The Product Optimization Framework for Customer-Centric eCommerce
- Customer lifetime value has three levers: marketing (what you say), merchandising (what you sell) and customer experience (what you do). Product optimization works on the second.
- The framework has four data layers: sales volume, customer experience, behavior and intent. Use the last 12 months of data and the same set of products for every layer.
- Toxic products sell well but create dissatisfaction and churn; sticky products create repeat purchases and loyalty. The goal is to remove the first and add more of the second.
- The buy-to-detail rate (orders divided by product views) measures intent; purchase frequency, return rate and the Average RFM Impact Score measure post-purchase behavior.
- Aggregated data is analyzed either with a weighted priority score per product or with a high/low two-point scale that sorts products into eight categories.
Customer-centric eCommerce companies need a reliable product optimization framework to find out whether their product assortment helps them reach their goals: acquiring better customers and turning new customers into repeat and loyal customers. The framework below judges each product on four data layers (sales volume, customer experience, behavior and intent), so you can see which products build customer lifetime value and which destroy it.
If you focus on customer-centricity, increasing customer lifetime value is your top priority. There are only three things your departments can do to improve this metric:
- Marketing: what do you say?
- Merchandising: what do you sell?
- Customer Experience: what do you do?
Product optimization is one of the most impactful pillars of CLV, yet it is often neglected and usually left to the merchandising department. It should be the result of collaboration between departments. Instead of working in separate silos, shift towards customer-centricity and use a framework that analyzes every layer of product performance.
What the product optimization framework helps you do
-
Get rid of toxic productsToxic products look good in terms of sales volume but generate customer dissatisfaction and high churn rates. Identify them by looking at what first-time buyers and loyal customers purchased before they never came back.
-
Prioritize the sticky productsSticky products generate repeat purchases, high customer satisfaction and loyalty. Identify them by looking at the products, brands and categories your best customers prefer and keep buying.
-
Add new sticky products to your offerYou need more sticky products to increase sales and customer lifetime value. To identify new ones, look beyond sales volume at the factors that influence CLV: customer behavior and customer experience.
With a better product assortment, you can create better offers to attract new and existing customers, build better experiences around your most loved products, and generate more loyal customers.
The four data layers of the Product Optimization Framework
The product optimization framework was created by the Omniconvert team and is an essential part of the customer value optimization process. It helps customer-centric companies look beyond typical product performance metrics and understand each item's impact on customer experience and satisfaction.
| Data layer | Questions it answers | Key metrics | Where the data comes from |
|---|---|---|---|
| 1. Sales volume | Which products sell best? Which are most profitable? | Customers, orders, revenue, margin | Analytics tool or order management system |
| 2. Customer experience | Which products are most appreciated? Which create promoters or detractors? | Net Promoter Score, review score | NPS surveys, review app, competitor reviews |
| 3. Behavior | Which products ruin customer relationships? What do loyal customers buy? | Average RFM Impact, purchase frequency, return rate | Customer and order data, RFM segmentation |
| 4. Intent | Which products convert the visitors who view them? | Product views, orders, buy-to-detail rate | Analytics tool |
Layer 1: Sales volume data
Most stores analyze product performance by looking only at sales volume. You can extract this data from tools like Google Analytics or your eCommerce order management system. Sales volume data answers two essential questions:
- Which are our best-selling products?
- Which are the most profitable products?
Export these fields for each product: product ID, category, brand, customers, orders, revenue and margin.
From a customer-centric perspective, sales volume data is necessary but not sufficient. The next three layers analyze product performance with the customer in mind.
Layer 2: Customer experience data
The customer experience layer shows how different products generate different levels of satisfaction. It answers two essential questions:
- Which are the most appreciated products?
- Which of our products create promoters or detractors?
To find the most appreciated products, look at product reviews. Export them from your review app. If your shop does not have enough reviews, do some market research and look at stores that sell identical or similar SKUs. Competitors' reviews can also give you ideas for new sticky products to diversify your assortment.
To see whether your products generate promoters or detractors, export the Net Promoter Score by product. Pre-delivery NPS evaluates the shopping experience. Post-purchase NPS evaluates the experience customers have when they use the product. Together, they show what customers love and are likely to recommend by word of mouth or on social media, and what causes a poor experience and churn.
Export these fields for each product: product ID, category, brand, customers, orders, revenue, margin, Net Promoter Score and review score. Ideally, you have both experience metrics, NPS and review score, for every SKU.
Layer 3: Behavior data
The behavior data layer analyzes what happens after purchase, depending on which products customers bought in the last 12 months. It answers two questions:
- Which products are ruining our relationships with customers?
- What are our most loyal customers buying?
Export three metrics per product: Average RFM Impact, purchase frequency and return rate. Calculate them as follows:
- Purchase frequency = Total number of orders / Total number of customers
- Return rate = Number of returned products / Number of sold products
- Average RFM Impact Score = SUM (density in the total sales in each RFM group × CLV/CAC ratio per group)
Worked example (illustrative numbers). A product has 1,200 orders from 1,000 customers, so its purchase frequency is 1,200 / 1,000 = 1.2. Of 2,000 units sold, 160 were returned, so its return rate is 160 / 2,000 = 8%. Suppose 60% of its sales come from an RFM group with a CLV/CAC ratio of 4 and 40% from a group with a ratio of 1.5. Its Average RFM Impact Score is (0.6 × 4) + (0.4 × 1.5) = 2.4 + 0.6 = 3.0.
A high RFM Impact Score means the product sells to the customers who bring the most value relative to what they cost to acquire. RFM segmentation gives you the groups this score needs.
Layer 4: Intent data
Most eCommerce companies look at conversion rates by traffic source, location or device. For product performance analysis and assortment optimization, you need the conversion rate by product, also called the buy-to-detail rate:
Buy-to-detail rate = Total number of orders / Total product views
Example: 150 orders from 5,000 product views = 150 / 5,000 = 3%.
Export these fields for each product: product ID, category, brand, product views, product revenue, orders and buy-to-detail rate.
How to analyze aggregated product data
Once you have the data for each layer, you are close to identifying the products that will support your customer-centric approach. You have two alternatives:
- Priority score calculation
- High/low two-point scale
Priority score calculation
-
Give a score for each attributeNormalize the data. Score each attribute on a scale from 0 to 100. For example, the best product in terms of the number of customers attracted gets the highest score.
-
Prioritize attributesWithin each data layer, prioritize the attributes according to what you want to optimize. For example, in the sales volume layer, the most important metric could be margin or customer count.
-
Establish a weight for each data layerSet the weight of each layer according to your business goals. If you rely heavily on customer acquisition, give a higher weight to the sales volume and intent scores. Otherwise, focus on customer experience and post-purchase behavior.
-
Calculate the priority score per productThe priority score is the sum of the scores for each layer. The higher the priority score, the more resources you should invest in making the product visible to potential and existing customers.
Worked example (illustrative numbers). A retention-focused store weights the layers at 20% sales volume, 30% customer experience, 30% behavior and 20% intent. A product scores 80, 60, 70 and 50 on those layers. Its priority score is (0.2 × 80) + (0.3 × 60) + (0.3 × 70) + (0.2 × 50) = 16 + 18 + 21 + 10 = 65.
High/low two-point scale
The alternative to the priority score is a two-point scale, high or low, to evaluate sales, retention and customer experience. You still calculate the score per attribute for each data layer, but you do not calculate a priority score. Based on the aggregated score, you decide whether each value is high or low.
Three areas, each high or low, give eight combinations. That splits your products into eight categories and makes it clear how to optimize your assortment.
The eight product types and how to handle them
| Product type | What it usually signals | How to act |
|---|---|---|
| True gems | Your best-sellers | Give them the attention they deserve: feature them in on-site product recommendations, build ads, and create excellent experiences around them. |
| Fake diamonds | Low-quality or over-promising products, seasonal products, or products with low consumption or long purchase cycles | Find out which of these causes applies before you invest more in them. |
| "Needs more polish" | A problem in customer experience | Investigate further and decide whether to replace the product or the supplier. |
| "Shines, but not for long" | Low retention caused by the purchase cycle, seasonality or a lack of marketing effort | Look for ways to increase retention for these products. |
| Potential hidden gems | Low sales volume caused by a lack of proper marketing or low market demand | Check which cause applies; if it is marketing, give them more visibility. |
| "Rare, but loved" | Durable, long-lasting products that customers appreciate but that are not marketed enough | Market them more. |
| Rare semi-precious | High retention caused by high necessity and few options on the market | Keep them available and protect the repeat demand. |
| Simple stones | Low values for all data layers | Stop selling them before they cause more harm. |
The framework takes time, but it is one of the most effective ways to analyze and optimize a product assortment with a customer-centric mindset. It helps you design more customer-centric experiences and manage products based on multiple data layers, including customer data. As a result, your store can improve acquisition and retention with better offers built around an optimized product assortment.
If you do not have enough data or resources for this quantitative analysis, start smaller: spot toxic products by monitoring the order return rate by product, brand and category.
The analysis also identifies products that need further investigation through qualitative research. For the customer layers, Customer Intelligence in Nexus by Omniconvert gives you real-time RFM segmentation (Soulmates, Loyal, New, Promising, About-to-Dump and Breakups), CLV tracking by segment, cohort and channel, and NPS broken down by RFM segment, product and location. Once you know which products to promote, Omniconvert Explore lets you run experiments on product pages to test how you present them.
Frequently Asked Questions
A product optimization framework is a structured way to evaluate every product in your assortment against more than sales. The Omniconvert framework combines four data layers: sales volume, customer experience, customer behavior after purchase, and purchase intent. The result tells you which products to promote, which to investigate, and which to stop selling.
Toxic products look good in sales reports but damage customer relationships. They create dissatisfaction, returns and churn. You find them by looking at what first-time buyers and previously loyal customers bought before they stopped buying from you, and at products with high return rates or low Net Promoter Scores.
Sticky products generate repeat purchases, high customer satisfaction and loyalty. You find them by looking at the products, brands and categories your best customers prefer and keep buying. Sticky products deserve priority in recommendations, ads and on-site experiences because they increase customer lifetime value.
Extract data for the last 12 months and use the same set of products in all four data layers. This keeps the analysis comparable across layers. If you do not have enough data or resources for the full analysis, start by monitoring the order return rate by product, brand and category to spot toxic products.
The buy-to-detail rate is the conversion rate of a product page. The formula is total number of orders divided by total product views. For example, a product with 150 orders from 5,000 product views has a buy-to-detail rate of 3%. It is the metric of the intent data layer.
The Average RFM Impact Score is the sum, across all RFM groups, of each group's density in the total sales of a product multiplied by that group's CLV/CAC ratio. A product that sells mostly to groups with a high CLV/CAC ratio scores high, which means it attracts or keeps valuable customers.
Both start from a 0-100 score for each attribute. The priority score adds weights to attributes and data layers and sums them into one number per product, so you can rank products and allocate resources. The two-point scale skips the priority score and marks sales, retention and customer experience as high or low, which sorts products into eight categories, from true gems to simple stones.
Product optimization is usually left to merchandising, but it works best as a collaboration between departments. Marketing decides what you say, merchandising decides what you sell, and customer experience decides what you do. All three influence customer lifetime value, and all three need the results of the analysis.
Along with CLV and cohort analysis, product assortment optimization is one of the most important elements of quantitative research and an essential pillar of a customer-centric strategy. The framework takes time, but it shows which products build loyal customers and which quietly drive them away. Use it to design better offers and experiences around your most loved products, and follow up with qualitative research on the products the analysis flags.
See which products your best customers buy
Nexus by Omniconvert scores every customer on recency, frequency and monetary value, tracks CLV by segment, cohort and channel, and breaks NPS down by RFM segment, product and location. Built on 13 years of customer data across 7,000+ websites and 15+ industries.