Personalization @Intuit — Part 2 Functional Overview

Personalization @Intuit — Part 2 Functional Overview In Part 1 we looked at some of the scenarios on how we Personalize to create customer delight in their interactions with us. In Part 2 we will look at a functional framework in how we go about solving for those scenarios. Let’s start with the basic definition of Personalization. […]

Personalization @Intuit — Part 2 Functional Overview

In Part 1 we looked at some of the scenarios on how we Personalize to create customer delight in their interactions with us. In Part 2 we will look at a functional framework in how we go about solving for those scenarios.

Let’s start with the basic definition of Personalization. Personalization enables suggestion of message or content that is relevant to the individual user, based on user’s implicit behavior (Optimization) and/or explicitly preferences (Customization). Optimization uses implicit interests and learns what you like from your actions. Customization on the other hand is driven by explicit preferences. Optimization usually is at the message/content served level while Customization is at a “feature” level, where you see/hide etc. what you want explicitly through QuickBooks Account Preferences.

Concepts

There are several distinct facets to Personalization

Who is the customer/prospect (and how much do we know about him)?

  • Identity — information about each individual customer’s identity — user, small business entity, visitor (web/mobile), social & email Id.
  • Profile — Demographics, activity, profile, transaction history, and so forth.
  • Behavior — information about customer’s interactions.

What gets personalized?

  • Message — optimize the message are you trying to convey.
  • Content– optimize the content — which can be for the entire page,a module or a component.

Where is the interaction happening — These are the touch points across the entire QuickBooks online ecosystem.

When is the personalization decisions made?

  • Before — this usually is based on past behavior, actions and analytical segments.
  • During — generally higher payoff as it leverages contextual data to optimize.

How is the personalized message selected? — There are two different approaches to how the personalization decision is made

  • Deterministic — generally through static data . While this approach does give more control over the segment definition and makes it easy to anticipate success, it has limitation as data may quickly become stale or irrelevant.
  • Predictive — uses machine learning algorithms to predict the best course of action. The models generally are adaptive, iterative and self-improving.

Why is real-time personalization important?

The ability to connect with customers on their own terms, during their interactions is important and can dramatically improve the effectiveness of engagement efforts. With this background, Personalization can be broadly defined as optimizing for an end goal or an Objective that leverages characteristics like segments, user preferences, locale and context.

  • Objective — The end goal that we want the user to accomplish or the function on what we want to personalize upon.
  • Segment — A collection of attributes that group users.
  • Context –Account, Transaction, Activity or User Session.
  • User Preferences — explicit preferences or settings.
  • Locale — product for the local market (intersects with Globalization).

Personalization,Globalization and Experimentation

Personalization builds upon Globalization and Experimentation. The following section provides a conceptual framework on how you can coherently integrate these loosely coupled systems to tailor your products for your customers.

Personalization vs. Globalization

Globalization (G11n) is about building internationalized (I18n) products & services for the global market and then localizing (L10n) the product capabilities per locale — menus, sitemap, currency, language, country, user registration rules. Personalization builds on top of a compliant product features & capabilities which are available for a market.

Personalization vs. Experimentation

Experimentation is the process of testing & learning on visitors/users by serving two (A/B) or more distinct experiences on a population — i.e. multiple users simultaneously. For e.g. an experiment may have two factors each with two levels — Layout (top or left) & Creative (creative 1 or 2) served to distinct random populations. Experiments are a function of factors served to populations with the objective to measure a statistical significant lift.

Personalization on the other hand is about personalizing the experience of a specific user. It is similar to Experimentation as the building block is factors, however unlike Experimentation, has coefficients or weights that are evaluated on a particular user (and not on a random population).

While both Personalization and Experimentation has the lowest common denominator as factors, each factor is mutually exclusive for a given scope (product flow, page, module, component,etc). To illustrate, hypothetically, if page layout is a factor on the page you are experimenting, you would not experiment and personalize on layout simultaneously. Assuming you are personalizing on layout on the page, you may subsequently perturb on the creative diffrent content and serve say creative 1 or 2. The deviations to a factor (creative 1 or 2 in this example) is often called Perturbation — as you are continuously experimenting to refine the model. The interaction of Personalization with Experimentation is shown below.

To conclude Part 2 , Personalization is a message or content that is relevant to the individual user, build on top of Globalization and interacts closely with Experimentation. In Part 3, with this background and foundation, we will look at how we have built out a Personalization platform that scalably solves for customer delight.

Interested in working on these areas to power prosperity for Small Businesses? We are hiring…


Personalization @Intuit — Part 2 Functional Overview was originally published in QuickBooks Engineering on Medium, where people are continuing the conversation by highlighting and responding to this story.

Source: Intuit