How to monitor competitors without guessing what they are thinking

The five pages that carry signal

Almost everything a company publishes is noise. These five pages are where a decision leaves a visible mark, and each answers a different question.

1. The pricing page. The highest-signal page they have, because a price is a decision somebody had to defend internally before it shipped. A change tells you what they now believe the product is worth, and the shape of the tiers tells you who they now think the buyer is: seat minimums, usage limits, a new tier that only says contact us. It does not tell you what customers actually pay. Discounts, grandfathered plans and negotiated contracts all sit behind that page. A new enterprise tier says they want larger deals. It does not say they are winning them.

2. The homepage, or wherever the positioning lives. The headline is the one sentence they chose to lead with, so a rewrite means they changed who they are talking to, or what they think the reason to buy is. Watch which use case moved to the top and which customer logos appeared or disappeared. It does not tell you whether any of it is working. Marketing pages get rewritten for real strategy and for a new marketing lead wanting a visible win, and from outside those look the same.

3. The changelog or releases page. The most reliable of the five, because shipping costs money and nobody ships by accident. Read it for rate and direction: which parts of the product get attention, how often they release, whether they are going deeper for the customers they have or wider for ones they do not. It does not tell you what is coming next, whether a feature works, or whether anyone uses it. Remember it is also a marketing surface. Small releases get written up large, awkward ones get left out, and a quiet changelog is not proof of a quiet quarter.

4. The careers page. The only one of the five that points forward, because a role is a plan funded ahead of the work. It gets its own section below.

5. Review sites and forums. The one page on this list the company does not write, and the place customers say what it will not. It tells you what breaks and what people wish existed. It does not tell you proportion, because reviews come from the two extremes and from whoever happened to be asked at renewal.

How often to check each one

Match the check to how fast the page can actually change. Checking faster than that produces a stream of nothing, and a stream of nothing is how you teach yourself to ignore the one week that mattered.

• Changelog: weekly. The fastest moving page and the cheapest to read. Most still publish an RSS feed, so five competitors land in one reader and the whole job takes a minute.
• Pricing: monthly. Price moves are rare and large. Daily checking buys you nothing for eleven months, and you could not respond inside a week anyway, because changing your own price is a decision with customers attached to it.
• Careers: monthly. Hiring plans move on quarters. The pattern across three months is worth more than any single posting.
• Positioning: monthly. A real repositioning is loud and you will not miss it. The monthly pass is for the quiet edits: a changed subhead, a dropped logo, a use case that climbed the page.
• Review sites: monthly, and again before any decision that leans on them.

Then write it down, because the record is the real output here. One line per change: the date, the page, what changed, and a link. Save a copy of the page as well, with the Wayback Machine or a saved PDF, because companies edit pages without notice and the version you are comparing against quietly stops existing.

One pricing page is a snapshot. Six months of dated pricing pages is a story, and it is a story nobody can buy, which is the argument for starting now rather than when you need it.

Some things change constantly and mean nothing: a rotating testimonial, a footer year, reordered navigation, a new analytics script. Decide once that these are not news, so you are not deciding it again every Monday.

A careers page is a roadmap somebody has already funded

A job posting is the earliest public evidence of a plan. Someone wrote the description, got the budget approved, and expects the work to exist in six months. No other page gives you that much lead time.

Read it in this order.

1. Count the openings by team. Three backend engineers and no sales roles is a company building. Four account executives and a sales engineer is a company that has something that sells and is pushing on distribution instead. The ratio tells you more than any single role.
2. Look for the first hire in a function. A fifth engineer means more of what they already do. A first compliance hire, a first partnerships lead, a first support person in another timezone: each of those is a door being opened, and the first is worth ten of the fifth.
3. Read the location. A role in a country they do not sell in yet is the loudest thing on the page.
4. Read the description, not the title. Titles are generic. Descriptions are written by the hiring manager and leak the stack, the customer segment, and sometimes the actual project. The requirements list is where a company tells you what it is about to work on.
5. Note the seniority. Hiring the head of a function means the function is now big enough to need an owner. Hiring juniors into it means it is settled and needs hands.

What a careers page cannot give you is timing or certainty. Postings sit live for months, get reposted, and get canceled when a quarter goes badly. Treat one as an intention with a budget attached: more than a rumor, less than a fact.

Count the complaints, do not score them

The tempting move with review sites is to reduce a pile of reviews to one number: sentiment 6.8, or trending negative. Do not. Nobody can act on it, nobody can check it, and it buries the specific complaint that was the only useful thing in there.

Count instead. Fix the sample first, say the last fifty reviews or everything from the last quarter, then tally by theme and keep a link to every review you counted.

The result reads like this: eleven of the last fifty reviews say the import fails on large files, and here are the eleven. That is something you can build against, answer on a comparison page, or raise in a sales conversation. Trending negative is something you can only feel.

Three habits keep a count honest.

• State the denominator. Eleven of fifty and eleven of five hundred are different findings.
• State the window. A complaint from two years ago may describe a bug that was fixed eighteen months ago.
• Never let it stand in for customers. People write reviews when they are angry, when they are delighted, or when a vendor asked them at renewal. It is a count of reviews, and it should say so.

Start with the review directories, G2 and Capterra, then find the forum or subreddit where your customers actually talk, which is less organized and usually more honest.

The best thing in there is not a complaint at all. It is a thread where several people describe the thing you are building and cannot find. That is a list of people to talk to, in public, with the problem already stated in their own words, which is the raw material market research for startups sends you looking for.

Write down what you expect before you look

This is the habit that separates monitoring from browsing, and it costs a minute.

Before you open the page, write one line: what you expect to see, and what would surprise you. I expect the entry tier to still be twenty nine dollars and the free plan to still exist. Then look.

Two things follow. The first is that you can no longer be quietly wrong. Once you see a change, your mind builds a story in which it was obvious all along, and that story teaches you nothing about the market and quite a lot about your own confidence. A written prediction is the cheapest way to find out whether your model of a competitor is any good.

The second is that you start noticing things. Without a prediction you skim a pricing page and see a pricing page. With one, the missing free plan jumps out, because you were looking for it.

When you are wrong, write one line on why. After three months you will know which competitor you understand and which one keeps surprising you, and the second is where your attention belongs.

Then decide the action in advance. If they go below our entry price, we do this. If they ship the thing we are building, we do this. A trigger written before the event turns a page change into a ten minute decision. Written after, the same change becomes an afternoon of speculation and a roadmap nobody planned.

What monitoring cannot tell you

You can see that a page changed. You cannot see why, and the gap between those two is where founders lose weeks.

A price rise is confidence, or a revenue hole to fill before a raise, or a new finance lead applying a rule from their last company. A shipped feature is a considered bet, or the thing one engineer wanted to build, or a promise made to a single large customer. A new job posting is growth, or a backfill for someone who quit on bad terms. From outside, each of those pairs looks identical, and the story you settle on says more about your mood that week than about them.

So keep two columns and never let them merge. What the page said, dated and linked. What you think it means, marked as a guess. It is the same discipline as separating a sourced claim from an assumed one in your research, and it fails the same way when you skip it.

The list of things this method cannot reach is longer than the list it covers: revenue, retention, churn, whether a launch worked, what is on the roadmap, what was said in the room, whether they have heard of you at all. No public page carries any of it.

Which leaves one rule that keeps the whole exercise useful. Act on a competitor's move only when it changes a fact your own plan depends on. A price your customer can now get cheaper elsewhere is that kind of fact. A competitor shipping a feature you had planned usually is not, unless that feature was your entire reason to exist. Most changes should end in the record with the word noted and no further consequence. Reacting to a move you cannot explain hands your plan to a company you cannot see, and they may be getting it wrong.

If you would rather not keep five calendar reminders, competitor watch runs this method for you: up to five companies picked from the ones your research already found, their pricing, positioning, changelog and careers pages checked on the cadence you choose, and every change carrying the dated page it was read on so you can go and disagree. Complaints come back as counts with their sources, never as a score. A quiet period runs nothing, costs nothing, and says so, which is what makes the weeks that do report worth opening. And it will not tell you a motive either, for the same reason you should not.