Competitor validation mistakes that can derail your plans

Most competitor research fails quietly. The output still looks like research: a table, some columns, a paragraph about differentiation. The error was made earlier, in how the competitor set was chosen or how an artifact was read, and nothing in the table shows it.

Nine of those errors follow. Each comes with the symptom you can catch it by and the step that fixes it, because a warning without a step is only another paragraph.

What to take away

• The most expensive mistakes happen before you look at a single competitor: choosing the wrong set, or drawing the category to fit the answer you wanted.
• A feature table records what exists. It records nothing about why, and why is the only part you can use.
• Every correction here is the same move: replace an inference drawn from a website with an observation taken from a buyer.

1. Counting only the companies that look like you

How it shows up. Your list is three startups with similar websites. Nothing on it about the spreadsheet, the agency, the in-house person, or doing nothing at all.

The step. Ask ten people with the problem what they did about it last month. Write the answers down. That list is your competitor set. Products that appear on it stay. Products that do not appear come off, however much their homepage resembles yours. The pillar on reading competitors explains why substitutes usually win; this is how you find out which substitute you are actually up against.

2. Reading absence as opportunity

How it shows up. "Nobody does this" appears in your notes as a plus.

The step. Before you record a gap, spend an hour trying to find who tried. Old versions of their pages, a year of their changelog, forum threads about the feature, and the federal trademark register, where a mark filed for a product nobody remembers is a record of somebody's attempt. If someone built it and pulled it, you have found a hole they climbed out of, and you need the reason before you jump in. If nobody has ever tried, ask why a problem this visible has attracted no money.

3. Treating the pricing page as the price

How it shows up. Your comparison uses the numbers printed on their site.

The step. Ask three of their customers what they actually pay and on what terms. In anything sold to an organization the list price is an opening position. The gap between list and paid is the room the competitor has to respond to you, and you want to know how wide it is before you set your own number in pricing validation.

4. Confusing loud with large

How it shows up. You rank competitors by how often you notice them: advertising, conference stands, posts.

The step. Rank them instead by things that cost money to sustain. Headcount over time. Offices. Customer logos that have been on the site for years rather than months. The size of the support library. Marketing volume is a spend decision that can be reversed next quarter. Persistence cannot be faked cheaply, and it is rarer than it looks: the federal statistics on business births, deaths and survival are a reminder that establishments enter and leave constantly, so years of continuous operation is itself information.

5. Building the feature table first

How it shows up. A grid with check marks, thirty rows, and a highlighted column for you.

The step. Build the table backwards. For each feature, write what would have to be true about their customers for it to have been worth building. Delete every row where you cannot answer. What remains is the list of things you genuinely know about their market, and it is much shorter than the grid.

6. Interviewing yourself about their customers

How it shows up. The sentence "their customers are frustrated because" with no name attached to it.

The step. Find three people who use the competing product. Ask what almost stopped them buying, what they still do outside the tool, and what would make them move. Log the answers in their words. Your assumption goes in a separate column marked as yours. The rules for asking about what happened rather than what they think apply without exception.

7. Leaving the switching cost off the page

How it shows up. Your plan says "better product" and stops there.

The step. For the incumbent you are replacing, list what leaving costs: data to move, people to retrain, the remaining contract term, the standing of whoever chose it. Put a number or a name beside each line. Your advantage has to clear the total. If you have not written the total down you cannot know whether it does.

8. Letting one competitor name your category

How it shows up. You describe yourself as "like them, but".

The step. Write your offer without naming anyone. Then ask five buyers what they would compare it to. If they name a different category from the one you had in mind, the price expectation and the buyer come from their category, not yours, and the offer you test has to be written for that category.

9. Doing it once

How it shows up. A competitor document with a date on it from six months ago.

The step. Put three artifacts on a calendar: the pricing page, the job listings, the changelog. Check them monthly and write one line per change. Ten minutes a month. The pattern only appears over time, and a single reading cannot show direction.

The corrected version

The output of competitor work done properly is short. One paragraph naming the substitute you are actually replacing and what it costs to leave. One list of things you learned from their artifacts, each with the artifact beside it. One set of quotes from their customers. And one line stating what none of this can tell you, which is whether anyone will buy from you. That last line is the handoff to a test.

Common questions

Should I sign up for their product?

Yes, under your own name, on a real trial, doing a real task. Log where you got stuck and what you had to read. Do not misrepresent who you are to get a quote or a demo; you learn less from a conversation built on a false premise, and the reputational cost is real in a small market.

What if a competitor has already done exactly my idea?

Then you have evidence that somebody pays for this class of thing, which is more than most ideas have. Find out what their customers still do by hand. The remaining workaround is where the next offer lives.

How many competitors should I study?

Three, thoroughly, beats twelve at a glance. Choose the one buyers name most often, the one they compare on price, and the substitute that is not a product at all.

Is there a line I should not cross?

Public artifacts, their customers who agree to talk, and trials in your own name are all fair. Pretending to be a buyer, using someone's confidential documents, or recruiting their staff to extract information are not, and none of it tells you anything a refusal from a real buyer would not tell you better.