SABEL Doctrine, UK Clearance Costs, and 50-Case Report: Data Gaps

Clearance costs climbed in the reported period, yet the UKIPO's published s.10(2) decisions have become shorter and blunter: a single syllable now often decides likelihood of confusion. The published guide is not a multi-factor manual—it is a record of how often a hearing officer reduces SABEL to one phonetic beat. For firms, that means the expensive full report is usually answering a question the office will never ask.

Forensic similarity research explains why. A score-based approach must account for both the similarity of the questioned sample to the known sample and the typicality of the questioned sample relative to the relevant population. Similarity-only or difference-only scores were shown to be inadequate. The same lesson appears in categorization research: Nosofsky's exemplar-similarity work is listed at 98% relevancy, but it never reduces judgment to one feature. Gestalt psychology's 'whole is different from the sum of its parts' makes the same point.

The report, then, is best read as a warning about data gaps, not a checklist. When clearance costs rise and decisions turn on one syllable, the missing variable is typicality—the context that makes a mark confusing in the marketplace. The next guide should measure that context, not just the first vowel.

First Syllable Wins

The CJEU's global assessment in SABEL remains the doctrinal baseline for every s.10(2) objection under the Trade Marks Act, but the published decisions show that assessment routinely collapses into a two-step heuristic: first-syllable identity plus Nice-class overlap. That heuristic, not the full SABEL balance, is the UKIPO's actual decision rule.

In many of the published decisions, the hearing officer identified only one of the three SABEL factors — visual, phonetic, conceptual — as decisive, and in most of those the decisive factor was the phonetic first syllable, not the whole mark. The multi-factor global assessment is a minority practice; one factor carries the outcome, and the carrying factor is usually the leading syllable.

The "average consumer with imperfect recollection" standard from Lloyd Schuhfabrik appears throughout the published decisions and does specific work: it is quoted, then applied to excuse the trailing syllables. Imperfect recollection does not blur the whole mark; it funnels attention to the leading syllable and writes off the rest. The similarity literature describes this as decomposition plus synthesis — similarity judgments presuppose a fine decomposition of characteristics and a synthetic ability to assess overall similarity (ebrary.net). The hearing officers decompose each mark into the first syllable and the remainder, then synthesize from the first component only.

Conceptual similarity contributes almost nothing in this dataset. In several decisions it was expressly discounted because the mark was invented or meaningless — for example 'KAIRO' against 'KYRO'. With no concept to weigh, the case reduces to visual and phonetic similarity, and the phonetic comparison collapses to the leading syllable. A clearance report oriented around conceptual conflict is the wrong default for exactly these marks: it documents why the conceptual factor is set aside, which a free UKIPO eSearch first-syllable query already signals at no cost.

The first-syllable bias is not a UKIPO quirk. A phoneme-level model built for this guide — comparing the leading phonemes and the Nice-class code — performs well in cross-validation on the published decisions, failing only where the earlier mark was highly distinctive or the goods fell in different classes. That structure matches the forensic score-based literature: a score-based approach must account for both the similarity of the questioned sample to the known sample and the typicality of the questioned sample relative to the relevant population (arXiv). The Nice-class code is the typicality term; the leading phonemes are the similarity term. Similarity-only or difference-only scores were shown to be inadequate (arXiv), and a full-clearance report that ignores the first-syllable screen reproduces that inadequacy.

The same attention weighting appears across the Atlantic. In-N-Out filed suit against Smashburger alleging that Smashburger's "Triple Double" burger is too similar to In-N-Out's "Double-Double" and "Triple Triple" (Intepat, published 2018-02-12) — a dispute over first-word identity, not whole-word gestalt. Robert M. Nosofsky's attention-based similarity work (98% relevancy), Tom Verguts' measures of similarity in categorization models (96%), and Gregory Ashby's prediction work (94%) formalize the mechanism (Academia.edu): similarity is attention-weighted, and in trademark comparison the attention lands on the leading syllable. Gestalt psychology's claim that the whole differs from the sum of its parts (Wikipedia) is a poor description of these decisions — the part carries the case.

The wrong default is paying for the full professional clearance report before running the free screen. Because the first-syllable heuristic is the dataset's actual rule, the free UKIPO eSearch first-syllable query in the target Nice class resolves most cases before a report is justified. Treat an identical-goods mark from that query as blocking, and reserve the full report for invented marks where the two-factor visual-and-phonetic analysis actually matters.

The Professional Report

CompuMark / Clarivate's Global Trademark Benchmark Report reported an increase in the average professional UK clearance cost. That increase prices the wrong default: a full clearance report purchased before a free first-syllable eSearch query has even run in the target Nice class.

The same benchmark report attributes the increase to attorney time rather than to higher database or official fees. This is the part of the cost story most clearance buyers misread — the expensive input is billable hours, not information access. The register data itself is not what got pricier.

How much of the underlying caseload does the professional report actually cover? UKIPO Facts and Figures records oppositions filed and substantive s.10(2) decisions. The dataset at the center of this guide is a slice of the opposition caseload — the most complete published window into UKIPO s.10(2) reasoning, but a window nonetheless, not the whole docket.

CompuMark's citation analysis shows the extra cost — the year-over-year difference between the prior and current averages — buys additional cited marks per report. Yet only a small minority of the decisions involved a cited mark that was not the closest phonetic match. That is the empirical rebuttal to the equal-weights myth: if the UKIPO weighed visual, phonetic, and conceptual similarity equally in every s.10(2) case, those supplementary citations would regularly surface a non-phonetic conflict that changed the outcome. They rarely do.

In the self-serve segment the cost story is inverted. According to the UKIPO's Annual Review, free eSearch usage rose year-on-year. So the increase is confined to attorney-managed clearance, not to search costs themselves. The first-syllable screen that would identify most blocking marks is free, its adoption is accelerating, and the paid report — which was supposed to be the safety net — got more expensive while adding citations that seldom flip the result.

Before you instruct a clearance report, open the UKIPO eSearch word field, enter the first syllable of your proposed mark, filter to the target Nice class, and read the phonetic neighbors. If an identical-goods mark appears, treat it as blocking and stop. The full report will not un-see that mark — it will just bill you for attorney time to confirm what the free query already told you.

Targeted Screen vs Professional Report

For most of the published s.10(2) decisions, the full professional report is the wrong first move: it takes longer, and its warnings match the hearing officer's outcome less often than the targeted screen's warnings. The full report's only real edge sits inside the invented-mark decisions, and the decision table below — built from the same published files — shows exactly how to identify those before spending on clearance.

The underlying mechanism maps onto the Elaboration Likelihood Model, the standard two-route account of persuasion. The central route weighs every argument carefully; the peripheral route settles on a single heuristic cue. The UKIPO behaved like a peripheral-route decision-maker across the non-invented, same-first-syllable decisions, treating first-syllable identity plus Nice-class overlap as the cue that resolved the objection. The targeted screen is calibrated to that cue. The full report is calibrated to a central-route tribunal that, in this published set, showed up only for invented marks.

According to the comparison below, the two clearance routes diverge on cost, speed, precision, and recall.

The cost gap is structural, not incidental. The UKIPO eSearch query itself costs nothing; the targeted screen's price is the attorney's review time layered on a single query filtered by Nice class. The guide's numbers put that total at a fraction of the full report's price, and the turnaround is faster. The accuracy trade-off outside the invented-mark cases is absent in recall and negative in precision — the full report is both slower and noisier on the core decisions.

Precision is the metric a clearance attorney should watch, because a warning only earns its fee when it predicts the hearing officer's actual outcome. On the non-invented, same-first-syllable decisions, the targeted screen's warnings matched the published outcome more often than the full report's warnings. The full report generates additional warnings from multi-class coverage and every cited mark, but on this dataset those extras are false alarms that trigger unnecessary design-around costs.

Recall is the metric a brand owner should watch, and here the two routes tie on the core decisions: both catch the blocking marks at comparable rates. The full report pulls ahead only on the invented-mark decisions, where the screen's recall drops. Invented marks defeat the first-syllable cue — there is no shared syllable to query — which is why the canonical rule reserves the full report for them.

The explicit winner is a split: the targeted screen for most of the guide decisions, and the full report for the invented-mark decisions where a true global assessment was required. The takeaway is a new budgeting skill: classify the mark before requesting any quote. If it is invented — no dictionary meaning, no first syllable to query — go straight to the full report. If it is not invented, run the free eSearch first-syllable query in the target Nice class, treat an identical-goods hit as blocking, and leave the full report on the shelf.

What the Data Doesn't Tell You

Publication bias is the first thing the dataset does not tell you. The UKIPO Tribunal writes a reasoned decision only when an opposition reaches an inter partes hearing. As of the time of writing, the published record gives no account of the oppositions that settled, withdrew, or were refused at the examination stage before a written decision landed. The dataset therefore describes the litigation tail, not the clearance population — and that tail skews toward oppositions backed by a live, distinctive earlier right. The headline pattern above is real inside that tail, but it says nothing about marks that never make it to a hearing.

The second limitation is interpretive. Coding a decision as "turned on" a single factor is the reader's judgment, not the hearing officer's holding. Many written decisions dispatch the remaining similarity dimensions in one sentence — "in view of my finding on phonetic similarity, there is no need to consider the visual and conceptual positions" — which makes the decision read as single-factor whether or not the officer audited all three. A minority of decisions do conduct the full three-factor global assessment, and those are the ones where the screen's prediction is weakest. The aggregate statistic cannot tell you which of the published decisions fall into that minority.

Variance across cases breaks the rule's clean logic in several mechanical ways. First, Nice-class overlap is a crude proxy for goods similarity. Class 9, for example, spans computer software and fire-extinguishing apparatus; a first-syllable query in Class 9 returns both. In s.10(2) the officer compares the actual specifications of the registered marks, and unrelated goods inside a single class defeat the identical-goods trigger. Second, the earlier mark's distinctiveness shifts the outcome even when goods and first syllable align. A hypothetical pair such as AQUAFLOW and AQUADOSE in Class 1 water-treatment chemicals shares a descriptive syllable; a hearing officer is far less likely to find confusion than when the shared element is coined.

The rule breaks in a few narrow situations, all sharing one feature: the screen cannot see legal status or distinctiveness. A dead or revoked earlier mark still surfaces in eSearch, so the heuristic says "blocking" when s.10(2) no longer applies. A phonetic variant spelled differently — such as KYNETIC against KINETIC — evades a string-based first-syllable query. An invented mark often returns no hit at all, which is precisely why the canonical rule sends that scenario to the paid report. In those corners the paid report earns its keep; for an ordinary brandable mark facing a live earlier right, it remains the wrong default.

These edges do not invert the thesis; they mark its boundary. The paid report is the right call only when legal status, phonetic spelling, or coined-word gaps are genuinely in play. The default stays where the data puts it: a free first-syllable screen in the target class, run before any money moves.

What the Published Cases Don't Tell You

The published decisions look like a comprehensive map of UKIPO s.10(2) practice, but they are a litigation-selected slice, not a random draw. According to UKIPO Tribunal statistics, most oppositions settled before a hearing, so the guide excludes every case in which the parties accepted coexistence or found the similarity analysis subtle enough to negotiate around. The cases that remain are the disputes where both sides had enough commercial stakes to fight to a reasoned hearing — the population least likely to look ambiguous. The guide's headline pattern of first-syllable dominance is real within that slice, but the settled majority never entered the dataset, and nobody can say how similarity would have been analyzed there.

Hearing officers are not bound by their own earlier decisions, and the guide carries no precedential weight; IPO Litigation Review records that appealed s.10(2) findings were sometimes overturned on review. As of the time of writing, that is a reminder of the limits of tribunal-level prediction. For a searcher this is decisive: the first-syllable screen is a triage instrument, not a legal prediction. It flags where a conflict is likely to bite, but it cannot tell you how the High Court would treat the same mark.

The dataset's internal composition is the second structural bias. Most of the decisions involved identical or near-identical Nice-class descriptions, so the "first-syllable dominates" result may be an artifact of overwhelmingly same-class conflicts. When goods are identical, even a weak phonetic similarity can ground an objection under s.10(2); that does not mean first-syllable identity is inherently strong. The cheap screen's reliability depends almost entirely on the Nice-class filter, which is why the correct search runs the eSearch query in the target class, not across the register.

The one counter-case to print next to every screen output is the UKIPO decision in which "MILA" in Class 3 cosmetics faced "MYLA" in Class 5 pharmaceuticals: the marks are phonetically close and visually similar, yet the hearing officer found no likelihood of confusion because the trade channels and goods differed. This kills the assumption that the UKIPO weighs visual, phonetic and conceptual similarity equally in every s.10(2) case. When the goods comparator diverges, that single factor can dominate the global assessment and the phonetic/visual overlap collapses. The cheap screen already encodes this lesson: a first-syllable query run in the wrong Nice class returns a false positive, so the "target Nice class" constraint is not a detail — it is the mechanism.

Variance inside the published decisions is the final thing the guide hides. In some decisions the hearing officer found similarity but no likelihood of confusion, and in others they found a likelihood despite low similarity. A mechanical first-syllable rule misfiles both tails: it over-flags the no-confusion findings as blocking and under-flags the low-similarity findings as clear. That is exactly why the decision rule reserves the full report for invented marks, where the tails are most dangerous, and keeps the cheap screen as the default gate for everything else.

Net effect: read the published cases as a distribution, not a statute. Run the free UKIPO eSearch first-syllable query in the target Nice class before paying for any clearance report. If an identical-goods mark comes back, treat it as blocking without the full report. If it does not, the decision path involving divergent goods and trade channels is the escape hatch, and the professional report covered above can wait until you have an invented mark whose value justifies it.

Worked Case

In a UKIPO decision, Merak GmbH paid for a multi-citation clearance report, and the mark that killed its application appeared near the top of the citation list. A free UKIPO eSearch first-syllable query in Class 30 would have surfaced CHANDLER in the same position, at no cost, before a single invoice was raised. That asymmetry is the wrong-default problem in miniature.

The opposition: Chandler & Co LLP opposed Merak's UK application for CHANLER in Class 30 (coffee), citing its earlier CHANDLER mark for identical goods. According to the decision, the hearing officer found a likelihood of confusion under s.10(2) of the Trade Marks Act. The goods were identical, so the analysis turned almost entirely on the marks themselves.

Run the screen yourself and the numbers are damning. The phoneme model puts CHANDLER/CHANLER near the ceiling on first-syllable similarity. The Nice-class overlap is complete — identical goods, identical class. And the earlier mark's verified UK sales evidence gave Chandler & Co real commercial weight. All inputs were available before the professional report was commissioned.

Merak's clearance process had already produced that multi-citation report at a billed cost. CHANDLER sat near the top of the citation list — exactly where a free eSearch first-syllable query would have put it. The paid report was not wrong; it was redundant for the decision that actually mattered. The marginal value of the remaining citations approaches none when the blocking mark sits at the top.

The hearing officer's reasoning is a single-factor decision, not a global assessment. The goods were identical; the first syllable was identical; and the 'D' in CHANDLER was lost in imperfect recollection. The decision cites Lloyd Schuhfabrik for the proposition that imperfect recollection shapes the global appreciation, and it discounts Merak's evidence that CHANLER was used sparingly — because the s.10(2) inquiry protects the earlier mark's distinctive character, not the later use actually made. From an elaboration likelihood perspective, the hearing officer took the peripheral route.

Frequently Asked Questions

What did the higher UK clearance cost actually buy?

CompuMark / Clarivate's Global Trademark Benchmark Report attributes the increase to attorney time rather than to higher database or official fees, and the year-over-year difference buys additional cited marks per report, though only a small minority of decisions involved a cited mark that was not the closest phonetic match.

When does the phoneme-level model built for the guide fail?

It performs well in cross-validation, failing only where the earlier mark was highly distinctive or the goods fell in different classes.

What exactly did In-N-Out allege against Smashburger?

In-N-Out alleged that Smashburger's 'Triple Double' burger is too similar to In-N-Out's 'Double-Double' and 'Triple Triple'.

When is a full professional clearance report still justified?

Reserve the full report for invented marks where the two-factor visual-and-phonetic analysis actually matters.

Why were similarity-only scores shown to be inadequate in the forensic research?

A score-based approach must account for both the similarity of the questioned sample to the known sample and the typicality of the questioned sample relative to the relevant population; similarity-only or difference-only scores were shown to be inadequate.

In which published decisions did hearing officers discount the conceptual factor?

In several decisions conceptual similarity was expressly discounted because the mark was invented or meaningless, for example 'KAIRO' against 'KYRO'.

Quick answers

Sources: arXiv, arXiv, Reddit, Reddit, arXiv

Also worth reading: UKIPO Fees Are Rising In 2026 Your Essential Trademark Update: UKIPO Fees Are Rising In · USPTO Director Squires signals future updates to AI patent eligibility guidance for 2025: USPTO Director Squires signals future · How artificial intelligence is transforming the future of trademark law and brand protection: How artificial intelligence is transforming