Digital Life cycle – AI Use cases

There is no 'one definition' in our industry for what constitutes #digital or Digital transformation as it keeps evolving. Some people understand Digital as newer channels of Customer communication while some others realize that in order to be Digital your entire organization & many departments need to be an enabler. This article is aimed at explaining how #ai can be applied at each step of your Customer digital life cycle, what are the challenges at each step and some ideas around how to overcome some of these challenges.

1.????Target Market Segment

This is the first step in your Customer Digital life cycle which involves identification of your target customer segment. If its an existing product or service to be marketed, then you’ll have historical data of Customers who have availed this product or service and those who haven’t. You can create a target Persona based on the above Customer variables using any of the statistical techniques like k-means clustering, PCA, common factor analysis, discriminant analysis, etc. If it’s a new product or service to be launched, then you can design a market survey & collect data of Customers or prospects who would like to avail such a product or service and those wouldn’t. Conjoint analysis is one technique which can be leveraged.

2.????Digital Marketing channel

Now that I know my target Customer from step 1 above, the next question is where do I find this Customer? How do I know where are my Customers spending their time digitally? Whether they spend more time on social media or on eCommerce sites or on their WhatsApp, etc. The question I need to answer is what would be the preferred channel for each of my Customer? Should I reach out to them on Email, WhatsApp, SMS, Social media, cold call, etc? For existing Customers, we’ll have historical data around the click response on each channel, cost to campaign on each channel, Customer contact details, his digital activity data, etc which can be used to arrive at the preferred channel of communication for each Customer using modelling techniques like ensemble learning, etc.

3.????D2C campaign execution

Now that I know my target Customers and also their preferred channel of communication, the next question to be answered is what is the right communication to be sent to my Customer? Gone are the days when you used to send one standard nudge or message to all Customers or even a target segment. In today’s Digital world, you need a Martech platform which can manufacture a personalized nudge for each of your Customer. There are many COTS product vendors like CleverTap, MoneyThor, Knowesis, Unica which can be a part of your Martech platform stack but you need to build your platform in such a way that you are able to execute campaigns on batch as well as real time basis with a clear segregation of duties between your each platform layer such as data layer, CDP, nudge manufacturing layer, nudge disbursal layer, channel end point communication, CMP, etc.

4.????Campaign performance measurement

Now that you have identified your target Customers, arrived at their preferred channel and sent personalized campaigns to them, the next question would be to determine if you’ve been doing the right thing all along. Building & deploying any AI /ML solution is only 50% of the challenge, the remaining 50% is always around how to measure the success of your AI/ ML solution and improve your solution based on this feedback data. In the Digital world, it is possible now to measure your Customer clicks on each of your campaigns on different channels. You can use Google Analytics, Firebase, etc to collect your click data and analyze this real time streaming data in a Big Query or similar environment to come up with campaign performance metrics like 'campaign to click ratio', channel wise click ratio, segment performance, content performance, offer performance. You can use UTM parameters to identify your different campaigns and these need to be passed via your campaign link to the destination website or App for further processing.

5.????Conversion Funnel drop offs

You have done everything that you could to identify your target Customers, arrive at their preferred channel, successfully nudged them and brought them to your website or App for final conversion. But then, what if your website or App journey becomes the show stopper? You need to analyze what happens at each step of your Customer journey, where are the drops happening, what could be the reason for the drops and then come up with the right interventions. There could be many reasons for the drops such as 'too many inputs required from the User', too many screens to hop before conversion, wait time for internal validation, OTP didn’t arrive on time, session timed out, no autofill for returning user details, no Readme or guided tour, no explainer videos for help. You can further deep dive into whether we are rejecting Customers based on our stringent risk & compliance rules, whether these rules are mandatory or have been applied over a period or time, whether they are relevant in today’s situation. In areas such as lending, firms come up with a lot of internal models for income estimation, credit decisioning, line assignment, offer finalization, risk assessment and we need to be constantly evaluating these model outputs as we may be losing some good Customers at this stage. We need to leverage alternate data as well as partner level scores to help us understand our Customers better wherever we lack sufficient data. Some firms like Experian, Transunion provide industry utilities for data room exercises in the Financial services world whereby you can combine your data with your partner data along with Bureau data to come up with an improved understanding of your Customer & the risks involved.

6.????Portfolio deepening

The last & final stage of Customer Digital lifecycle is after you have successfully converted your Customer but now need to cross sell more products to ensure that you have long term stickiness. At this stage, you need to know what products to pitch to each of your Customers and in what sequence. You already have historical data of important Customers and 'not so important' Customers from the value they provide to your firm. You can build models to identify what factors can contribute to a long term higher PPC (Product per Customer) whether it is an early registration on your mobile App, high no of transactions, any specific digital product that will lead to more Digital activity, etc. Different companies have used multiple linear regression, decision trees, random forest techniques?to arrive at these factors & their relative importance. Based on this analysis, you can design a sequence of nudges to be shown to each Customer in the first 100 days of onboarding as an example.