How the Thirteen du Pont Factors Actually Get Weighed

Roughly 4.6 million live trademark filings — registered marks plus pending applications — share the United States federal register, by GleanMark's count across its index of more than 14 million USPTO records. And more than half of the people who file into that crowd hear about it: of the 46,443 applications filed directly with the USPTO in January 2022 (a cohort old enough that nearly every file's examination has finished; applications routed through the Madrid international system are excluded), 25,444 — 54.8% — have drawn a refusal letter of some kind on the register.

When the refusal is about a collision with somebody else's name, the fight runs under Section 2(d) of the Trademark Act — the "likelihood of confusion" ground — and the analysis runs on the same thirteen questions every time. They are called the du Pont factors, after a 1973 case involving the DuPont chemical company, and since mid-July we have been publishing a factor-a-week breakdown of the list — the thirteenth and final installment lands Wednesday, October 7. This article is the piece that series was building toward, and it stands on its own: what nobody prints on the poster is that the thirteen factors were never a scoring rubric. Nobody adds up seven-out-of-thirteen and declares a winner. The real weighing follows three rules, each set down in a case you can read, and each with consequences a founder can act on.

The court that wrote the list refused to write a formula

The list comes from In re E.I. du Pont de Nemours & Co., decided in 1973 by the federal appeals court that then heard trademark cases. Three tribunals apply it to this day: the USPTO examining attorney who reviews your application; the Trademark Trial and Appeal Board, or TTAB — the USPTO's internal trademark court — which hears appeals from examiners and disputes between companies; and, above both, the Court of Appeals for the Federal Circuit, a real federal appeals court.

The du Pont court listed its thirteen "evidentiary elements" and then, in the same opinion, refused to rank them:

"The evidentiary elements are not listed above in order of merit. Each may from case to case play a dominant role."

And it refused to mechanize them:

"Each case must be decided on its own facts. There is no litmus rule which can provide a ready guide to all cases."

Twenty years later the Federal Circuit said the same thing more compactly in In re Shell Oil Co. (1993): the factors "may play more or less weighty roles in any particular determination." That is the entire official doctrine of weighing — no arithmetic, no ranking, no minimum count. What fills the vacuum is three working rules the courts have made explicit.

Rule one: two factors are the keystones

Board decisions state the first rule in a stock sentence: "In any likelihood of confusion analysis, two key considerations are the similarities between the marks and the similarities between the goods or services." The citation the Board attaches to it is a 1976 case, Federated Foods, Inc. v. Fort Howard Paper Co., for its holding that the fundamental inquiry under Section 2(d) "goes to the cumulative effect of differences in the essential characteristics of the goods and differences in the marks." The USPTO's own examination manual, the TMEP, says the same at § 1207.01: the similarity of the marks and the relatedness of the goods or services are the key factors, there is "no mechanical test," and each case is decided on its own facts.

In practice this means most examinations begin and end on factors one and two. Week two of our series watched it happen in a live file: MAMA MATCHA, packaged matcha in Class 30, was refused over MATCHA MAMA, a juice bar in Class 43 — the same two words in the opposite order, a product against a shop, and the refusal still went final in July 2026. The two keystone factors work as a sliding scale: the more similar the marks, the less related the goods have to be, and vice versa. "We're in different industries" loses to "related enough" far more often than founders expect. If you want the full walk through all thirteen factors at examination depth, our guide to the du Pont factors covers each one.

Rule two: a single factor can end the case

In 1991 the Federal Circuit decided Kellogg Co. v. Pack'em Enterprises, Inc. Kellogg, owner of FROOT LOOPS, opposed Pack'em's application for FROOTEE ICE (with an elephant design) for frozen flavored ice bars — an opposition being a formal objection to somebody's trademark application, decided inside the USPTO rather than in a courtroom. The Board ended the case without a trial, ruling for the applicant on exactly one factor: the marks were simply too dissimilar. It said it would rule the same way even if Kellogg proved at trial that FROOT LOOPS was famous, the goods closely related, and the buyers the same casual shoppers. Kellogg appealed, arguing you cannot decide a thirteen-factor test on one factor.

The Federal Circuit disagreed, in words that have carried ever since:

"We know of no reason why, in a particular case, a single duPont factor may not be dispositive."

The court grounded that in du Pont's own language — each factor "may from case to case play a dominant role." The founder-facing consequence: a strong enough showing on one factor does not just tilt the analysis, it can close it. That cuts both ways. Marks dissimilar enough can save you against a giant, as they saved FROOTEE ICE. And a collision hard enough on marks-plus-goods can sink you no matter how good the rest of your story is — which is what the next rule guarantees, because nobody will even reach the rest of your story.

Rule three: the tribunal skips the factors with nothing behind them

In 1997 the Federal Circuit decided In re Dixie Restaurants, Inc. Dixie had applied for THE DELTA CAFE (with a design) for restaurant services and been refused over DELTA, registered for hotel, motel, and restaurant services. On appeal, Dixie's argument was procedural: it asked the court to reverse "for failure to fully and properly consider the required 13 factors of DuPont." The court's answer was that there is no such requirement:

"not all of the DuPont factors are relevant or of similar weight in every case … Indeed, any one of the factors may control a particular case."

And it found "no error in the board's decision to focus on the DuPont factors it deemed dispositive." The du Pont opinion itself had built in this limit — the factors must be considered "when of record," meaning when there is actually evidence in the file about them. The Board says the same in its own decisions: "While we have considered each factor for which we have evidence, we focus our analysis on those factors we find to be relevant."

Read the three rules together and the machine comes into focus. Two factors carry most cases; any one factor can decide a case by itself; and the factors nobody put evidence behind are not weighed at all — they are skipped. Which raises the question that turns doctrine into strategy: where does the evidence behind each factor come from?

The founder's asymmetry: presumed against you, or pay to play

Thirteen factors' worth of reading these cases side by side surfaced a pattern the doctrine never states outright. The factors split into two families, and the split is about who supplies the evidence.

Two factors are presumed against you when the paperwork is silent. Factor three (trade channels) and factor four (buyer care) get filled in by default, and the default is the worst case. Narrowing your own application usually does not escape it: if the other side's registration has no limits, its goods are presumed to reach your channel and your buyers too. If the other side registered for "software" with no limits, the USPTO assumes its software reaches every channel and every customer, yours included. That is what got MAPPYMAIL refused over MAPMAIL in week three. MAPPYMAIL had narrowed its own description to single-letter mailings, excluded the bulk mailing MAPMAIL is known for, and filed evidence of how differently the two businesses work. None of it mattered, because the comparison runs on the paperwork, and MAPMAIL's registration has no limits. Claim your buyers are careful professionals and the answer is: the description decides, and the least careful buyer of the goods as described sets the bar. A $3 impulse purchase and a $300,000 procurement get the same treatment unless the description itself shows the difference; evidence about who actually buys from you cannot narrow it.

Six factors move only when someone builds the evidence. Factor seven (actual confusion) needs logged incidents — and its absence proves almost nothing, as sixteen quiet years of RED BULL tequila demonstrated in week seven. Factor eight (coexistence without confusion) needs proof the two brands' customers actually overlapped — Guild Mortgage's forty-plus years of quiet coexistence were weighed on court order and found worth exactly nothing, because nothing showed the same buyers ever saw both names. Factor nine (a family of marks) needs a used, advertised naming pattern — McDonald's proved one and kept MCSWEET off the register after 24 years of real use. Factor ten (agreements between the parties) needs a deal with machinery in it, not one page of mutual reassurance. Factor eleven (your right to exclude) needs a registration you kept alive — a lapsed one is, in the Board's words, "a new ball game." And factor thirteen, the catch-all, has one reliable winning shape, and it too turns out to be a registration you kept alive. More on that at the close.

The asymmetry is the practical takeaway of the whole series. The factors that could hurt you are on by default; the factors that could save you are off until you spend money and diligence turning them on — confusion logs, coexistence files, family-mark advertising, properly built agreements, renewals filed on time. Weighing is not a lottery. A factor nobody proves does not count for you.

The scoreboard: thirteen weeks, thirteen factors

One line per factor: what each week of the series ran — the factor-13 installment closes the series on October 7 — and the lesson that survived verification. Weeks one through four taught through live register examples and du Pont's own text rather than famous case names; the table reports what the posts actually said.

The register's last word

There is an epilogue the series kept tripping over, week after week, until it became impossible to treat as coincidence. An astonishing number of the marks in these landmark fights, winners included, later died because nobody filed the routine paperwork that keeps a registration alive.

The TIME TRAVELER registration that beat Bay State's consent agreement was cancelled in 2020 when its owner stopped filing maintenance paperwork. Both RALLY marks from the du Pont case itself are dead — one cancelled in 2021 after 62 years on the register, the other expired. Both combatants from the E.D.S. fight died the same way within six weeks of each other in early 2016. Both BLUE MOON marks that fought a 2003 Federal Circuit case over beer versus restaurant services are gone — including the very certificate Coors won there. The small registration that beat Shell Oil lapsed in 2017. The DAVEY registrations lapsed the same way twice, in 2006 and 2018 — the first lapse cost a Board appeal, and the register shows a Davey entity at the same Australian address refiling after each loss. And ANYWEAR BY JOSIE NATORI, the registration that nearly killed the ANYWEAR application in the factor-13 case, was itself cancelled in July 2024 — while the ANYWEAR registration it nearly blocked lives on, renewed.

This is not rare. In the twelve months ending August 31, 2026, more than 212,000 US registrations were cancelled for missed Section 8 filings — the sworn declaration of continued use that keeps a registration alive. Winning the argument and keeping the registration are different jobs, and the register only remembers the second one. The mechanics and the deadlines are in our Section 8 and 9 maintenance guide; GleanMark's monitoring, for its part, checks the register twice daily, so changes to a watched mark surface the same business day.

Which brings the series to its natural last word: In re Strategic Partners, Inc. (2012), the ANYWEAR case, application serial 77903451. The Board looked at ANYWEAR for footwear against ANYWEAR BY JOSIE NATORI for clothing and conceded the fight was lost on the merits: "We would conclude, under usual circumstances, that confusion is likely to occur among consumers in the marketplace." Then it threw out the refusal anyway — because the applicant already owned a registration for essentially the same mark on the same goods, that registration had coexisted with the blocking registration for over five years, and after five years the law closes the window for attacking a registration on confusion grounds. Factor thirteen, the Board wrote, "accommodates the need for flexibility in assessing each unique set of facts" — and here it "outweighs the others." Five years later a veterans' charity walked the same path with three and a half years of coexistence and lost; the window had not closed.

Sit with what that means. On the two factors that carry most cases, the Board said no — the marks and the goods pointed to confusion. The thirteenth factor said yes and won, and the fact it weighed was not brilliance in the brief or luck in the panel. It was a registration the applicant still owned, more than five years old and still sitting on the register — which was no sure thing: the applicant's other prior ANYWEARS registration had died in the middle of this very fight, cancelled in 2010 because a Section 8 declaration was never filed. The winning registration's own first maintenance deadline came up five months after the decision; that filing was made, and the mark was renewed in 2016. Its next renewal came due on August 29, 2026. As of early October the register shows no filing, only the USPTO's reminder, so the owner is inside the six-month grace period that ends in late February 2027. The du Pont list was never a formula, and this is the proof: the factor that outweighed all the others here was a fact the applicant had created for itself, by getting a registration and keeping it. After thirteen factors, that is the weighing lesson worth keeping — the factors that can save you count only if you built the record for them, and the cheapest piece of that record is the one most people forget.