Checklist · Intellectual Property

Classifying Goods and Services for a Trademark Filing: A Practical Checklist

Type: Checklist
Jurisdiction: National
Date: 2024-09-07
Topic: Intellectual Property

A federal trademark protects only the goods and services an application actually claims, and the USPTO sorts all of commerce into the 45 standardized classes of the Nice Classification. This checklist walks a filer through five disciplined phases: inventorying every offering, locating the correct class and pre-approved wording in the USPTO ID Manual, drafting identifications that are definite and properly scoped, deciding the class count and budgeting the 2025 per-class fees and surcharges, and clearing the coordinated classes an examiner will cross-search. It explains why classification is "strictly ministerial" yet decisive for cost, examination, and the completeness of protection, and it dissects the one-way door that lets you narrow but never broaden, the Class 9 versus Class 42 software trap, and the fraud exposure created by claiming goods you do not sell. Worked hypotheticals, primary authority, and cross-references accompany each phase. Confirm current fees and the live ID Manual before filing.

Classifying Goods and Services for a Trademark Filing: A Practical Checklist

Intellectual Property > Trademark | Published 7 September 2024 - Updated 29 January 2025 | UpLaw - uplaw.us

A trademark registration is a fence, and classification decides where the posts go. This checklist takes a filer from a blank inventory to a properly scoped, correctly classed, fee-optimized identification of goods and services — and flags the traps that quietly cost applicants money, priority, and sometimes the whole registration.

Keywords: trademark classification, Nice Classification, USPTO ID Manual, identification of goods and services, international class, multi-class application, per-class fee, 2025 USPTO fees, free-form identification surcharge, coordinated classes, Class 9 software, Class 42 SaaS, specimen per class, bona fide intent to use, 37 CFR 2.85, no-broadening rule

Why a "ministerial" step decides so much

Trademark law draws a sharp line that trips up almost every first-time filer. Your rights flow from the mark you use and the goods or services you use it on. Classification — the act of sorting those goods and services into numbered baskets — does not, by itself, create or enlarge any of those rights. The Trademark Manual of Examining Procedure says so in as many words: classification is "strictly ministerial" and exists "for the [USPTO's] administrative convenience." See TMEP § 1401.03(b). Put the same shirt in Class 25 or mis-file it in Class 24 and your legal claim to the shirt is unchanged; you have only annoyed an examiner.

And yet classification is one of the most consequential decisions in the entire filing. It is ministerial in theory and dispositive in practice, for three reasons that compound one another.

First, fees scale with classes. The USPTO charges per class, at filing and again at every maintenance milestone. A brand that belongs in four classes pays roughly four times what a single-class brand pays — forever. Get the class count wrong on the low side and you leave product lines unprotected; get it wrong on the high side and you pay for coverage you cannot support and may later have to defend.

Second, classification shapes the examiner's conflict search. When the examining attorney searches for confusingly similar prior marks under Section 2(d) of the Lanham Act, 15 U.S.C. § 1052(d), the identification and its classes tell the examiner where to look — and, just as importantly, where a later third party will look when deciding whether to oppose you. The classes you claim advertise your competitive lane to the world.

Third, the identification defines the boundary of what you own. The registration certificate protects the mark "for" the listed goods and services and nothing else. Draft the identification too narrowly and a competitor can operate one product category over with impunity; draft it inaccurately and you invite a non-use cancellation or, worse, a fraud claim.

So treat this as a load-bearing task, not a form field. The good news is that it is a disciplined task: the standards are written down, the wording is pre-vetted in a free government database, and the mistakes are predictable. This checklist imposes that discipline across five phases, with the reasoning ("WHY") and the failure modes ("TRAPS") attached to each. For the panoramic tour of what lives in each of the 45 baskets, read it alongside The 45 Nice Classes: How Trademark Classification Shapes Cost, Scope, and Conflict; for where this step sits in the larger filing, see Preparing a USPTO Trademark Application: A Practical Checklist.

The architecture in one paragraph

The United States sorts goods and services under the Nice Classification, the international system created by the Nice Agreement Concerning the International Classification of Goods and Services (1957) and administered by the World Intellectual Property Organization. Nice divides all of commerce into 45 classes — Classes 1 through 34 for goods, Classes 35 through 45 for services — and WIPO republishes the list, currently the 12th edition, with revisions each year. By regulation, the USPTO adopts whatever version of Nice is in force. See 37 C.F.R. § 2.85(a). Two rules of the road follow immediately and matter more than any class number: an identification must be definite enough to be understood and classified (37 C.F.R. § 2.32(a)(6); TMEP § 1402.01), and after you file you may narrow but never broaden it (TMEP § 1402.06). Everything below is, in one way or another, an application of those two rules.

How to use this checklist

The five phases run in order, and the order is not arbitrary. You inventory before you classify because you cannot count classes for offerings you have not yet named. You find approved wording before you draft because the USPTO's database both tells you the right class and hands you examiner-tested language. You draft before you price because the class count that drives the fee is an output of the identification, not an input. You clear coordinated classes last because you can only search around a lane once you know which lane you are in.

Work each phase to completion before starting the next. Check the box only when the underlying judgment is actually made, not when it is merely noted. Where a phase turns on a fee, a class assignment, or a Manual entry, verify it against the live source the day you file — the USPTO ID Manual refreshes weekly, and fees change by rulemaking.

Phase 1 - Inventory every offering

[ ] List every product the brand sells or will sell (physical goods), in ordinary language.
[ ] List every service the brand performs for others — the "for others" qualifier is doing real work.
[ ] Treat software as its own line item and record its delivery mode: a downloadable app or program is a good; hosted, non-downloadable software accessed over the internet is a service.
[ ] Capture any retail or e-commerce activity: running a store, even an online one, is a service distinct from the goods on the shelves.
[ ] Note content, media, and events the brand produces (publications, videos, classes, live experiences) — these frequently hide extra classes.
[ ] For each item, mark whether it is in current use in commerce (supporting a Section 1(a) filing) or covered only by a bona fide intent to use (Section 1(b), 15 U.S.C. § 1051(b)).
[ ] Flag any item that is aspirational — on the roadmap but not yet backed by a genuine, documented plan to launch.

WHY. The single most common and most expensive classification error is under-classifying: the founder who is sure the business is "one thing" almost always occupies three or four classes. A modern brand rarely lives in a single basket. The physical product sits in a goods class; the companion app sits in Class 9; the website that sells everything is a Class 35 retail service; the customer-support hotline or subscription-management portal may be its own service. You cannot price or protect what you have not first written down, so the inventory must be exhaustive before a single class number is assigned.

TRAPS.

The "for others" trap in services. A service class covers activities performed for the benefit of others. Cooking for your own restaurant guests is a restaurant service (Class 43); cooking for your own convenience is nothing the USPTO recognizes. A company that hosts its own website is not, on that basis, rendering "website hosting services" to anyone. Ask, for every candidate service, "who is the beneficiary, and are they a third party?" If the answer is "us," it is probably not a service.
The ancillary-service mirage. Ordinary support that comes bundled with a product — a warranty, a help line, free shipping — is usually not a separately registrable service; it is incidental to selling the goods. See TMEP § 1301.01(a). Do not manufacture a Class 35 or Class 39 claim out of activities that merely accompany the sale.
The roadmap trap. An intent-to-use basis requires a bona fide intention to use the mark on each listed item, judged by objective circumstances. Loading the inventory with everything the pitch deck promises is how applicants end up with claims they cannot support at the specimen stage — or in a later audit. If a line is genuinely speculative, park it; you can file a continuation or a new application when the plan firms up.

Worked example: Thornbury Provisions

Imagine a hypothetical startup, Thornbury Provisions, that began as a single small-batch hot sauce and is expanding. Its founder insists the brand is "a hot-sauce company — one class." A proper inventory tells a different story:

The hot sauce and a line of spice blends — condiments and seasonings, which fall in Class 30.
Branded glass bottles and an insulated tumbler sold as merchandise — drinkware and household containers, Class 21.
Branded aprons, tote bags, and t-shirts sold on the site — Class 25 for the apron and tee (clothing), Class 18 for the tote (leather/canvas goods), depending on materials.
The online shop that sells all of the above — retail store services featuring food and merchandise, Class 35.
A weekend pop-up tasting room the founder plans to open next year — restaurant/food-and-drink services, Class 43, on an intent-to-use basis.

A "one-class" instinct has become as many as five. Thornbury need not file in all five — that is the Phase 4 judgment — but it cannot make that judgment intelligently until the whole board is on the table. This is also the moment to decide the mark itself is worth the investment; if the brand name is weak or unavailable, classification is premature. For that upstream question, see Choosing a Mark the Law Will Protect: A Practical Checklist and What Can Be a Trademark: The Subject Matter of Trademark Law.

Phase 2 - Find the class and approved wording in the ID Manual

[ ] Search each inventory item in the USPTO's Trademark ID Manual (the "Acceptable Identification of Goods and Services Manual") to find its class and a pre-approved description.
[ ] Record the true class count that falls out of those searches — this is the number that drives cost.
[ ] Flag classification surprises, where an item lands somewhere non-obvious (a water bottle in Class 21, not Class 32; a downloadable app in Class 9 while the same product as a hosted service is in Class 42).
[ ] Adopt ID Manual wording verbatim wherever an entry fits — this avoids the free-form surcharge and speeds examination.
[ ] Where nothing in the Manual fits a genuinely novel offering, draft custom language that mirrors the Manual's structure and specificity rather than inventing a new format.
[ ] Cross-check against real registrations for comparable products in the USPTO's Trademark Search database to see what language actually issued.

WHY. The ID Manual is the highest-leverage free tool in the entire process. It answers, in one search, three questions that otherwise take an experienced practitioner to resolve: What class is this in? What language will the examiner accept? And is there a trap I should see before I pay? Because the entries are pre-approved, using them removes the most common ground for an office action — an indefinite identification — and, since the 2025 fee rule, it also removes a surcharge. The Manual is a controlled vocabulary; treat it as one.

TRAPS.

The weekly refresh. The Manual is revised every week. An entry you relied on in a template last quarter may have been deleted, split, or narrowed. Pull it live the day you draft.
The Class 9 / Class 42 software line. This single distinction swallows an enormous amount of modern technology practice and gets its own deep dive below. Decide it deliberately; if your product genuinely exists in both a downloadable and a hosted form, you likely belong in both classes.
"Close enough" is not close enough. The Manual is organized by precise goods, not by industry. "Software" alone is never acceptable; the Manual will make you state the software's function and often its field of use. Skimming for a superficial match and pasting it in is how filers end up in the wrong class.
Verbatim, not approximate. The surcharge relief and the presumption of acceptability attach to using an entry as written. Paraphrasing a Manual entry converts it, for fee and examination purposes, into free-form text — you lose the benefit while thinking you kept it.

Reading the Manual like a classifier

Each Manual entry pairs a description with a class and, frequently, a note about what additional detail the examiner will require. The discipline is to search by the good or service, not the industry. A "coffee company" is not a thing the Manual classifies; roasted coffee (Class 30), coffee mugs (Class 21), a coffee-subscription retail service (Class 35), and a café (Class 43) are. Break every inventory item down to the concrete deliverable, search that, and let the Manual assign the class. When you reach a genuinely new kind of offering with no on-point entry — which is rarer than founders think — model your custom description on the closest analogous entries: same level of specificity, same "type + function + field" skeleton, no open-ended connectors. For the surrounding mechanics of turning this into a filed application, see Filing a Federal Trademark Application: A Step-by-Step Guide and Registering a Trademark with the USPTO: A Complete Guide.

Phase 3 - Draft accurate, properly scoped identifications

[ ] Replace any bare class heading or class number with a specific, definite listing ("t-shirts, sweatshirts, and baseball caps," never "clothing" or "IC 25").
[ ] Describe the real product line fully, aiming for accurate breadth — because you can narrow later but never broaden (TMEP § 1402.06).
[ ] Confirm every identification is definite enough to classify and understandable to an ordinary English speaker (37 C.F.R. § 2.32(a)(6); TMEP § 1402.01).
[ ] Strip out open-ended and vague terms — "including," "such as," "and the like," "accessories," "systems," "solutions" — and any third-party trademarks.
[ ] Assign each item to its correct international class (37 C.F.R. § 2.85), and make sure each class actually contains only goods or services that belong there (TMEP § 1402.01(a)).
[ ] Claim only the classes you genuinely use or intend to use — resist the urge to "cover everything."
[ ] Confirm the filing basis for each class (use vs. intent to use) and that you can back it up.

WHY. Definiteness and accuracy are the twin gatekeepers. An indefinite identification draws an office action and delays the filing; an inaccurate one — claiming goods you do not offer — puts the registration itself at risk. The drafting sweet spot is language that is definite enough to satisfy the examiner and broad enough to cover your real line and its foreseeable minor variations, without wandering into goods you do not actually make. The USPTO itself illustrates the balance: "hats" is a better identification than "a wide variety of ladies moderately priced fashion hats made of felt," because the shorter version is definite, accurate, and leaves room to sell a wool hat next year without a new filing. Over-specify and you fence yourself in; over-reach and you invite attack.

TRAPS.

The class heading is never enough — in the U.S. Many countries let an applicant claim an entire Nice class heading and take the whole class. The United States does not. A heading such as "pharmaceuticals," "jewelry," "clothing," or "telecommunications services" is too broad to be an acceptable identification here; you must itemize. This is the single most common indefiniteness rejection. (For a comparative note on how differently the rest of the world treats headings, see the discussion of IP Translator below.)
The one-way door (TMEP § 1402.06). Amendments may clarify or restrict, but they may never add goods, add services, or expand the scope of what was originally filed. Start with "hats" and you can later amend down to "baseball caps"; you cannot amend up to "all headgear." This asymmetry means your original draft sets a ceiling you can never raise. Draft the ceiling where your real business is, not where it was the week you incorporated.
The definiteness–breadth tension. Definiteness pushes toward narrow, concrete nouns; the no-broadening rule pushes toward capturing your full real line at the outset. Resolve the tension by listing your actual products at the natural level of generality the Manual uses ("hats," "t-shirts," "handbags"), not by piling on adjectives (which narrows) and not by retreating to headings (which is indefinite).
Vague connectors and puffery. "Including" and "such as" are non-limiting and therefore indefinite; the USPTO reads them as leaving the list open. "Accessories," "systems," "solutions," and "products" name nothing in particular. Marketing language migrates into identifications when drafters copy from a business plan; scrub it.

The examiner is not bound by the last examiner

A recurring hope among filers is that because the USPTO accepted a given identification for someone else's registration, it must accept the same words for them. It is a reasonable instinct and a legally unfounded one. Each application "must be examined on its own merits," and prior acceptances do not bind a later examiner. See In re Nett Designs, Inc., 236 F.3d 1339, 1342 (Fed. Cir. 2001); accord In re Loggerhead Tools, LLC, 119 U.S.P.Q.2d 1429 (T.T.A.B. 2016). Prior registrations are a useful research input — they show what language has cleared — but never a guarantee. Use the Trademark Search database to see what worked; use the ID Manual to know what will work now.

Worked example: the narrowing trap in real time

Return to a fashion-forward hypothetical. Halcyon Row launches selling only felt fedoras and files an identification reading "felt fedoras" in Class 25. It is definite, accurate, and — as it turns out — a cage. Eighteen months later Halcyon Row adds knit beanies, bucket hats, and a small line of scarves. None of those is a "felt fedora." Because of the one-way door, the company cannot amend its pending or registered identification to reach them; it must file a new application (and pay again) to cover the expanded line, and it loses the earlier filing date as to those new goods. Had Halcyon Row filed "hats" — or, better, "hats; scarves" if scarves were genuinely planned and backed by a bona fide intent — it would have carried the fedoras plus room to grow, all under one priority date. The lesson is not "claim everything," but "claim your real line at the Manual's natural altitude." Timing this well interlocks with the priority and intent-to-use analysis in Timing Your Trademark Filing: A Practical Guide to Clearance, Priority, and Intent to Use and Intent-to-Use Trademark Filings: A Practical Guide.

The over-claim trap: how a padded identification becomes fraud

The opposite error is more dangerous than it looks. Claiming goods or services you do not actually offer — padding the identification to look bigger, or copying a competitor's list wholesale — exposes the registration to attack on two fronts.

The first is non-use cancellation. A registration can be cancelled for goods on which the mark was never used or has been abandoned. See 15 U.S.C. § 1064. If you claim ten items and use the mark on six, the four phantom goods are a standing invitation.

The second, and graver, is fraud on the USPTO. In the leading modern case, the Board cancelled a registration in its entirety where the owner had declared use on goods it had never sold. See Medinol Ltd. v. Neuro Vasx, Inc., 67 U.S.P.Q.2d 1205 (T.T.A.B. 2003). The Federal Circuit later reined in Medinol's "knew or should have known" formulation, holding that fraud requires a subjective intent to deceive, not mere negligence or an honest mistake. See In re Bose Corp., 580 F.3d 1240 (Fed. Cir. 2009). Bose makes accidental over-claiming survivable, but it does not make it safe: proving your own honest mistake is a litigation you would rather not have. The prophylactic is simple — claim what you use or genuinely intend to use, item by item, and no more.

The third pressure point is administrative and increasingly routine: the USPTO's post-registration audit program, under which the Office randomly selects registrations and demands proof of use for particular listed goods or services. See 37 C.F.R. § 2.161(h). Fail to prove use of an audited item, and it is deleted; fail to respond adequately, and the whole registration can fall. A padded identification is a landmine that the Office may step on for you years later. The connection between what you claim now and what you must prove later runs straight through the maintenance filings covered in Keeping a Registration Alive: A Trademark Maintenance and Renewal Checklist and Filing the Section 8 Declaration of Continued Use: A Practical Checklist.

Phase 4 - Decide the class count, fees, and structure

[ ] Confirm the current per-class base fee and every applicable surcharge on the live USPTO fee schedule before you commit numbers.
[ ] Avoid the "insufficient information" surcharge by filing a complete application (currently reported at $100/class).
[ ] Avoid the free-form identification surcharge (reported at $200/class under the 2025 rule) by using ID Manual wording — the biggest behavior-based cost lever you control.
[ ] Avoid the lengthy free-form text surcharge by keeping any necessary custom wording concise (an added charge applies per extra block of characters beyond the first 1,000 per class).
[ ] Choose the structure: a single multi-class application (37 C.F.R. § 2.86) versus separate single-class filings.
[ ] Plan one specimen per class, each showing the mark used on the goods or services in that class.
[ ] Confirm you can later delete a refused class or divide the application (37 C.F.R. § 2.87) so clean classes register while a problem class is worked out.

WHY. Fees scale linearly with class count — at filing and again at every Section 8 and Section 9 maintenance step over the life of the registration. See 15 U.S.C. §§ 1058, 1059. Because the class count is an output of the identification, Phases 1 through 3 are, in economic terms, the budgeting exercise; Phase 4 just prices it and picks a container. Two decisions dominate the cost: how many classes you really occupy (a substance question resolved above) and how you word the identification (a behavior question resolved by using the Manual).

The 2025 fee restructure. On January 18, 2025, the USPTO retired the old TEAS Plus / TEAS Standard tier system and replaced it with a single per-class base application fee (widely reported at $350 per class) plus behavior-based surcharges. The surcharges are the point of the reform: they price the extra examiner labor that sloppy or custom filings create. Three matter for classification:

an insufficient-information surcharge (reported at $100/class) when the application is missing information the Office would otherwise have to chase;
a free-form (custom) identification surcharge (reported at $200/class) when you draft your own identification instead of selecting from the ID Manual; and
a lengthy-identification surcharge for free-form text beyond a set character count per class.

The behavioral math is stark. A four-class application that uses ID Manual wording throughout avoids roughly $800 in custom-identification surcharges compared with the same filing drafted free-form — before counting the office actions that indefinite custom language tends to draw. Any figure quoted here is a snapshot; the tiered TEAS numbers are now obsolete, and the current numbers change by rulemaking, so verify the live fee schedule the day you file.

TRAPS.

One specimen does not cover a multi-class application. Each class needs its own specimen showing the mark in use on the goods or services of that class. Filers routinely upload a single logo photo for a four-class application and are surprised by a specimen refusal in three classes. Plan specimens class by class; for the standards, see Preparing Specimens That Pass Examination: A Practical Checklist.
Multi-class convenience, multi-class fragility. A single Section 66(a) or Section 1 multi-class application is cheaper to manage and shares one serial number, but a refusal in one class can hold up the others. Know your escape hatches: you can delete the problem class, or divide the application under 37 C.F.R. § 2.87 so the clean classes proceed to registration while you fight the refusal in a child application. Division preserves the filing date for the separated class.
Quoting dead fees. Any advice, template, or article referencing "TEAS Plus" or "TEAS Standard" pricing is pre-2025 and wrong. Likewise, do not resurrect the defunct pre-1973 U.S. classification numbers that occasionally surface in old materials; the U.S. has used the Nice system for domestic applications since 1973.
Deferring the count. If you omit classification, the examiner will assign it — and that delay, plus the risk of the examiner reading your identification differently than you intended, is avoidable. Classify affirmatively.

Worked example: pricing Thornbury

Recall Thornbury Provisions from Phase 1, which could touch five classes (18, 21, 25, 30, 35), plus a planned Class 43 tasting room. Suppose Thornbury decides its core, in-use offerings are the sauces and seasonings (30), the drinkware (21), and the online shop (35), and that the apparel and totes are early enough to defer. That is a three-class, use-based application today. Using ID Manual entries verbatim for all three, Thornbury pays three base fees and zero identification surcharges. Drafting the same three identifications free-form would add roughly $600 and several likely office actions. When the tasting room opens, Thornbury files a separate Section 1(a) application in Class 43 on a real first-use date — cleanly, without having carried an unsupported Class 43 claim (and its audit exposure) for a year. The structure decision here — three-class now, single-class later — is exactly the kind of judgment the full From Brand Idea to Registered Mark: A Trademark Registration Toolkit is built to support.

Phase 5 - Clear the coordinated and related classes

[ ] Search not only your classes but the coordinated and related classes the examiner will cross-search for conflicts.
[ ] Work the known clusters: the fashion cluster (Class 25 clothing ↔ Class 35 retail ↔ Classes 18/14/9 for bags, jewelry, eyewear) and the software/tech cluster (Class 9 ↔ Class 42 ↔ Class 38 telecommunications).
[ ] Assess conflicts by the relatedness of the goods and services, not by shared class numbers.
[ ] Run design search codes for any logo or device elements alongside the word-mark search.
[ ] Feed the results back into Phases 1–4 — a conflict may change which classes you file, or push you toward a coexistence strategy.

WHY. This is the phase where filers confuse the map for the territory. Classification is administrative; likelihood of confusion is not. Two marks in the same class can lawfully coexist if their goods are unrelated (identical marks on "hammers" and on "financial newsletters," both arguably reachable within a class, do not necessarily conflict). Two marks in different classes can absolutely collide if their goods are commercially related and move through the same channels to the same buyers. The examining attorney's Section 2(d) search — and any future opposer's diligence — follows relatedness, which is why the USPTO maintains lists of coordinated classes that examiners cross-search as a matter of course.

The controlling framework is the second of the thirteen DuPont factors: "the similarity or dissimilarity and nature of the goods or services." See In re E. I. du Pont de Nemours & Co., 476 F.2d 1357 (C.C.P.A. 1973). Relatedness asks whether consumers encountering similar marks on these particular goods would reasonably assume a common source — a question about the marketplace, not the class index. For the full factor analysis, see Likelihood of Confusion in Trademark Law: A Brand Owner's Guide and the step-by-step workup in Working Up Likelihood of Confusion: A Practical Checklist.

TRAPS.

Clearing your class but not the cluster. A clothing brand that clears Class 25 and ignores Class 35 can sail through examination and then draw an opposition from a similarly named apparel retailer — because clothing and clothing-retail services are textbook related. The classes differ; the relatedness does not.
Treating "different class" as "safe." It is not a defense to a 2(d) refusal or an opposition that the prior mark sits in another class. Relatedness, channels of trade, and consumer overlap decide the question.
Skipping design codes. If your mark includes a device (a stylized animal, a monogram, a distinctive shape), the examiner searches design search codes, not just letters. A clearance that checks only the word mark misses a whole category of prior rights.

A proper classification pass and a proper clearance search are two halves of the same coin: classification tells you where you sit, clearance tells you who else is nearby. Run them together. For the full clearance methodology, see Clearing a New Trademark: A Practical Checklist and Trademark Clearance Searching: A Comprehensive Guide.

Worked example: the fashion cluster bites back

Halcyon Row (from Phase 3) clears "hats; scarves" in Class 25 and finds no confusingly similar registered mark in Class 25. It files. Six weeks after publication, an opposition lands from Halcyon & Co., owner of a registration for retail clothing store services in Class 35. The goods and the services are commercially related — a consumer who knows "Halcyon" for a clothing shop may well assume the "Halcyon Row" hats come from the same house. Halcyon Row cleared its class and missed its cluster. Had Phase 5 been run properly, the Class 35 registration would have surfaced during clearance, and Halcyon Row could have chosen a different mark, negotiated a consent, or narrowed its channels — before spending the filing fee.

Deep dive: the Class 9 / Class 42 software line

No single classification question generates more confusion — or more wasted filings — than how to class software. The rule sounds simple and hides real subtlety.

The core distinction is the mode of delivery.

Downloadable software is a good, in Class 9. A mobile app the user installs, a program delivered as a file, firmware embedded in a device — these are goods. The user takes possession of a copy.
Non-downloadable, hosted software is a service, in Class 42. Software-as-a-service, a web application the user accesses through a browser without downloading, "platform as a service" — these are services under the heading "providing temporary use of non-downloadable software." The user never takes a copy; they use the provider's running instance.

Modern products routinely exist in both forms at once: a SaaS analytics platform (Class 42) with a companion downloadable mobile app (Class 9), perhaps plus a messaging or data-transmission feature that reaches into Class 38 (telecommunications). If your product genuinely lives in two of these modes, you generally need to file in both classes — and pay for both — because a Class 42 registration does not cover the downloadable app, and vice versa.

Two further wrinkles trip up even experienced filers:

Function and field are mandatory. "Software," "software platform," or "mobile application" is never an acceptable identification standing alone. The USPTO requires the software's function ("software for managing inventory," "software for editing photographs") and, for many entries, its field of use ("...in the field of commercial real estate"). See TMEP § 1402.03(d). The ID Manual entries are built this way; follow their skeleton. A bare "downloadable software" will draw an indefiniteness refusal every time.

The line can move under your feet. A product that starts life as a hosted service (Class 42) and later ships a downloadable client (Class 9) has crossed a class boundary — which, thanks to the one-way door, cannot be fixed by amendment. It needs a new application for the new class. Founders who expect to add a downloadable component should consider claiming Class 9 from the outset on a bona fide intent-to-use basis, rather than discovering the gap after launch.

Worked example: Lumen Labs

Lumen Labs, a hypothetical analytics company, offers (1) a browser-based dashboard customers log into (hosted, non-downloadable), (2) a downloadable iOS/Android app that syncs to the dashboard, and (3) an in-app secure messaging feature. A one-line "software" identification would be wrong three times over. The correct spread is roughly: Class 42 for "providing temporary use of non-downloadable software for data analytics and business-intelligence reporting"; Class 9 for "downloadable mobile application software for viewing and analyzing business data"; and, if the messaging feature is marketed as its own service, Class 38 for the transmission function. Three classes, three fees, three specimens — and a registration that actually matches the product. The broader IP architecture for a product like this (copyright in the code, trade-secret protection for the models, the trademark layer) is mapped in Intellectual Property Strategy for Mobile Apps: A Comprehensive Guide.

Special situations and edge cases

Retail and e-commerce services (Class 35). Running a store — physical or online — is a service in Class 35, described as "retail store services featuring [category of goods]" or "online retail store services featuring [category]." It is distinct from, and additional to, the classes for the goods sold. A brand that both makes a product and sells it (and others') through its own shop occupies at least the goods class and Class 35. Note that Class 35 also houses advertising, business consulting, and marketplace services — do not let its breadth tempt an over-claim; itemize the actual retail activity.

House marks across a product family. A single "house mark" used across many products may need multiple classes even though it is one mark. The classification follows the goods and services, not the number of marks. A house mark on apparel, fragrances, eyewear, and a retail store spans Classes 25, 3, 9, and 35 respectively.

Kits and combination products. A product that bundles items from different classes is classified by its components or by the primary purpose of the assembled kit, per the Manual's guidance on combination goods. Do not assume a boxed set is one class; check the entry.

Trade dress and non-traditional marks. Classification attaches to the goods or services, not to the type of mark. A product-configuration trade dress, a color, or a sound is still classified by what it identifies. The classification analysis here is orthogonal to the protectability analysis for such marks, which lives elsewhere.

International filings under Madrid. If you plan to extend protection abroad through the Madrid Protocol, your U.S. application (or registration) is the "basic" filing, and its identification and classes propagate to every designated country. A sloppy or overly narrow U.S. identification exports its problems worldwide; a fraud problem can trigger a central-attack cancellation that cascades. Get the U.S. classification right first. For the outbound mechanics, see Acquiring, Licensing, and Exploiting Trademark Rights.

A comparative footnote: why "class headings" fail here but work elsewhere

United States practice forbids claiming a bare Nice class heading; you must itemize. Much of the world once did the opposite, treating a heading as claiming everything in the class ("class-heads-cover-all"). That divergence produced one of the most consequential trademark decisions of the last two decades in Europe: Chartered Institute of Patent Attorneys v. Registrar of Trade Marks ("IP Translator"), Case C-307/10 (CJEU, June 19, 2012), which held that class-heading terms must be interpreted with sufficient clarity and precision, and that an applicant using a heading must specify whether it intends to claim all goods in that class. The practical upshot — that vague heading claims will not do — moved EU practice toward the specificity the United States had required all along. The lesson for a U.S. filer working internationally is that the American habit of itemizing is now closer to the global norm, not an outlier; draft precisely and you draft portably.

Common mistakes, restated as a pre-flight scan

Run this list before you hit submit. Each maps to a phase above.

Under-classifying — treating a multi-class business as one class and leaving real product lines unprotected (Phase 1).
Filing a bare class heading or class number as the identification — indefinite, and the most common office action (Phases 2–3).
Writing custom descriptions when a Manual entry exists — a $200/class penalty since 2025, plus slower examination (Phases 2, 4).
Collapsing "our software" into one class when it exists as both a downloadable good (Class 9) and a hosted service (Class 42) (software deep dive).
Over-narrow drafting — trapping yourself behind the no-broadening rule with adjective-laden identifications (Phase 3).
Over-broad or padded drafting — inviting non-use cancellation, audit deletion, and, at the extreme, a fraud claim (Phase 3).
Forgetting the per-class specimen — assuming one specimen covers a multi-class filing (Phase 4).
Clearing your class but not the cluster — missing coordinated/related classes an examiner or opposer will find (Phase 5).
Quoting dead fees or dead classes — TEAS-tier pricing and pre-1973 U.S. class numbers are both obsolete (Phase 4).

When an office action does arrive — and on identification grounds they are common and usually fixable — the response playbook is set out in Overcoming a USPTO Trademark Office Action: A Practical Checklist. If your filing is on an intent-to-use basis, the downstream proof step is covered in Completing an Intent-to-Use Application: A Statement of Use Checklist.

Where classification sits in the life cycle

Classification is one gate in a longer sequence: pick a protectable mark, clear it, class and identify the goods and services, file, prosecute through examination and publication, register, and then maintain. Each gate constrains the next. A weak or unavailable mark makes classification moot. A mis-scoped identification narrows what registration buys you and complicates every renewal. And the classes you choose today are the classes you will prove use in — and pay for — at every Section 8 and Section 9 filing for as long as you own the mark. For the end-to-end view and a filing-desk workflow, see Prosecuting a Federal Trademark: A Filing-Desk Checklist from Application to Renewal, the collected Federal Trademark Registration Lifecycle: A Checklist Collection from Clearance to Incontestability, and, for what the resulting registration is actually worth, What Federal Trademark Registration Actually Buys You: A Practical Guide.

Primary authority

Regulations. 37 C.F.R. § 2.85 (adoption of the Nice Classification); § 2.86 (applications covering multiple classes); § 2.87 (dividing an application); § 2.32(a)(6) (identification requirement); § 2.34(a) (use-in-commerce requirements, per class); § 2.161(h) (post-registration audit / proof of use).

TMEP. §§ 1401–1402 (classification and identification generally); § 1401.03(b) (classification is strictly ministerial); § 1402.01 and § 1402.01(a) (definiteness and correct-class requirements); § 1402.03(d) (identifying computer software by function and field); § 1402.06 (amendments may not broaden the identification); § 1301.01(a) (what qualifies as a service).

Statutes. Lanham Act § 1(a)–(b), 15 U.S.C. § 1051(a)–(b) (use and intent-to-use bases); § 2(d), 15 U.S.C. § 1052(d) (likelihood-of-confusion refusal); § 45, 15 U.S.C. § 1127 (definition of "commerce"); §§ 8 and 9, 15 U.S.C. §§ 1058, 1059 (per-class maintenance and renewal); § 14, 15 U.S.C. § 1064 (cancellation, including for non-use and fraud).

Cases. In re E. I. du Pont de Nemours & Co., 476 F.2d 1357 (C.C.P.A. 1973) (relatedness of goods/services as a confusion factor); In re Nett Designs, Inc., 236 F.3d 1339 (Fed. Cir. 2001) (each application examined on its own merits); In re Loggerhead Tools, LLC, 119 U.S.P.Q.2d 1429 (T.T.A.B. 2016) (same); Medinol Ltd. v. Neuro Vasx, Inc., 67 U.S.P.Q.2d 1205 (T.T.A.B. 2003) (fraud from claiming unused goods); In re Bose Corp., 580 F.3d 1240 (Fed. Cir. 2009) (fraud requires subjective intent to deceive). Comparative: Chartered Institute of Patent Attorneys v. Registrar of Trade Marks ("IP Translator"), Case C-307/10 (CJEU 2012) (class-heading terms require clarity and precision).

Treaty and tools. Nice Agreement Concerning the International Classification of Goods and Services (1957), administered by WIPO (12th edition, revised annually); USPTO Trademark ID Manual (revised weekly); USPTO fee schedule (per-class base fee plus behavior-based surcharges under the rule effective January 18, 2025).

This checklist is general information, not legal advice, and fees and classifications change. Confirm the current USPTO fee schedule and the live ID Manual before you file, and consult qualified trademark counsel about any particular mark.

Related UpLaw Resources

Related Guides

Related Checklists

Related Toolkits

From Brand Idea to Registered Mark: A Trademark Registration Toolkit

Related Articles