Why your brand name is a bad keyword
Everyone's first keyword is their own brand name. It is the obvious move, it is what the empty state invites, and it is usually the weakest keyword in the set by a wide margin.
Not because monitoring your brand is a bad idea. Because of what a keyword actually is: a string match against text written by people who have never heard of your naming decisions.
This is the mechanics post. What to track strategically is a different question — this is about why the matching goes wrong and what you can do to it.
Failure one: your name is an ordinary word
Software naming has been trending toward short English words for fifteen years, and every one of them is a keyword that matches thousands of sentences with nothing to do with you.
Consider the tool literally named Mention. In a social listening context, that is close to the hardest keyword in the category — the word appears in every conversation about the thing the product does. That is not a criticism of the name; it is a good name. It is an illustration of how independent the naming problem is from the monitoring problem.
The general shape: the more ordinary the word, the more the match rate is driven by the English language rather than by your customers.
There is no clever query that fixes an ordinary word. There are only three partial mitigations:
- Negative terms.
acmeexcludingacme plant hire. Cheap, effective against the specific collisions you already know about, useless against the ones you have not seen yet. - Platform scoping. A repo name belongs on GitHub and Hacker News, not TikTok. Narrowing by platform is usually cheaper than narrowing by words, and people forget it exists.
- Ranking rather than filtering. A plant-hire firm sharing your name will match; the point is that it should never reach the top of the list.
We should say the honest thing here rather than the marketing thing: tracking Apple is still
hard. Tracking AcmeMail is not. If your product is named after a common word, expect to spend
real effort on this, and expect it never to be perfect.
Failure two: it only finds conversations you are already in
This one is structural and it is the reason a brand-only setup plateaus.
Your brand name finds threads that already mention you. By definition, those are people who already know you exist. That is worth having — it is your support surface, your reputation, and your early warning — but it is not new business, and it is a shrinking fraction of the conversations you would want to be in.
The threads where someone is describing your problem and asking what to use contain no brand name at all. A keyword set built out of brand names cannot surface them. Not because it is tuned badly — because the string is not there.
That is why the third category matters more than its usage suggests:
| Kind | Example | What it finds |
|---|---|---|
| Brand | AcmeMail, acmemail.com | Everything said about you |
| Competitor | alternative to Notchmail | People shopping, or comparing |
| Buyer intent | "shared inbox" + slow | People with your problem who have never heard of you |
Almost everybody sets up the first, about half set up the second, and the third is the one that gets skipped and is worth the most.
Failure three: when you are small, it finds nothing
A new product's brand keyword returns approximately zero results, and that is the correct answer. The mistake is concluding that monitoring is not useful yet.
At that stage the entire value is in the other two categories — the competitor and problem phrases, which are busy regardless of whether anyone has heard of you. A five-keyword set for a pre-launch product should probably contain one brand term and four others.
What matching genuinely cannot do
Worth being clear about the limit, because a lot of tooling in this category implies otherwise.
A keyword — even with negatives, even with boolean operators, even with platform scoping — is a test on the presence of strings. It cannot tell you that "is Acme down?" is an outage report and "the Acme is down the street" is not. It cannot distinguish someone recommending you from someone warning against you. It cannot see that a thread asking which tool to buy is worth more than a thread where your name appeared in a signature.
That gap is what ranking is for, and it is worth understanding what each approach does to it:
- Binary filtering — keep or drop against a description. Removes obvious junk. A hundred survivors still arrive as a hundred equal items.
- Scoring — a number per mention, with the reasoning attached. The top of the list is where attention goes; everything else stays searchable rather than unread.
On our plans, scoring is part of AI triage and runs on every plan, including Free. What the free plan limits is volume — 10 keywords and 500 relevant mentions a month — and filtered noise never counts toward it (how relevance is scored), so a collision-heavy brand name does not eat the allowance.
The bit nobody does: prune
Keyword sets rot. Competitors get acquired or renamed, campaigns end, a product line is discontinued, someone adds a term during a launch and never removes it. Six months later a meaningful share of the volume is coming from terms nobody would add today.
Two habits fix it:
Pause rather than delete when something ends. You keep the history, and you can resume without re-backfilling. Paused keywords still count against your plan limit, which is the useful forcing function — a full keyword list makes you decide.
Re-read the list monthly. It takes five minutes and it is the single highest-return maintenance task in this whole category.
A five-keyword set that works
If you are on a free plan anywhere, five is roughly the number you get (ours gives ten). Here is a set that beats "brand name plus four variations":
- Your brand name, with negatives for the collisions you already know about, scoped to the platforms your buyers actually use.
alternative to [your biggest competitor]— the highest-intent phrase in the language.[competitor] vs— the shortlist conversation.- A problem phrase in your buyer's words, not your marketing's. Whatever they would type at 2am, not the category term you use in a deck.
- A second problem phrase, deliberately broader than the fourth, to catch what the fourth is too specific for.
Only one of those five is your brand. That ratio is the actual advice in this post.
Our keywords documentation covers negatives, per-platform targeting and the backfill behaviour, and the free plan is 10 keywords across Reddit, Hacker News, Bluesky, GitHub and Google News with no card, which is enough to test whether that ratio holds for you.
Named products are used here as illustrations of a naming problem, not as criticism — we keep a dated feature and pricing dataset on all of them, including where they beat us, on our comparison pages.