Pillar Two · Understand

Understand the Product, Not Just the Customer

SKU-level product intelligence for post-purchase product analysis

A brand can carry an excellent overall satisfaction score while one SKU is quietly generating quality complaints, low reorder intent, or disproportionate dissatisfaction. Store-wide averages hide product-level problems.

In short: Understand is the second pillar of the Post-Purchase Framework. It means shifting the unit of analysis from how customers feel about the brand to what customers are experiencing with the specific product they received - so one underperforming SKU never gets averaged away.

The commercial questions

Why the unit of analysis matters

Reviews, NPS, support tickets, returns data and purchase analytics each capture a fragment of the answer. On their own, none of them reconstructs the full picture of what happened, why it happened, and what it means for retention.

The principle

What to analyse

In repeat-purchase categories, a product-level issue that goes unnoticed for even one order cycle can quietly compound into a full cohort of customers who simply don't come back. Look for patterns across:

CategoryLook for
Product qualitySensory issues, packaging damage, breakage or leakage, batch variation.
UsageConfusing instructions, prep or dosing friction, unclear expectations.
ValueWhether the product feels worth the price and matches the promise.
Waste and abandonmentWhy a product was discarded, or why use stopped.

The most useful patterns often live in the interaction between variables - a strong unboxing score paired with a poor first use, or high overall satisfaction next to one SKU with unusually low repeat intent. Store-wide scores can't reveal that. Product-level analysis can. Best practice: analyse feedback at the SKU or product level first, then segment further - by region, format, variant, or batch - only where the data supports it.

The diagnostic

Four causes, one distinction

For every pattern you find, ask which of these it is. That distinction determines what happens next - and prevents teams from treating every problem as a marketing or support issue.

A product problem

Quality, formulation, packaging, or format issues that show up consistently for one SKU.

An expectation problem

The product performs as designed, but the customer expected something different.

An operational problem

Fulfilment, shipping, or handling issues introduced after the product left the warehouse.

A communication problem

Unclear instructions or usage information, not a defect in the product itself.

Measurement

High-signal KPIs for Understand

KPIWhat it tells you
Satisfaction by SKUWhere experience quality actually sits, product by product.
Issue rate by SKUWhich products are generating problems.
Repeat intent by SKUWhich products are quietly at retention risk.
Top issue drivers by productWhat's actually causing dissatisfaction.
Waste / discard reasonsWhy a product goes unused.
Time to first negative signalHow early a problem shows up.
Variance between products, variants or regionsWhere the real differences are hiding.

In practice

How TrueSignal operationalises Understand

TrueSignal turns post-purchase responses into product-level intelligence - surfacing which products create the strongest experiences, which generate the most issues, and where expectations aren't being met. The useful output isn't "customers are unhappy." It's "Product X is creating a specific experience problem that appears to be affecting repeat intent" - a sentence someone can actually act on.

Common questions

Understand, answered

What is the difference between product intelligence and product analytics?

Product analytics typically tracks behavioural events - clicks, purchases, page views. Product intelligence, as this pillar defines it, is declared feedback tied to a specific product: what a customer says about the exact SKU they received, not what they clicked.

How do I know if a problem is a product issue or a customer expectation issue?

Look at what's actually being reported. A product problem shows up as consistent sensory issues, damage, or batch variation for one SKU. An expectation problem shows up as the product performing as designed but not matching what the customer assumed they were buying. The distinction determines whether the fix is a product change or a listing/messaging change.

Do I need to segment feedback by region, batch, or subscription status?

Only where the data supports it. Analyse at the SKU or product level first, then segment further - by region, format, variant, or batch - once a pattern is visible. Segmenting before there's a clear pattern usually just fragments a small sample size.

Want more thinking like this?

Get new frameworks, research and practical insights from Agaya Cloud on post-purchase intelligence, customer experience and retention.

TrueSignal

Put Understand into practice

TrueSignal turns post-purchase responses into product-level intelligence - satisfaction, issues, and expectations, broken down by SKU.

Explore TrueSignal →