You launched a sleep-tracking app called Whoop. Two months in, you get a sternly-worded letter from the lawyers at WHOOP Inc, the fitness-strap company. Their position: you're going to confuse customers. Your position: you make a sleep app, they make a wristband, nobody's confused. Who's right?
The honest answer is "nobody knows until a court decides," and the test the court will apply is called likelihood of confusion. It's the single most important concept in US trademark law, and almost nobody outside IP attorneys actually understands what it weighs.
The earlier post on how trademark classes work mentions likelihood-of-confusion briefly. This one breaks it apart so you can predict, with decent accuracy, whether your name is going to get you sued.
What the law actually says
US trademark infringement under the Lanham Act (15 U.S.C. § 1114) prohibits using a mark "likely to cause confusion, or to cause mistake, or to deceive" in the same commercial sphere as a senior registered mark. Likely is doing all the work here. Not "definitely will," not "could theoretically." A real, plausible chance that an ordinary consumer would confuse the two.
"Likelihood of confusion" doesn't have a bright-line rule. Federal courts decide it on a multi-factor balancing test. The specific factors come from each circuit's controlling precedent: a named appellate case that articulated the test, which then binds every subsequent likelihood-of-confusion decision in that circuit. Every US federal circuit has its own articulation, and most cite a few common cases by name. The three most-quoted in practice (and at the USPTO):
- Federal Circuit / USPTO TTAB: the DuPont test, 13 factors. From In re E.I. du Pont de Nemours & Co., 476 F.2d 1357 (CCPA 1973). Applied by the USPTO Trademark Trial and Appeal Board in every registration- and opposition-stage confusion decision.
- Second Circuit (NY): the Polaroid test, 8 factors. From Polaroid Corp. v. Polarad Electronics Corp., 287 F.2d 492 (2d Cir. 1961).
- Ninth Circuit (CA): the Sleekcraft test, 8 factors. From AMF, Inc. v. Sleekcraft Boats, 599 F.2d 341 (9th Cir. 1979).
The other circuits (1st, 3rd, 4th, 5th, 6th, 7th, 8th, 10th, 11th, DC) each have their own multi-factor test from a circuit-specific lead case. Factor lists overlap substantially across circuits; the same six or seven concepts show up in nearly every articulation. See the Wikipedia overview for the full circuit-by-circuit list. The six concepts that carry the most weight in practice (across every articulation):
The six factors that actually decide cases
1. Similarity of the marks
Visual (do they look similar?), phonetic (do they sound similar?), and conceptual (do they mean similar things?). All three matter independently. Whoop and WHOOP are identical phonetically and conceptually. That's the strongest possible match on this factor.
2. Similarity of the goods or services
This is where trademark classes come into play. Sleep-tracking app (broadly: software / health tech, class 9 / 42 / 44) versus fitness wristband (broadly: hardware / consumer electronics / health tech, class 9 / 10). Both sit in the health-tracking space; both touch class 9. A court would call these "related goods" even though the products are different.
Counter-example: Delta Air Lines (class 39, transport) vs Delta Faucet (class 11, plumbing) coexist because no consumer is buying an airline ticket and a sink in the same shopping mindset.
3. Channels of trade
Where do the products meet customers? If both are sold on Amazon health-tech filter pages, that's a problem. If WHOOP is in retail fitness stores and the sleep app is on the App Store, that's slightly less of a problem. But in 2026 every consumer software product touches the App Store, so channel-of-trade analysis has weakened over time.
4. Strength of the senior mark
Strong marks get broad protection. Weak marks get narrow protection. The strength scale:
- Fanciful: invented words (Kodak, Xerox). Maximum protection.
- Arbitrary: real words with no relation to the product (Apple for computers, Camel for cigarettes). Strong protection.
- Suggestive: hints at the product (Coppertone for sunscreen, Greyhound for buses). Moderate protection.
- Descriptive: directly describes the product (American Airlines). Weak. Needs to acquire distinctiveness ("secondary meaning") before any protection attaches.
- Generic: the word for the product category itself (escalator, aspirin). Both started as brands and lost protection. No protection.
WHOOP is arbitrary or suggestive depending on interpretation. Moderate-to-strong protection. By contrast, a mark like Better (as in the mortgage company Better.com) sits on the descriptive end of the scale, which limits how broadly it can block third-party uses of the same word.
5. Actual confusion evidence
If WHOOP can produce customer-support emails, returns, or survey data showing consumers actually confused your sleep app for their wristband, that's powerful direct evidence. Courts give heavy weight to actual confusion when it exists.
In practice, this factor often isn't visible until after launch. Most filings happen without any actual-confusion evidence, and the case is fought on the other factors.
6. Defendant's intent
Did the defendant know about the senior mark and adopt theirs anyway? Did they copy specifically to free-ride on the senior brand's reputation? Independent name selection and a clean prior search (USPTO check, like dibbed.dev does) shift intent in the defendant's favor. The opposite (visiting the senior's site, code comments showing prior awareness) makes intent grim.
How the test plays in practice
Mattel v. MCA Records (2002)
Aqua released the song "Barbie Girl" in 1997. Mattel sued for trademark infringement and dilution. MCA won. The Ninth Circuit ruled that the song was an artistic, expressive use protected by the First Amendment, and that no reasonable consumer would think Mattel had licensed or endorsed the song. The opinion famously ends with "the parties are advised to chill" (296 F.3d 894 (9th Cir. 2002)).
Louis Vuitton v. Haute Diggity Dog (2007)
A dog-toy maker sold a chew toy called "Chewy Vuiton" parodying Louis Vuitton handbags. LV sued for trademark infringement and dilution. The dog-toy maker won. The Fourth Circuit held that parody marks which clearly signal themselves as parodies (not affiliated with the senior brand) can defeat likelihood of confusion. Key reasoning: an ordinary consumer seeing Chewy Vuiton on a dog toy wouldn't think LV had entered the dog-toy market. 507 F.3d 252 (4th Cir. 2007)
The "Whoop" hypothetical
Back to the example. Likely outcome on the factors:
- Marks: identical (very bad for the defendant)
- Goods: related health-tech, partial class overlap (bad)
- Channels: both App Store and health-tech retail (bad)
- Strength of senior mark: moderate-to-strong (bad)
- Actual confusion: TBD post-launch (neutral until evidence emerges)
- Intent: the defendant didn't know about WHOOP when picking the name (good, provable via dated screenshots and search history)
Honest assessment: this case probably settles before discovery. The senior mark holder wins on enough factors that the cost-benefit of fighting it doesn't work for the defendant, even with a technical shot. Rename early. The $5–20k cost of renaming in year one is much smaller than the $50–200k of litigation, and it preserves the option to pivot.
How to use this when picking a name
The cheapest move is to never get into this situation. Before committing:
- Run the candidate through dibbed.dev. Live USPTO data; any active mark on the same wordmark is a red flag.
- For an active mark, look at its class. Software/tech (class 9, 42) overlap is the dangerous one. Plumbing fixtures (class 11) is not.
- Look at its goods description in TSDR. Narrow descriptions are easier to coexist with than broad ones.
- For anything ambiguous, pay $500-2000 for a clearance opinion from an IP attorney. A single email gets a quote.
The fundamental truth of US trademark law is that "available" is a much weaker statement than people think. A clean USPTO search is necessary but not sufficient. The likelihood-of-confusion analysis covers cases USPTO's examiners might let through.
When in doubt, pick a different name. Naming is the cheapest part of building a company. Relaunching after a forced rename is the most expensive.
Ready to check a name? Run it through dibbed.dev →