Patent Search: Start From the Invention Point, Not the Keywords

In my 20 years of practice as a patent attorney and litigator, I have reviewed a great many searches done by clients, by junior colleagues and by other firms. The searches that fail rarely fail because the searcher used the wrong database. They fail because the searcher started in the wrong place.

This article explains the routine I use myself, the mistake I see most often, and one experience that convinced me to stop trusting keyword-first searching.

What is the most common mistake in patent searching?

The mistake I see most often from newcomers is searching only the current technology within their own industry. They cannot correctly distinguish between their industry and the technical field that patent law cares about.

Your industry is a commercial category: who your customers are, who your competitors are, which trade shows you attend. A technical field, for patent purposes, is defined by the technical problem and the means used to solve it. Those two maps overlap, but they are not the same map.

This matters because prior art is not limited to your industry. Under European practice, for example, the problem-and-solution approach asks what a skilled person would do when faced with the objective technical problem, and that person may look into neighbouring fields where the same problem has been solved. US obviousness analysis similarly considers analogous art that is reasonably pertinent to the problem. China's examination guidelines follow the same logic. A search that stays inside your industry leaves exactly those documents unfound — and those are the ones an examiner, or an opponent in an invalidation action, will use.

Search the technical problem, not your industry. If a competitor's engineers would never read the document but an engineer facing the same technical problem would, it is still prior art against you.

The following is an illustrative example, not a real client case. Suppose a company makes kitchen appliances and has developed a way to stop steam from condensing on a sensor window. If it only searches kitchen appliance patents, it may conclude nobody has done this. But "keeping an optical window free of condensation" is a problem that people working on cameras, vehicle sensors and medical instruments have also faced. So the search should not stop at the company's own industry; guided by the specific technical problem and the technical means, it should extend to other fields that may offer relevant solutions.

What steps should a patent search follow?

My own routine is fixed, and it has three steps.

1. Understand the technology. Before typing anything into a database, I make sure I understand how the product or process actually works: what the components do, what the steps are, and what problem the engineers were trying to solve. If I cannot explain it in plain words, I am not ready to search.

2. Find the invention point. The invention point is the specific feature or combination that differs from what already exists and produces the technical effect. It is not the product as a whole and it is not the marketing description. Most of a product is usually conventional; the invention point is the part that is not.

3. Combine the invention point's keywords with classification codes. I build the search query from two kinds of material: keywords that describe the invention point, and the classification codes (IPC or CPC) where documents about that technical means are likely to be filed. Keywords alone depend on how an author chose to word the document. Classification codes alone return too much. Combined, they narrow the search to the right place while tolerating differences in wording.

Then I read the results against the invention point, not against the product. A document that looks nothing like the client's product can still disclose the invention point, and a document that looks very similar can be irrelevant because it lacks it.

Why does keyword searching miss relevant prior art?

I learned this lesson clearly on one matter. Both the client and our own team of patent agents had searched by keywords, and nobody could find any close prior art. That result was suspicious rather than reassuring.

When we looked again, we realised the keywords had several possible choices — the same technical idea could be expressed in more than one way, and the searches had followed only some of them. So I went back to my usual habit: focus on the invention point and build the search around it. Once we did that, we found the documents that were actually relevant.

The point is not that the team was careless. The point is that a keyword-first search is only as good as the vocabulary the searcher happens to think of. Documents are written by many different authors, in different eras, sometimes translated from other languages. Anchoring on the invention point forces you to describe the technical substance in several ways and to use classification codes, which do not depend on wording at all.

"We searched and found nothing close" is a warning sign, not a conclusion. Go back to the invention point, list the alternative ways it could be described, and add classification codes before you trust an empty result.

How is a prior art search different from a freedom-to-operate search?

People often ask for "a patent search" without saying what decision it will support. The purpose changes how the documents are read.

• Novelty and inventiveness search: Is the invention point already disclosed, or would it be an obvious step from what is disclosed? Any public document anywhere in the world can count, whether or not it is a patent and whether or not it is still in force.
• Freedom-to-operate (FTO) search: Does the product fall within the claims of a patent that is in force in the market where you plan to sell? Here only live patents in the relevant jurisdictions matter, and what counts is the claims, not the description.

The first two steps of my routine — understand the technology and find the key features — apply to both. But an FTO conclusion requires mapping each product feature against each claim element, which is a legal judgement. This is not legal advice; consult a qualified patent attorney for your situation.

When should you search, and who should do it?

Search early. The best time is before a project is committed, when the result can still change the design or the decision to file. A search done after the product is finished can only tell you about problems; a search done at the start can help you avoid them.

Engineers should take part. They understand the technology better than anyone, which is the first step of the routine. A patent professional brings the second and third steps: separating the invention point from the conventional parts and knowing how documents are classified and how examiners read them. It makes sense for the two to work together.

Whatever tools you use, read the actual documents. A list of titles is not a search result; a document read against your invention point is.