Predictive CX: Enhancing Platform Experience

(This article has 3 sections and is a 5 minute read)

I) Context

Modern platform experiences are complex and user expectations are sky high. Experiential economy is an high stake game where there is a premium paid for better user experience and the cost paid is usually in terms of loosing the customer or the business overall.

Platform experiences is generally from the point of view of task success (app open, make a payment, video attempt success etc).

Note: This is not the CRO/Funnel Experience optimisation use case as usually is understood where data can traditionally play a role. Data can, should, and have increasingly been able to play a role in platform experience optimisation (e.g. app load time improvements etc).

Let's just talk about app open as a task success attribute as it's common to most D2C apps irrespective of business they drive.

E.g.

• Is the app opening successfully what % of the user base?
• What % of the instances? (crash rate %)
• How quickly is the app opening? (app load time etc).

1.a Why is this key to Customer Experience?

As app load time and speed determine the performance, it is crucial that you optimize it.

Statistically speaking, the ideal load time of an app should be 3 seconds. In fact, according to Google, 53% users abandon a mobile website that takes more than three seconds to load. This may be acceptable up to 5 to 8 seconds.

1.b. What factors will be at play ? (We can see a variety of attributes and data points are at play - Hinting: role of predictive CX)

• Client Side - Mobile Capabilities and Resources
• Server Side - API calls order and optimization (if CDN is involved then the optimization thereof)
• Network Side - what network the client is on and the speed etc.

1.c. How do smart data centric organisations (& CX functions) improve it?

Platform CX can be improved by using a mix of all following approaches Technology, Design and Data Science

• Technology Approach: Tech call optimization what APIs to call first what to call later basis the tech user flow needs
• Design Approach: Improve perception of speed Often as a design approach companies also utilize the perception of speed concept by making the welcome page to inform and engage the user with the brand or launches while the app loads with the landing page expectations etc in due time)
• Data Science Approach: Predictive CX Approach Here's where the science approach can be used by utilising the variety of data at our disposal from client side logs, server calls logs and network side attributes in addition to behavioural attributes typically captured by customer experience analytics team to arrive at a predictive approach to task success optimisation outcomes.

?

?

II) Predictive CX Approach & Challenges

Platform Experience Optimisation use cases using AI - driving device level optimisation of experience using attributes native to user, device and network situations during browse/consumption.

These works on principles of what can be called early (pre-fetched), what can be delayed (deferred) and what can pre-known (cached).

Challenges:

• Predictive CX is hugely underrated. Use of data and analytics to enhance tech processes is hugely under-utilised.
• Part that it can play to platform experience optimisation is understood but under-employed.
• Maybe because it sits at cross sectional skill set behavioural analytics and core tech prowess. It's almost impossible to find leaders who will have both skill set as of today..... thought leadership is required in terms of setting such POD that will have specific focus on this area that gets lost amidst feature roadmap and front end feature/design optimisations.
• And, while all of these are done BAU within the tech organisation for ages, they become extremely potent with AI as guiding arsenal

Mature organisations that make cross-functional teams (Tech + Data) across tech and data at charter cadence usually find the benefit of this approach. Been lucky to part of such charters across the organisations I've worked with in both my current and past roles.

Enumerating some of the case studies for end user understanding (masking org names for confidentiality - so let's just focus on high level use cases )

III) Predictive CX : Case Studies

Below we want to highlight couple of use cases basis 'Pre-fetch' and 'Cache' for optimising app load times (CX) using predictive power of behavioural & temporal data

Use Case A: Pre-fetch

Predicting Session App Depth and pre-fetching experiences during the app load surgically to improve overall CX

• App Start Up experience considers minimum time required to render landing page experience (Home Page – with statement rendered).
• Hence, it tries to defer calls such as those to membership rewards experiences that are deep-seated
• Usually, customer’s drop off from the timeline as the engagement is high here (most basic things needed are services, my balance, my last transaction and details of transaction led servicing such as fraud, dispute and pay later)
• However, there will always be session behavior (profile, browsing or time of month based) where customers will feel the need to go into the ‘depth of the app’

Couple of key decisions

a) Can we predict will the session end with app depth ? (session moving to at least one tab other from home page)

b) Can we predict which tab / experience that will be? (e.g. Membership rewards, offers or account management etc)

?

Use Case B: Cache

Caching key behavior attributes for better /faster rendering differentiated experiences

• Apps with differentiated experiences on launch (e.g. Freemium vs Premium only variants of app - that are lets say different by geography) will require caching of pre-attributes depending on user profile (new vs repeat) to thereby enhance the overall CX in terms of app load
• Imaging same case – if the geography is a differentiates the experiences strategically (Fremium vs Premium) – then it’s embedded in user past behaviors (changing country on app launch, new user vs repeat user etc) and travelling profiles (which are baked in kind of phones, OS etc)
• Tech teams can work with data inputs to optimize basis such data driven prediction of travel and cache the base country for ‘N’ days -? for repeat sessions basis traveler profile and travel patterns, thereby expediting the experience for the end users across these profiles.
• Limit error rates by supreme predictions of such behaviors (the regret of prediction)

Couple of key decisions

a) Can we predict travelling customer attributes?

b) Can we predict base period to cache per customer or for a segment of customer?

?