Choose what you are counting before removing duplicates
A search can show several documents that lead back to related filings. That is useful evidence, but it is easy to turn a long result list into an inflated story about how many inventions exist. Before you tidy the list, write the question your count must answer. Are you estimating how many technical groups need reading, tracking applications, or identifying particular publications for a review?
The UK Intellectual Property Office's patent-data handbook explains that family grouping helps avoid double counting and that the definition must fit the analysis. Its family discussion remains useful background, although the handbook was updated in 2015. Use the current database record for the actual members rather than copying an old screenshot's total.
| Your immediate question | Useful working unit | Information to retain underneath |
|---|---|---|
| How many records did my search retrieve? | Search rows, with exact duplicates marked | Original query, filters, date and export |
| How many documents need to be cited or compared? | Distinct publications, including kind codes | Application links and exact version read |
| How many related technical groups should I review first? | Families under one named definition | Every member and unresolved difference |
| Can I make or sell this product? | A separate legal review of the actual product and territories | Relevant claims, official status and professional analysis |
Keep a raw sheet and a summary sheet. The first preserves the evidence; the second helps you navigate it. Grouping should make the work easier to inspect, not erase documents that might matter later. This guide supplies a research method, not a patentability or freedom-to-operate opinion.
Record the family definition, not just the word family
The EPO's family overview describes related applications linked through priority claims. A priority claim refers to an earlier filing used as a basis for the later application. For this worksheet, you are recording the database's stated link, not deciding whether that priority claim is legally valid. The EPO also warns that different databases can produce different memberships.
A DOCDB simple family groups applications considered to cover the same technical content. The EPO applies specific rules when handling priorities, continuations and divisions. Do not recreate that classification by matching a title or keeping only the earliest date: the published priority list is not a complete explanation of the grouping algorithm.
An INPADOC extended family follows a broader network of direct or indirect priority links. Its members can cover similar rather than identical technical content. Use it to discover related branches, then inspect what those branches contain. A broad network is not a count of interchangeable documents.
Write a method line above your table, such as “EPO DOCDB simple families, retrieved October 20, 2026; publications retained in a separate sheet.” If a service supplies its own family definition, name that service and definition instead. Never merge totals from two definitions under an unexplained heading such as “patents found.”
Start with one traceable publication
Copy the complete identifier from the document you actually opened. Keep the office prefix, number and kind code, the suffix that identifies a publication type. Also record its application number and publication date in separate fields. An identifier stripped of its suffix can make two versions look like a duplicate row when they are distinct publications.
Open the record's family view and save the source URL and retrieval date. In the European Patent Register, the EP Patent family help explains the labels: E means equivalent in a simple family, M indicates a related family member, and P and D describe parent and divisional relationships. Read the displayed relationship rather than treating every row as another independent invention.
Choose a starting document because it is relevant and readable. Keep its role explicit: “used for initial technical reading,” for example. That choice does not mean it is the first disclosure, the newest claim version or the document relevant to your target market. Those are separate questions.
Worked public example: EP1000000 and its related publications
On October 20, 2026, we inspected the Register record for EP1000000, titled “Apparatus for manufacturing green bricks for the brick manufacturing industry.” It lists European application 99203729.1, filed November 8, 1999, and priority NL19981010536 dated November 12, 1998. The record also shows the original priority format NL 1010536.
The following is a selected excerpt from its EP Patent family view, not a complete worldwide inventory or a current legal-status assessment. The view labels the displayed groups as equivalents.
| Publication selected | Publication date shown | What to put in the worksheet |
|---|---|---|
| EP1000000 A1 | May 17, 2000 | European application publication; retain as its own document. |
| EP1000000 B1 | February 12, 2003 | Patent specification on the same European application record. |
| US6093011 A | July 25, 2000 | A separate US publication listed as an equivalent; preserve its identity. |
These three selected publications can occupy one family summary entry while remaining three rows in the document sheet. The two EP rows do not become two unrelated inventions because their dates differ. The US row should not disappear merely because you already read an English-language EP document. The grouping explains a relationship; it does not establish matching claim scope.
To repeat the exercise, search the Register for EP1000000, open “EP About this file,” then “EP Patent family.” Compare the publication entries and priority numbers. Keep the kind code as a separate column if the interface separates it from the number. If your view differs from this dated example, record the difference rather than forcing the current data to match this table.
Build a summary that never replaces the member list
Use a local label such as F-001 for your own notes, but do not present it as an official database family identifier. Where the database exposes its identifier, retain that in another column. A local label lets your research discussion stay readable while the source identifiers remain available for checking.
| Field | What to save | Why it helps |
|---|---|---|
| Local group and source definition | F-001; database; simple, extended or named provider family | Makes the grouping rule visible. |
| Source family identifier | Exact displayed ID, or “not exposed in this view” | Avoids inventing an official identifier. |
| Document identity | Publication prefix, number, kind code and application number | Separates documents from applications. |
| Date fields | Priority, filing and publication dates, each labeled | Prevents one date from silently replacing another. |
| Relationship evidence | Displayed relationship, priority links and source URL | Explains why the member was grouped. |
| Reading evidence | Exact publication, page, paragraph, claim or figure reviewed | Keeps a technical observation tied to a document. |
| Open question | Difference to inspect, missing record or status not checked | Gives the next reader a useful action. |
Keep your inclusion rule nearby. For example, a mechanism review may include a family only after at least one document shows the feature of interest. Merely sharing a family with a matching result is a reason to look, not a finding that every member contains that feature. Mark unread members as unread.
Reconcile the totals before writing a conclusion
Here is a hypothetical spreadsheet audit, independent of the real EP example. An export contains eight rows. One row repeats the exact same publication identifier, leaving seven distinct publications. Two publication pairs belong to two applications, while three other publications each belong to another application. That gives five applications. The database places those five applications in two simple families.
Your summary can therefore say: “Eight exported rows; seven distinct publications; five applications; two simple families in this checked set.” Each number answers a different question. Calling the export “eight inventions” would be unsupported, and replacing the seven-document sheet with two representative PDFs would throw away useful evidence.
If a member has no family assignment, put it in an unresolved list. Do not silently drop it or join it to the closest-looking title. Check the identifier, source record and relationship evidence. The audit should account for every original row through inclusion, an exact-duplicate marker, an explained exclusion or an unresolved entry.
A manageable report might include two family summaries and a seven-row appendix. You can then read across groups without repeatedly counting publication stages. This is a navigation aid for your selected results, not a measurement of every invention in the field or the value of a competitor's portfolio.
Keep representatives, dates and legal status separate
A displayed family representative can change. LexisNexis documents how PatentSight selects a representative and gives an example in which a later EP publication replaces a US publication as the displayed representative. Save the exact document you cited, not just the current top result for the family.
Keep priority, filing and publication dates in their own columns. An earlier priority claim is not evidence that the document you downloaded was publicly available on that date. When timing matters, identify the particular disclosure and its publication evidence. Do not assign one convenient date to every paragraph of every related document.
Likewise, a country appearing in a family list is not a verified statement that enforceable rights exist there today. WIPO's patent FAQ explains the territorial nature of patents and the absence of a universal world patent. Check the relevant official record for the specific right and date; obtain qualified advice before using the result to make a filing or commercial decision.
For practical reading, use our claim-comparison worksheet. For launch questions, keep the distinction between patentability and freedom to operate attached to the research packet. Family grouping answers neither legal question by itself.
Finish with a reproducible handoff
Save the query, filters, retrieval date, database, family definition, raw results and your grouped summary together. Add a short note explaining the representative you read and the unresolved members. Another person should be able to move from a summary row back to the exact publication without guessing what you meant by “the patent.”
When a later search changes the count, compare identifiers before declaring that new inventions appeared. A newly retrieved publication, a different filter or a changed grouping may explain the difference. Record the reason you can establish and leave the rest open. That keeps the next research step focused on evidence instead of an impressive but ambiguous total.
