T).
• Right segment from their millions of subscribers for a particular offer (10 GB for $X).
• Right offer in customers' micro-context (100MinVoice+5GB Data for 7 Days).
They are not equally comfortable arriving at long-term prescriptions for their customers and answering questions like "What are the newest customer behaviors influencing their spending patterns?" and "Who are the customers who are very much unlike who they were earlier and why?"
Purpose-aware hyper-personalization and marketing entropy reduction hold a lot of potential for advanced ML leverage in telcos.
Advanced Analytics Help Telcos Listen To Untold Needs
To provide a cognitive mapping from a business domain to mathematical models, consider the four functional elements "Hear (H)," "See (S)," "Think (T)" and "Do (D)," which are based on fundamentals of empathy mapping.
Hear corresponds to what customers experience in the form of personalized communications or offers (without necessarily registering the same in their minds).
See corresponds to those events or offerings that are, in fact, registered in the minds of end customers but without necessarily connecting to the business objective and/or purchase.
Think corresponds to those events or actions that seem to be getting deeper on an offering and/or toward a purchase, which may be observed through a transcendence of the face value of the offer/commodity/product/service.
Do corresponds to concrete actions, including soliciting messages, purchase leads, purchases and verbal endorsement(s) by end users who are convinced to get what they are determined to pursue.
Both Hear and See axes are favorable to telcos as they have direct control over communication channels and are well-positioned to reach out to their customers at exactly the right moments. But telcos have not been very successful at riding on their strengths on H and S in making strides on the T and D planes.
ML Is The Bridge To Better Understand Customers And Nudge Them Along
Telcos need to approach customer-centricity and personalization goals quantitatively and measure them using metrics like CSAT and ARPU. Although using revenue to account for personalization sounds counterintuitive, personalization when delivered the right way catalyzes events along the Think and Do axes eventually culminating in greater levels of spending and engagement.
Advanced ML algorithms, including GenAI-backed models, may be used by telcos to compile HSTD scores for various combinations drawn from sets of products, offers, plans, channels, segments and CSAT initiatives. Think of it as a book where one can look up by persona and intent to find answers to questions like the ones below and then use that as the driver for subsequent decisions.
What is Alice's awareness of our bundled data-voice combo? What is its relevance? Has she seen our weekly subcription plan? Is she thinking of a switch?
Who best defines the class of "self-driven customers"? Those with assertive actions along the D axis regardless of events along the H and S axes?
Generative ML models may be employed to discover the not-seen-before elements under H/S/T/D axes and understand their complex relationships which in turn may be used to arrive at the right prescriptions to achieve desired values under D axes using H and S as means, and T as the facilitator.
ML Is The Lever To Reduce Marketing Entropy
Entropy is broadly the degree of disorder or uncertainty in a system. From a marketing perspective, it translates to a measure of lack of coherence in different marketing activities. A disparity between strategic objectives and the tactical efforts in that direction adds to entropy in the system.
For example, in the telecom context, base revenue and acquisition revenue are two important revenue sources, and even within the same region, depending on various factors, including market share and growth strategies, different telcos have different targets for the two. Along the year, based on relative progress under respective baskets, the allocation of marketing budgets, including the intensity of campaigns, nature of offers and frequency of promotions, may all have to change. When marketing activities do not continuously adapt and re-calibrate themselves to stay in line with strategic objectives, overall entropy is on the rise.
Telcos may leverage digital twins and employ deep learning models to understand the complex interrelationships between marketing efforts, their effects on customers and the net result on business, and use marketing entropy as a watchdog to holistically assess synergies in marketing activities. Given the telecom scale (millions of subscribers, billions of transactions, thousands of offers, hundreds of campaigns, dozens of partners), and the lack of formal techniques to effectively contain marketing entropy, advanced ML comes as a viable means of bridging the marketing divide.
It Is Time To Widen The AI Horizon!
Telcos were one of the early adopters of AIML and have seen it add value to their CVM bottom lines. It started a decade ago with the leverage of descriptive analytics at scale for behavioral insights followed by diagnostics, especially in operational areas, including discovering reasons for customer dissatisfaction and churn. Then followed the predictive era with propensity models for various actions and subsequent remediation effectiveness taking the front stage. Now is the time to adopt prescriptive analytics, including generative models, to deliver strategic long-term objectives that drive both customer centricity and business profitability.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
