When Every SaaS Looks the Same: The Commoditization Problem

The spreadsheet had 41 rows. It was 11:40 on a Thursday night and I was calling it research: pasting competitor homepages into a column, sorting by headline.

Row 3: The AI-powered platform for modern teams. Row 17: The modern platform for AI-powered teams. Row 29: AI-powered workflows for modern teams.

I laughed. Then I stopped laughing, because row 34 was mine and the only thing separating it from row 3 was a comma.

My first thought was that forty companies had copied each other and I’d been slow to notice. That was wrong, and being wrong about it cost me two more quarters. Nobody steals your differentiation. A system removes it, one planning meeting at a time, and every person involved makes a decision they can defend in the room.

Why does every SaaS product sound the same?

Because they aren’t copying each other. They’re all copying the same third document, the category page that sits between them and every buyer.

Scott Brinker’s 2025 marketing technology landscape counted 15,384 solutions across 49 categories, up 9% from 14,106 the year before. That’s an average of 314 products per category. No buyer evaluates 314 products. No buyer evaluates 14. Zylo’s 2025 SaaS Management Index put the average company’s portfolio at 275 applications, a number that describes exhausted buyers rather than curious ones.

So a layer formed in between to compress 314 into a shortlist. Review-site grids, analyst quadrants, RFP templates, procurement checklists, “10 best alternatives to X” listicles. Call it the comparison layer. For most buyers it’s the only surface where they ever meet most products.

A comparison layer can’t represent everything. It represents what fits in a column: price tier, seat count, SOC 2, SSO, API, integrations, a matrix of checkmarks. Everything a product actually is, the ordering of a workflow, the thing it refuses to do, the customer it’s wrong for, has no cell to live in.

That constraint doesn’t stay at the buying stage. It runs backwards into every roadmap in the category.

Who actually decides which features you ship?

Not the founder, and not the users. The feature list on the comparison grid decides, through a loop that closes in about one sales cycle.

A rep loses a deal. The post-mortem field says missing capability: [X]. X came off the buyer’s checklist, which came from an RFP template, which was assembled from the same grid every other vendor is on. The rep posts it in Slack. Three more losses tagged X and it enters the roadmap with a revenue number attached. It ships. The grid gets a new checkmark. Competitors see the checkmark and run the same loop. Six months later the axes of the grid haven’t moved, and four products that were different in 2024 are the same product in 2026.

Every step is rational. Nobody in that chain is lazy. The rep is reporting a real loss, the PM is derisking real revenue, the founder is answering a real board question. The system takes 41 independent rational actors and produces one product. Your roadmap isn’t written by your users. It’s written by the checkbox column your losing deals keep pointing at.

You can see this in usage data. Pendo analyzed feature usage across 615 subscriptions and found that 80% of features in the average software product are rarely or never used, representing up to $29.5 billion in public cloud R&D spend. That was 2019, and nothing since has reversed it. Those features weren’t built for people who open the app. They were built for people reading a table.

Is feature parity actually a strategy?

No. Parity is a tax you pay to get listed, and the receipt is worthless the moment the deal closes.

Parity buys you entry to the evaluation. It doesn’t buy retention, and retention is where commoditization actually sends the bill. Benchmarkit’s 2025 SaaS Performance Metrics put median gross revenue retention at 88%, down from 90% over three years, and median net revenue retention at 101%, which means the median SaaS company now expands its existing base by approximately nothing. Median growth fell to 26%. CAC payback stretched 12.5% at the median since 2022.

Read those four numbers as one sentence. Acquisition got more expensive, the base leaks faster, and expansion stopped covering the difference. That’s the financial signature of a product the customer can’t distinguish from its replacement. Once every option in the category clears the same checklist, switching costs collapse to migration effort and renewal turns into a procurement exercise instead of a decision.

Parity spends your engineering capacity building a product nobody has a reason to stay with.

The two forces that pull a category together

Neither of them is competitor espionage.

The first is sociological. Paul DiMaggio and Walter Powell named it in 1983 in The Iron Cage Revisited, American Sociological Review 48(2). Organizations in a shared field become homogeneous, and the strongest driver is mimetic isomorphism: under uncertainty, organizations model themselves on others in the field that look legitimate. Not because imitation works. Because imitation is defensible. A PM who ships the competitor’s feature and fails is unlucky. A PM who ships something nobody in the category has and fails is reckless. The asymmetry lives in the career, not the market.

The second is geometric. Harold Hotelling described it in 1929 in “Stability in Competition”: two vendors on a beach, customers spread evenly along it, each vendor edging toward the middle to capture the other’s territory, until both of them are standing side by side at the center selling the same thing. Every SaaS category has a beach. The comparison layer draws it, and the middle is wherever the grid’s axes happen to cross. What converges on you isn’t your competitors. It’s the whole category sliding toward a coordinate system a third party published.

Isn’t this just what a mature market looks like?

This was the objection I got most often once I started saying any of it out loud. Sameness is the natural end state, it happens to everybody, and the winner is whoever executes distribution best. It’s a comfortable story, because it turns the outcome into weather instead of a choice.

The economics doesn’t support it. Hotelling’s convergence only holds when prices are fixed. In 1979, d’Aspremont, Gabszewicz and Thisse published “On Hotelling’s ‘Stability in Competition’” in Econometrica and showed the Principle of Minimum Differentiation is invalid. When sellers set their own prices, no equilibrium exists at the center, because standing next to your rival triggers a price war that ruins you both. Fix the model properly, with quadratic transport costs, and the equilibrium flips to maximum differentiation. The two firms walk to opposite ends of the beach.

So convergence isn’t the mature-market equilibrium. It’s what happens when firms behave as though price were fixed. Which is exactly what a published tier table on a comparison grid enforces: your price becomes a column, legible to every competitor, anchored to the category median before a prospect ever talks to you. The comparison layer disables the one mechanism that would push firms apart, and then the resulting sameness gets filed under maturity.

The distribution argument fails on its own evidence too. If brand and go-to-market were enough, gross retention would hold steady while acquisition costs rose. Instead GRR fell to 88% while CAC payback stretched. You can’t out-distribute a product the customer can swap out.

How do you differentiate when everyone has the same features?

Take a position the comparison layer has no column for. Anything a checkbox can hold, a competitor can copy in a sprint, and the next grid refresh erases the difference.

Things that don’t fit in a cell: who your product is deliberately wrong for, a default you refuse to make configurable, the order a workflow forces on you, what happens in the first ninety seconds, the feature you deleted. A grid has no notation for this product will make you do it this way. Which is why opinionated products survive category compression and configurable ones get bought on price.

Two things that cost nothing, and one that costs you a feature.

Stop instrumenting the grid and start instrumenting your own cohorts. The customers who renewed at 24 months did it for a reason that isn’t on anybody’s checklist. Go find the reason and make it louder.

Write the paragraph naming who shouldn’t buy this, and publish it. It disqualifies the buyer who was going to churn anyway, and it’s the one claim no competitor will copy, because copying it costs them deals.

Then delete a feature. That 80% number isn’t an indictment, it’s a menu. Every unused feature is surface area you’re maintaining, documenting and describing, and most of it exists to satisfy a table.

What the spreadsheet actually showed

I went back to those 41 rows a year later with a different question. Not who copied whom. I checked that, and the answer was nobody. I couldn’t trace a single headline back to another company on the list. Most of those teams had never read each other’s homepages.

They’d all read the same page. The same review-site category, the same analyst summary, the same RFP template, the same listicle. Forty-one companies, independently, each doing the reasonable thing, converging on a description none of them wrote and all of them signed.

The convergence had no author. That’s the part the guy at 11:40 on a Thursday couldn’t see, and it’s the part that matters, because it means no amount of trying harder inside the coordinate system gets you out of it. You don’t escape by writing a better row. You escape by building the thing the sheet has no column for and then sitting with a worse score for a while, which is much less fun than it sounds when you write it down as strategy.

I spent two quarters rewriting row 34. Nothing was ever wrong with row 34.