The order most founders do this in is backwards. Pick a name, buy the domain, print the packaging, file the application, and find out months later that it was never going to register.
The filing fee is gone. So is the name you put on everything.
Most refusals come down to two grounds, and you can assess both before you spend anything.
Ground one: it describes what you sell
Under Section 2(e)(1), a mark that merely describes the goods or services cannot register on the Principal Register. The Trademark Office is not being difficult. A descriptive term belongs to everyone selling that thing, and letting one company own it would stop competitors describing their own products.
Marks sit on a spectrum. Fanciful terms are invented, like Kodak, and are strongest. Arbitrary terms are real words unrelated to the product, like Apple for computers. Suggestive terms hint at a quality without describing it, and register. Descriptive terms state a characteristic and generally do not. Generic terms are the thing itself and never register.
The trap is that founders love descriptive names because they explain the business. Creative spelling does not help either. Changing an i to a y or dropping a letter does not make a descriptive term distinctive. Examiners look at how a term sounds and what it means, not how it is spelled.
The fix is usually to move one step up the spectrum, or to accept a disclaimer on the descriptive portion and build distinctiveness around the rest.
Ground two: it is too close to something already there
Under Section 2(d), a mark is refused if it is likely to cause confusion with an existing registration. Two things drive it: how similar the marks are, and how related the goods are.
Similarity is not identity. It covers appearance, sound, meaning and overall commercial impression. Marks that look different on paper can be refused because they sound alike said aloud, or mean the same thing in translation.
Relatedness is not identity either. You do not need to be in the same class. If consumers might reasonably assume the products come from the same source, that is enough.
This is why a quick search of exact matches gives false comfort. The conflicts that sink applications are usually near misses in adjacent categories.
Check before you spend
The USPTO's TESS database is public and free. Search variations, not just your exact string: phonetic equivalents, singular and plural, common misspellings, translations if your market includes other languages.
While you are there, look at the goods and services descriptions on anything close. A similar mark in a genuinely unrelated field may be fine. A similar mark in yours is a real problem.
What to do with a weak result
Finding out a name is risky is a good outcome, not a bad one, as long as it happens early.
Options in rough order of cost: change the mark, which is cheapest before you have printed anything. Narrow the goods and services description to avoid the conflict. File on the Supplemental Register, which is available for descriptive marks and can move to the Principal Register after five years of use. Or file anyway with an argument prepared, which is what a trademark attorney is for.
What you should not do is file hoping nobody notices. Examination is not a formality.
Why we built a tool for this
MCInvent scores a mark against these grounds before you file, searching a corpus of more than seven hundred fifty thousand marks for conflicts and flagging descriptiveness risk, then explaining what to change. It is deliberately prep-not-file: it does not submit anything for you.
It also is not a law firm, and it says so on every result. It is analysis of registrability, not legal advice. For anything contested, or anything you are about to build a company on, see a trademark attorney. The value of checking early is that you walk into that conversation knowing which of your shortlist is worth paying to defend.
