Maximizing the 'Lifecycle Value' of Items in E-Commerce
Product 'Lifecycle Value'
Many e-commerce companies make more than 50% of their revenue from products introduced in the last three years and every product has a lifecycle with four phases: introduction, growth, maturity and eventual decline.
The Need For A Systematic Method For Product Introductions
A problem for most e-commerce companies is that they lack a systematic method to introduce and promote items to the members of their audience who are most likely to be interested - and to do this at scale. A common set of constraints creates this problem:
- New products have no purchase history or traffic data
- It’s difficult to match a new product to customers beyond general segments
- It’s difficult to identify trending products and promote them in a targeted manner
A Look at Product Lifecycles to Better Understand the Issue
For a typical new product, it’s lifecycle looks like this:
There are four stages: introduction, growth (varies for each product by the amount of success), maturity and eventual decline. A process for managing a product’s lifecycle would focus on the tasks that are required at each stage of the lifecycle.
For the purposes of this discussion, we will focus on the introduction and growth stages.
The Introduction and Growth Stages
During the introduction stage there are a number of key requirements:
- How to apply automated matching of user interests to appropriate new products in their introduction phase without any product history.
- Understanding which customers will be most interested in a product and ensure the product is shown to these customers in emails, onsite, and in mobile experiences.
- Once introduced, a product’s views, clicks, purchases and other data need to be tracked and used for iterative learning. This allows promising products to be identified.
- The ability to accelerate exposure (depending on feedback) by introducing the product to more users in a highly targeted manner using their profile preferences.
Problem: Not All Products Have the Same Lifecycle
A system that achieved these goals would be a good start. However, few products follow this lifecycle pattern precisely, and some products are so popular that they swamp the lifecycle for many others:
This dynamic creates another set of problems that need to be solved to enable successful and systematic lifecycle value harvesting. The key problem is that the most popular products often crowd out other products in the display areas of rudimentary discovery tools like top seller lists, collaboratively filtered (crowd-sourced) recommendations and filters like ‘most viewed’. This results in an experience where the customers are seeing the same products repeatedly (most e-commerce sites today suffer from this issue).
To improve this situation, several problems need to be addressed:
- How to promote highly successful items to a targeted audience so they can reach their full potential without crowding out other moderately successful items.
- How to maintain visibility to mature and moderately successful products while popular items.
- How to make product ‘room’ available in key channels like onsite category pages, promotional emails, lifecycle, and triggered emails, while ensuring that popular products are also promoted.
A process for maximizing lifecycle value can be designed to address these key issues. This system’s key goals should be:
- Making new items available for discovery and recommendation nearly immediately upon being added to a catalog. (Addressing the issue of minimum volume or product history).
- Identify items with positive trends and amplify success without crowding out other items in key channels like email and onsite. (Offering targeted content individualization that doesn't rely on crowd-sourced data like top-sellers, correlated purchases, most viewed, etc).
- Provide a better user experience by identifying and matching product interests, rather than showing repetitive product offerings. (developing iterative product exploration that helps identify a customer’s interests. Maintain customer profiles that dynamically adapt to ever-changing tastes and preferences).
A process that achieves these objections could look something like this:
Product Introduction Process Flow - Ideal State
Looking at each step, we can dive deeper into the details that make it functional.
Add Item to Catalog: Once an item is added to your catalog, the product feeds you generate will need to capture changes and share with supporting systems.
Indexing your Catalog Data: The best practice is to use semantic technology to extract attribute level information from each product and store it as tags. This technology should ‘crawl’ the catalog many times a day in order to detect and surface any changes - including the addition of new items. A huge benefit of this approach is that new products and product updates can become immediately available for use in individualized content selections.
Predictions of Customer / Product Match: The process will need a ranking capability to match products with customers. A sophisticated approach is needed that evaluates a product’s characteristics and matches them to a user’s profile, weighing various attributes and factoring in context. Using customer attribute level profile ranking allows products without traffic or purchase history to be matched to a customer.
Delivery Requirements: The needs for delivering individualized content into each channel differ dramatically and require mechanisms to handle unique specific issues. For example, the technology needed to handle spikes in predictive demand related to promotional email volume is computationally intensive for short durations and requires 'pre-rendering' of any item that might be ranked for selection. And when delivering predictions onsite, the imperative for fast page loads creates the need for a super fast response time.
Good predictions with poor presentation quickly fail: An impactful presentation has a large impact on the user experience. An ideal process ensures an appealing diversity of items in each presentation (e.g., most likely items to buy, new items of predicted interest, explorative items) and manage what is presented to prevent duplication or repetitive groupings. In order to achieve this, the system needs to possess intelligence around all the items that are being presented together to ensure each collection is cohesive. This crucial step prevents a recommendation that contains six different types of an item that are nearly identical (which is fairly common today in e-commerce experiences onsite and in email).
Below left is an example of a collection that fails to ensure diversity in its presentation. The prediction of interest in the displayed item may be accurate, but the overall presentation is repetitive and will deliver poor results. Below right is an example of a collection that contains top predictions, but it has also been through analysis to ensure a diversity of items. The diverse collection offers an opportunity to introduce new items as well as cross-sell additional categories.
On Left - Repetitive Presentation| On Right - A Diverse Presentation
By ensuring diversity in each collection, customers are exposed to a far larger percentage of your catalog which drives interest and engagement. And in the process, you also create space to offer new items that are predicted of interest for a user.
Scaling the Product Lifecycle Value Process
For any given product, this process is manageable. But when you multiply it for hundreds or thousands of new items per year, and then tailor it for each customer, the problem becomes unmanageable. A key factor for success is to understand and refine the process for introducing items and then applying automation at key points in the process:
- Distribution of catalog information through product feeds.
- Automated extraction of descriptive product information at the attribute level and making it actionable.
- An ability to rank matches between customers and items at open / view time.
- The capacity to deliver ranked customer / item matches onsite, into emails, and apps.
- The ability to intelligently present collections of items personalized for the context of each user.
- The ability to collect a user’s active and passive data and blend it with real-time product performance data and make it actionable for ranking.
Undoubtedly, there are more areas to the process where effective automation can be applied, but this list is a good place to start.