Research & PolicyFree
Patent Landscape Analysis and Filing Trends
Send the assignees and the technology you care about, get filings by cluster with each year corrected for what has not published yet.
River pulls the published filings for the assignees and classifications you name, groups them into technology clusters, and counts them by earliest priority date rather than publication date. Then it does the part that makes the counts usable. For every year in the window it computes how much of that year has passed its publication date, and reports the observed count, the eligible fraction and the corrected figure in separate columns. Incomplete years never enter a trend line, and the correction is shown rather than applied silently.
Every page on page one for this query treats a filing count as a count. It is not. Under section 122 of the Patent Act an application publishes promptly after eighteen months from the earliest filing date for which a benefit is sought. The USPTO's manual adds the second clock: the projected publication date is the later of eighteen months or fourteen weeks from the filing receipt. The real blind window is twenty-one months.
Built for the competitive intelligence analyst asked where a rival is investing, the R&D manager scoping around someone else's estate, and the corporate development team pricing a technical position. The filing picture belongs inside a competitive intelligence report rather than standing on its own. Read it beside what a filing admits year over year and an analyst report with its claims typed by basis, because a patent count agrees with neither by default.
A sixty per cent collapse that was actually flat
Run a landscape on 1 March 2026 across three assignees in automated storage. Published filings by earliest priority year come back 22 for 2019, 26 for 2020, 31 for 2021, 35 for 2022, 33 for 2023, 14 for 2024 and none for 2025. Read straight off the sheet that is a collapse: 35 down to 14 is a fall of 60 per cent, and it is the sentence the memo writes itself.
Now apply the window. Eighteen months back from the run date lands on 1 September 2024, and fourteen weeks before that is 26 May 2024. Only priority dates on or before that day have reached their projected publication date. For 2024 that is 147 of 366 days, 40.2 per cent of the year. Dividing 14 by 0.402 gives 34.9, which is 20.9 filings the record cannot yet contain. Against 2022 the corrected change is minus 0.4 per cent, not minus 60.
The correction only fixes timing. Three categories of filing never appear at all, and the statute says so. A provisional application is not published. Neither is an application abandoned before its publication date, which the manual describes the Office actively removing from the process. Neither is one filed with a nonpublication request, available where the applicant certifies the invention will not be filed abroad. A rival that files provisionally and lets them lapse leaves no trace to count.
How it works
Name the assignees
The companies or the field, plus the decision the filing picture is meant to inform.
Pull and cluster
Published filings grouped into technology clusters, counted by earliest priority date and not by publication.
Correct the window
Each year's eligible fraction computed from the run date, with the corrected count beside the observed one.
State the limits
What the pattern supports, the filings the record cannot hold, and the years no trend covers.
What you get
- Filings counted by earliest priority date, because publication date measures the office rather than the programme
- Observed count, eligible fraction and corrected figure in three columns for every year
- Trend lines drawn only across years the publication window can actually contain
- The three statutory categories of filing the published record can never show you
- Assignee names reconciled across acquisitions and subsidiaries, so a programme does not disappear
- What the pattern supports, what it does not, and the years it cannot speak for
Common questions
Why count by priority date rather than publication date?
Because publication date measures the patent office's queue and priority date measures when the work was done. A landscape counted on publication dates shifts every programme forward by a year and a half and compresses the differences between them. The cost is that the most recent years look empty, which is exactly what the window correction exists to state rather than hide.
Can I trust the corrected number for a partial year?
Treat it as a floor with a stated assumption, not a measurement. The correction assumes filings are spread evenly across the year, which is roughly true across a portfolio and badly wrong for a single programme that files in bursts. The run prints the observed count next to it so a reader can disagree with the assumption without discarding the data.
What if a competitor files in Europe first?
Then the earliest priority date is the foreign one, and a US-only search will miss the family until a US member publishes. The run works from the priority date wherever it was filed and says which offices it could and could not reach. A landscape built on one jurisdiction is a jurisdiction study, and it gets labelled that way.
Does a rising filing count mean they are winning?
No, and the document says so explicitly. Filing volume tracks legal budget and internal incentives at least as strongly as technical progress, and a programme can file heavily around a product it later cancels. What the count supports is a claim about where attention went two years ago. Where the real question is how big the opportunity is, a market size built in both directions is the better start.
How does it handle an acquisition?
By reconciling assignee names before counting. A startup bought in 2023 may keep filing under its own name, may assign the estate to the parent, or may do both across different families. The run lists the name variants it merged and the ones it left separate, so a reader can see whether a programme jumped or the label did.
Is this a freedom-to-operate opinion?
No. This is a picture of where filings cluster and what that can and cannot support about direction. Whether a specific product infringes a specific claim is a legal question that needs claim construction by a patent attorney, and the document says that in place of implying otherwise. Use the landscape to decide what to send them.
What do I get back?
A Sheet of filings with assignee, classification, priority date and technology cluster, plus the per-year table holding observed count, eligible fraction and corrected figure. A Doc reading the pattern, naming the years it will not draw a trend across, and listing what the published record structurally cannot contain.
Patent Landscape Analysis and Filing Trends
Fill in the form and your workspace opens with the work already underway.