Reading switching costs: why people stay with software they complain about

The short answer: people stay because leaving costs more than staying, and you can estimate that cost before you build by reading practitioner content for five specific barriers — stored data, learned workflow, connected tools, team habit, and procurement. Twelve to twenty sources in a category is usually enough to see which of the five actually binds.

Founders find a category where everyone complains and read it as an open door. Sometimes it is. Often the complaining has been going on for six years, the complainers are still customers, and the reason has nothing to do with the quality of the alternatives.

The five barriers, and how they sound

Switching cost is not one thing, and treating it as a single vague number is how it gets ignored. It decomposes into five barriers that behave differently and that a challenger can address separately.

Most categories are dominated by one or two of these. Knowing which is a design brief: a market held in place by stored data needs an importer more than it needs a better interface, and a market held by team habit needs a way for a single person to get value without a rollout.

Partial escapes are the real buy signal

The highest-value observation in this pass is somebody already leaving in pieces. Exporting to a spreadsheet every Friday, running a second tool alongside the incumbent for one part of the job, maintaining a private workaround that duplicates a feature they already pay for — each is somebody paying the migration cost in instalments.

These people are not merely dissatisfied; they have already decided the incumbent is insufficient and have accepted operational pain rather than wait for it to improve. They are also unusually easy to reach, because the workaround is normally what they made the video about. That overlap with idea-finding is why tracking which SaaS tools people actually recommend pairs so well with this pass: recommendation patterns tell you what has momentum, and workaround patterns tell you where it is failing.

Discount the noise before you count it

Categories go through waves where criticising the incumbent is simply the popular position, and a corpus gathered during one of those waves will overstate real movement. The tell is complaints that are fluent but non-specific — they describe a mood rather than a step in somebody’s week.

Filtering that is the same discipline as spotting hype versus signal in creator content. A complaint that survives the filter names the moment it happens, how often, and what the person does instead. Anything that cannot answer those three does not belong in the count.

Designing against the barrier that binds

Once you know which barrier dominates, the first release changes shape. Against stored data, the importer is a headline feature and deserves real engineering rather than a CSV endpoint nobody tested. Against team habit, the product needs to be genuinely useful for one person, because a single user is the only realistic entry point.

Against procurement, the work is not product work at all — it is a security page, an invoice option and an annual plan, and no amount of interface polish substitutes. Recognising that early prevents months spent improving something that was never the obstacle.

This is also where switching-cost research and objection research converge. The barriers above are objections stated as facts rather than as worries, and the phrasing people use is directly reusable, which is the point of finding buyer objections in creator content.

Reading the incumbent honestly

One trap worth naming: a challenger tends to study the incumbent’s weaknesses and skip its strengths, which produces a product that wins the argument and loses the evaluation. The recordings are full of things people quietly rely on and never mention as features, because they only notice them when they are missing.

Cataloguing those is unglamorous and load-bearing. It is the same structured teardown as a competitor teardown built from public video, run with the opposite question: not what does this do badly, but what would somebody miss on day two of using something else.

When the honest answer is do not enter

Sometimes the read is that the barrier is insurmountable at your size: the data is deep, the integrations are load-bearing, procurement is locked, and the complaints are all about price. That is a genuinely useful outcome, arrived at in a week rather than in a year, and it belongs on the same list as the other early signals that disqualify a SaaS idea.

Switching happens at moments, not gradually

Even in a market with low barriers, people rarely move because they decided to one Tuesday. They move at a trigger: a new job, a team reorganisation, a renewal date, a price increase, a migration they were already forced into, or a failure severe enough to make the status quo indefensible.

Those triggers are visible in the corpus and they are worth cataloguing separately from the barriers, because they determine timing rather than feasibility. A market with modest barriers and no recurring trigger moves slowly regardless of how good the alternative is; one with high barriers and an annual forcing event moves in a predictable window each year.

Practically, that shapes where the effort goes. If the dominant trigger is a new person arriving in a role, the product needs to be findable by someone researching the job for the first time. If it is a renewal date, the work is a straightforward comparison page and an import that runs in an afternoon.

What lowering the barrier actually looks like

The temptation is to answer high switching costs with a discount. It almost never works, because money is the one barrier on the list that was never binding — nobody stays on a tool they hate to save forty dollars a month, and the ones who would are not customers you want.

What does work is removing a specific, named cost. An importer that reads the incumbent’s actual export file, tested against a real one. A mode where both systems run in parallel for a month so nobody has to commit. A first result produced from imported data within minutes, so the migration proves itself before it is finished. Each of those is a feature, and each maps to a row in the barrier table rather than to a general desire to seem easier than the competition.

What the pass costs

Twelve to twenty sources, read for barriers and partial escapes rather than for sentiment, gives you a defensible view of whether a market can move. As of August 2026 that fits the $19 a month plan with 25 videos and 2 projects; $59 covers 80 videos and 8 projects if you are comparing several categories, and $199 covers 250 videos, 20 projects and 3 seats. See the pricing page.

Closing thought

Dissatisfaction is the easiest thing in the world to find and the least reliable thing to build on. What matters is whether anyone has started paying to leave — and that, unlike a mood, shows up as behaviour you can watch.