Why recent patent filing trends look like a decline

Recent patent filing trends look like a decline because most recent filings are not public yet. US, PCT and European applications are generally published about 18 months after their earliest priority date, grants come later still, and databases add documents some time after publication. A chart counted by priority or filing year therefore drops in its last two or three years even when the field is growing.

Where the lag comes from

Three delays add up before a filing shows on your chart.

Publication. Under 35 U.S.C. § 122(b)(1)(A), a US application is published promptly after the expiration of 18 months from the earliest filing date for which a benefit is sought. International applications under the PCT are published promptly after 18 months from the priority date (PCT Article 21), and European applications after 18 months from filing or priority (Article 93 EPC). Before publication, an application cannot be found by any search outside the office.

Grant. If a chart counts granted patents, examination adds years on top of publication. The delay varies by office and technology, so grant charts are a poor guide to recent activity.

Indexing. Data providers add newly published documents some time after publication. The delay depends on the provider and the office, and it is rarely zero.

Two US rules that hide filings longer

• Nonpublication requests. Under § 122(b)(2)(B)(i), an applicant can ask that an application not be published if the invention has not been and will not be the subject of an application filed abroad that requires 18-month publication. Those applications stay out of view until they issue.
• Applications filed before November 29, 2000. US 18-month publication came with the American Inventors Protection Act of 1999. Applications filed before that date were generally not published before grant. A long US series counted by publication date therefore jumps once later-filed applications began to publish, which reflects the change in law rather than a change in invention.

US 6,285,999 B1, "Method for node ranking in a linked database", shows how far apart the dates can sit. It claims priority to 10 January 1997, was filed on 9 January 1998 and issued on 4 September 2001. Filed before November 29, 2000, the US application would generally not have been published before grant, so the US document most likely entered the public record more than four and a half years after its priority date. On a priority-year chart, it belongs to 1997.

A worked example

Take a hypothetical field in which applicants file exactly 1,000 new applications in every priority year, spread evenly across the year. The true trend is flat. Assume every application publishes 18 months after its priority date and that your data provider adds documents three months after publication. You run the analysis at the start of August 2026, so you can see only what was published by about the start of May 2026.

A flat field reads as a collapse: 1,000, 1,000, 830, 0, 0. Real data is noisier because filings are not evenly spread and delays vary, but the shape is the same. The curve bends down in its last two or three years whatever the field is doing.

The older rows are not fixed either. A continuation filed in 2026 that claims the benefit of a 2021 application adds a document to 2021, so counts for older priority years keep creeping up for years after they look complete.

The same arithmetic gives a rule of thumb: a priority year is close to complete about 18 months after it ends, plus your provider's indexing delay. In the example, 2024 will look complete around the autumn of 2026, and 2025 not before the autumn of 2027.

How to correct for it

Count by priority or filing date, and stop where the data is complete

The earliest priority date is the closest public proxy for when an invention was made. Count by it (or by filing date), then either cut the series at the last complete year or mark the incomplete years clearly. In the example, 2022 and 2023 are solid, 2024 is provisional, and 2025 and 2026 should not be plotted as data points at all.

Compare ratios rather than raw counts

A company's share of the field in a given priority year is far less sensitive to the lag than its raw count, because the numerator and the denominator are both cut by the same delay. If a competitor filed 50 of the field's 1,000 applications in 2023 and 42 of the 830 visible for 2024, its share has held at about 5 percent; the raw drop from 50 to 42 means nothing. The correction is not perfect. Applicants who file only in the US and request nonpublication, or who use a different filing route, publish on a different schedule from the rest of the field.

Use publication-date trends carefully

Counting by publication date removes the cliff at the end, because every document counted is already public. It answers a different question, though: how much new material the offices released in a period, not when the inventions were made. A publication series moves with office workload and with changes in law such as the US switch to 18-month publication, and it describes inventions made roughly 18 months or more earlier.

Compare against the whole field

Any claim that a company, a cluster or a jurisdiction is growing or shrinking should be measured against the field over the same years and from the same data pull. If every assignee drops in 2024, the drop is lag. If one assignee drops while the field holds, that is a signal worth investigating.

Prefer applications to grants for recent activity

Grants trail applications by years. For any question about what is happening now, count published applications and treat grants as a lagging confirmation.

A quick check for any filing chart

Before relying on a trend someone hands you, get answers to these:

1. Which date is on the horizontal axis: priority, filing, publication or grant?
2. When was the data pulled, and from which source?
3. Are the incomplete years marked or removed?
4. Does the chart count documents or families, applications or grants?
5. Are the figures raw counts or shares of the field?

A chart that cannot answer the first two questions cannot tell you whether a field is slowing down.

Reading filing charts in One Transform

One Transform's Technology Landscape search charts filing activity over time and across jurisdictions for a field you define. Read the most recent years of those charts with the lag in mind, and compare clusters and assignees by their share of the field rather than by raw counts. For the other sections of a landscape report, see How to read a patent landscape report.

Next step

Run a Technology Landscape search on a field you follow, mark the priority years that are still filling in, and read the trend only up to the last complete year. Start at onetransform.ai.

This article is general information about patent practice, not legal advice. For advice on a specific matter, consult a registered patent attorney or agent.