Feature comparison matrix: AG1 vs Grüns

What a feature comparison matrix is

A feature comparison matrix is a table that sets two or more competing products side by side. Each row is a feature or offer term a buyer weighs, each column is a product or brand, and each cell says whether that product has the feature or how much of it.

How it works: rows, columns, cells

Rows are the buyer's questions

Each row should be something a buyer checks before paying. Can I try it cheaply. What does it cost to start. Can I send it back. Who else says it works. A row that no buyer would ever ask about is decoration, however proud the product team is of it.

Columns are the options the buyer actually has

Put your own product first, then the two or three rivals a buyer really considers. Ten columns is a market map, not a matrix, and nobody reads it.

Cells are yes/no or graded

A yes/no cell says whether the feature exists. A graded cell says how much: 30 days against 100 days, 4.3 stars on 2,803 reviews against 4.6 on 2,387. Yes/no cells hide exactly the differences buyers choose on. Two brands can both have a return policy and still offer very different ones. Use yes/no only where the feature is truly binary, and a number wherever one exists.

The parity trap

The classic matrix is a parity list: forty rows of ticks, drawn by the team that sells the product, where every row happens to be one they win. It persuades nobody who has seen the rival's own version, which has forty different rows.

It also teaches the team the wrong lesson, because a column full of ticks feels like a strategy. A row where both columns say yes tells you nothing about why a buyer picks one. The rows that matter are the few where the cells differ, and the few where you cannot fill them in.

Worked example: AG1 vs Grüns

AG1 and Grüns both sell greens supplements direct to consumers. We measured each brand on the same public sources; the Loop against Hears benchmark shows what that looks like on a live page. The matrix below takes only the offer features those sources measure, and puts in a cell only where there is a measured value. Where there is none, it says so.

What this matrix says

The loudest difference is not a feature at all. It is where the discount lives. 72% of Grüns' live ads carry an offer, against 4% of AG1's. 31% of AG1's live ads are fronted by a named creator, against none of Grüns'. One brand sells on the deal, the other on a face.

The review rows look like a Grüns win until you read the second one. Across all reviews, Grüns has 4.6 on Trustpilot and AG1 4.3. Across the newest reviews only, both sit at about 3.4. The gap is history, not the current experience.

The popup row shows why yes/no loses information. The measure has three states: none seen, a popup tool installed but no popup shown, and a popup seen. A tick box would have written "no" for both brands and hidden that Grüns has the machinery in place.

The first row is honest in a different way. "Ways in without talking to sales" counts free tiers, free trials and no-card starts. It was designed for software. Both brands score 0, and that zero does not mean neither sells a subscription. Subscription has its own row, and there nothing was measured for either brand.

Then there is AG1's column. Range, prices, return window, delivery promise and Amazon listings were not measured for AG1 on 28 September. A sales deck would fill those cells from memory. This matrix leaves them empty, because a guessed cell does more harm than a blank one: it gets quoted.

Three matrices from other categories

These are hypothetical, to show how the rows change with the category. The companies are made up and so are the values.

A subscription coffee brand (hypothetical)

Rows: price per bag at the entry plan, whether you can skip a delivery online, how many roasts are on offer, the delivery promise in days, and whether the first bag is discounted. Skip online is the only yes/no row, and only because you either can or you have to email.

A project management tool (hypothetical)

Rows: a free tier and its seat limit, trial length in days, the price per seat at the second plan, the number of listed integrations, and whether single sign-on sits on the top plan only. Here "ways in without talking to sales" is the row that matters most, because a demo call is where half the buyers leave.

A mattress brand (hypothetical)

Rows: trial nights, warranty years, the entry price for a queen, how many firmness options there are, and whether returns are collected or have to be shipped back. Trial nights and warranty years look alike, but buyers read them differently. One is a promise about the first month, the other about the tenth year.

How to make one

1. List the rivals a buyer really compares you with. Two or three, not ten.
2. Write the rows as buyer questions, before you look at anyone's answers. That keeps the rows from bending towards the ones you win.
3. Fill every cell from a source you can name and date: their store, their returns policy, their review page, their ads.
4. Use a number wherever one exists. Keep yes/no for things that are really binary.
5. Write "not measured" where you have nothing. Do not guess, and do not leave a blank that reads as "no".
6. Delete the rows where every column is the same, or move them below a line. Look at what is left.
7. Date the whole table and read it again in a month. Offers move quietly.

What it cannot tell you

A matrix counts what exists. It does not say how much a buyer cares about each row, so an empty cell in a row nobody weighs looks as bad as one in a row everyone weighs. A strategy canvas is the better tool for how hard each brand leans on each factor.

It cannot see what sits behind a login or an inbox. A welcome discount, a subscriber price or a retention offer can change the real price by more than any row on the public page.

And a cell is a snapshot. Grüns' 72% discount share is what its ads carried as the report stood on 28 September 2026. A campaign ending next week would move it, which is why the matrix is worth rebuilding rather than framing.

FAQ

What is the difference between a feature comparison matrix and a competitive matrix?

Often nothing. "Competitive matrix" is sometimes used for a broader grid that also covers price, positioning and channels. The shape is the same: options in columns, criteria in rows.

How many features should a comparison matrix have?

As many as a buyer checks before paying, usually eight to fifteen. More than that and the rows that decide the sale drown in rows that do not.

Should I use checkmarks or numbers?

Numbers wherever a number exists. A checkmark for "return policy" treats 30 days and 100 days as the same thing, and buyers do not.

Where do I get the data for a competitor's column?

From their public pages: store, returns policy, review sites, ad library and inbox. A battlecard uses the same sources to turn the differences into talking points.