What Your Customers Say vs. What They Mean: A Practical Framework

A customer writes: "I love the product but the bottle is terrible."

Read literally, that's one sentence with two opinions in it. Read for what it actually means, it's something more specific and more useful: the product itself is fine, but the packaging is undermining an otherwise good experience. Those are two different findings, and only one of them tells you what to actually go fix.

Most feedback tools stop at the first read. A star rating, a sentiment label, a word count of how many times "packaging" showed up this month - that's real information, but it doesn't answer the question a brand actually has, which isn't "what did people say" but "what should we do." Getting from one to the other takes one more step, and it's worth doing deliberately rather than leaving it to whoever happens to skim the review.

The four questions worth asking

Take any piece of feedback and run it through four questions in order:

What did they say? The literal words. "The bottle is terrible."

What do they mean? The interpretation. The packaging is creating a poor experience despite real satisfaction with the product itself - a different finding than "the product is bad."

Why does it matter? The consequence. If this is a one-off, it's not worth much attention. If it's showing up across many customers, it's a real signal that the packaging - not the formula, not the product design - is the thing costing you customers.

What should we do? The action. Depending on scale: nothing yet (watch for more), flag it to whoever owns packaging, or escalate if it's compounding with returns or negative ratings.

Skipping straight from "what did they say" to "what should we do" is how a lot of feedback gets misread. A team that reads "the bottle is terrible" as "the product needs work" ends up reformulating something that didn't need reformulating, while the actual problem - the cap that won't seal, the pump that clogs - goes untouched.

Sentiment alone isn't the answer either

It's tempting to think a sentiment score solves this: negative sentiment means a problem, positive means it's fine. But sentiment is one layer, not the whole picture. "I love the product but the bottle is terrible" might score as mixed or even neutral, which tells you almost nothing about what to actually do next. The theme extraction underneath it - this comment is about packaging, and it's a complaint, sitting next to praise for the product itself - is what turns a sentiment score into something actionable.

This is also why a single piece of feedback, read in isolation, can mislead. One customer calling something "fine" might mean mildly satisfied, mildly disappointed, or genuinely indifferent - there's no way to tell from one data point. The same word from fifty customers, clustered right after a formulation change, is a different kind of information entirely. Meaning often lives in the pattern, not the individual sentence.

What this looks like in practice

Take a real shape of feedback a D2C brand might actually receive:

"Great scent, but the pump stopped working after two weeks."

Say: positive on scent, negative on the pump mechanism.

Mean: the formula is working as intended - this isn't a reformulation problem - but there's a hardware/packaging defect specific to the dispensing mechanism.

Matters: if this is isolated, it's a manufacturing outlier. If it's showing up across a batch, it's a supplier or QC issue worth investigating before it becomes a wave of returns.

Do: check whether this is one customer or a cluster. If it's a cluster, that's a call to the packaging supplier, not a product reformulation meeting.

None of this requires reading fifty reviews by hand to notice - it requires tagging each piece of feedback consistently enough (what's the sentiment, what's the theme, is this a repeat pattern or a one-off) that the pattern can surface itself instead of depending on someone's memory of "didn't someone mention this before?"

Why this is worth building into your process, not just your head

The four-question framework works fine as a mental exercise on one piece of feedback. It becomes genuinely valuable once it's applied consistently across everything coming in - because that's when "why does it matter" stops being a guess and starts being a real count: how many people, over what period, saying essentially the same thing. That's the difference between reacting to whichever complaint was most recent or most annoying, and actually knowing what's worth your team's attention this week.

That's the whole idea behind how Sensia reads feedback: every submission gets tagged for sentiment, theme, and issues automatically, so "what does this mean" and "how often is this happening" are already answered by the time someone on your team opens it - instead of being work someone has to redo, comment by comment, every single week.