How to read a patent landscape report

Read a patent landscape report in the order it was built: the scope first, then the technology clusters, the assignee rankings, the filing trends, the citation network and, last, any white space flags. Each section depends on the one before it. If the scope missed part of the field or the company names were never merged, every chart after that point inherits the error, however clean it looks.

Start with the scope

Every patent landscape starts from a definition of the field: a seed patent, a technology description, a keyword query, a set of classification codes, or a mix of these. The rest of the report reflects whatever that definition included and whatever it missed. Find these details in the method section, and ask for them if they are not there:

• The seed or query. A report built from one seed patent reflects that patent's vocabulary and citations. A report built from classification codes reflects how examiners filed documents, which may not match how your business describes the technology.
• Date range and date basis. Priority date, filing date, publication date and grant date give different pictures. Know which one each chart uses.
• Jurisdictions and document types. Applications and grants together, or grants only. US only, or worldwide.
• Counting unit. Counting individual documents inflates any company that files the same invention in many countries. Counting families comes closer to counting inventions.

A quick test of scope quality: pull twenty documents at random from the set and read their first claims. If more than a few are off-topic, every count in the report carries that noise. Then write down the companies you already know are active in the field. If one is missing, the query probably used vocabulary that company does not use.

Technology clusters: how to check one

Clusters group the documents into technical themes. Some reports build them from classification codes; others group documents by the similarity of their text. The Cooperative Patent Classification (CPC), launched jointly by the EPO and the USPTO and in use since January 2013, is the usual basis for code-based grouping.

A cluster label is a summary written after the grouping was done. Treat it as a statement to verify:

1. Read five documents from the center of the cluster and five from its edges. The center should match the label. The edges show where the boundary was drawn.
2. Check the size. A cluster of three documents is usually a fragment, not a theme. A cluster holding half the set is usually too coarse to tell you anything.
3. Look at the assignee mix inside the cluster. A cluster dominated by one company may be one family counted many times over, or a genuine stronghold. Counting by family tells you which.
4. Be wary of labels that are only a classification code or a generic phrase such as "data processing". They usually mean the grouping found no clear theme.

Assignee rankings: fix the names before reading the numbers

Assignee rankings are the section most often wrong in a way that changes decisions. The same company appears under many names, and a ranking that does not merge them splits one large holder into several mid-sized ones.

Two further checks matter. First, find out whether the report ranks by the original assignee printed on the patent or by the current owner after recorded assignments. Portfolios change hands, and a ranking by original assignee can show a company that sold its patents years ago. Second, look at how patents owned by individual inventors were handled. A long list of individual owners near the top of a ranking usually means the assignment data is incomplete.

Filing trends over time and by jurisdiction

Filing charts are where readers draw the strongest conclusions, and the most common error concerns the last few years. US, PCT and European applications are generally published about 18 months after their earliest priority date, so a chart that counts by priority or filing year always shows its most recent years lower than they will turn out to be. A drop in the last two or three years of a filing chart is expected even in a growing field. Do not read it as a slowdown until you have corrected for it.

Check two more things on the time axis. Grant-based charts trail application-based charts by years, so they say little about current activity. And a sudden step in an older year can reflect a change in how an office published or classified documents rather than a change in inventive activity.

Jurisdiction charts show where applicants sought protection, which usually tracks where they sell, manufacture or expect competition. Read them together with the counting unit: if the report counts documents, a family filed in ten offices appears ten times. A sharp rise in one office can reflect a single large applicant entering that market rather than broad interest.

Citation network: central and peripheral patents

The citation network shows which documents cite which. Patents with many forward citations, or patents that connect otherwise separate clusters, sit near the center. They are often the foundational filings in a field and the ones most likely to come up as prior art against later patents.

Correct for three distortions before ranking anything as central:

• Age. Older patents have had more time to collect citations, so they look more central than newer ones of equal importance.
• Self-citation. A company that cites its own earlier filings builds a dense knot around itself. Separate self-citations from citations by others.
• Citation source. Examiner citations and applicant citations come from different processes. Check whether the report distinguishes them.

A peripheral position is not a sign of low value. A recent patent in a new direction simply has not had time to be cited.

White space flags

Many reports flag low-density areas as potential white space. Treat each flag as a hypothesis. An area can look empty because the query missed it, because the documents sit under a classification code the scope excluded, because the filings are too recent to be published, or because the subject matter is hard to patent at all. Before acting on a flag, run a targeted search on the gap itself, including non-patent literature and older terminology. The methods in Prior-art search beyond keywords apply directly, because a real gap should survive a search designed to fill it.

A checklist before you act on a patent landscape

1. What defined the scope, and who chose it?
2. Which date does each chart use, and which years are still incomplete?
3. Are the counts by document or by family?
4. Did a random sample of twenty documents hold up as on-topic?
5. Are any companies you know in the field missing?
6. Do the cluster labels match the documents at the center and at the edges?
7. Were assignee names merged across spelling variants, subsidiaries and acquisitions?
8. Does the ranking use the original assignee or the current owner?
9. Are the central patents central because of influence, or because of age and self-citation?
10. Has each white space flag been tested with a targeted search?

If a report cannot answer the first three questions, the rest of it is hard to rely on. If it can, the remaining seven usually take an afternoon, and they are the difference between a chart and a finding.

Where Technology Landscape fits

One Transform's Technology Landscape search produces this kind of report for a field you define. It groups the patents into technology clusters, ranks assignees, charts filing activity over time and across jurisdictions, shows a citation network and flags low-density areas as potential white space. Every conclusion is tied to its source documents, so each check in the list above can be run against the evidence rather than against a summary. Searches run in the background and results stream in as they are found.

Next step

Choose a field you already know well, run a Technology Landscape search on it, and work through the checklist. Where the report agrees with what you know, you have calibrated it; where it disagrees, you have found either a gap in your knowledge or a gap in the scope. 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.