I often hear that a patent search won't take long, or it will be a small search, or get asked to do a quick search, and in some cases that's probably right, but in other cases, by which I mean most, things are not always as small as they appear. I'm going to risk expulsion from some secret society by pulling back the curtain and revealing some of the black magic that occurs when someone asks for a patent search.

Why do people think a search will be small or not take long?

Perhaps they only know of three or four other companies that do what they do, or that the idea is a relatively simple concept, or that you only need to look for the doohickey part of the whole thingamajig, or that whenever they've done a search they only get a few results to look at, or maybe they just don't understand the scale of what constitutes patent literature.

Two estimates, made in the last three or four years, of the total number of patent applications filed put the figure at about 90 million, and between 80 and 100 million. If you want something a little more concrete, WIPO says that there are just over 56 million patent applications up to the end of 2018, which is close, but they've only been keeping count since 1985, so it's definitely an underestimate given there are a couple of hundred years of applications they haven't counted. This number is set to rise, and rise exponentially. To single out the highest filing country of recent times, China became the first country to file over one million patent applications in a calendar year, which happened in 2015, and just three years later they have increased that by half again.

The exact number, or even a credible estimate, doesn't matter. It's a big number, and there's no way any one of us can look through them all. The thing is though, we have to, and when it comes to your idea, and I mean Your. Actual. Personal. Idea., at the point you say "Can you do a quick search on...", every one of those 100 or so million patent applications is relevant, and they remain relevant until I can find a reason to exclude them. What's left after I find those reasons are your search results. That's going to be a relatively tiny number, and that's what makes a patent search look small or quick.

So how can you exclude millions of potential search results?

The fastest way to get a more manageable set of results is to apply a patent classification code. A patent classification code is assigned to a patent application when it is filed, and then further refined upon examination of that application. You're relying on people with experience in a subject matter field to say that this patent application belongs in this patent class, so when you apply that patent class in your search, your reason for excluding some of those 100 million patent applications is that other people­—experts—say they aren't relevant.

If we take one patent classification code relating to pharmaceutical compositions containing a chemical compound as the active ingredient, and apply it at its broadest level, A61K31, to that 100 million, we already bring ourselves down to fewer than 3 million patent applications. We've already excluded 97% by broadly applying a patent classification.

Patent classification systems are something for another day (but if you're keen here's the IPC and the CPC) so I won't go too deep, but if you're looking for something common such as esomeprazole (Nexium), the relevant classification is A61K31/4439, which brings us down to fewer than 70 thousand patent applications, so you have excluded over 99.9% of all patent applications.

Keywords can also be useful tools for excluding patent applications you don't need to consider. Some aren't so useful, such as the brand name for midazolam, which is Versed. Versed is also a word used in patent applications, commonly found at the end of a specification in a sentence that includes the words "... those versed in the art will appreciate ..." that indicates subject matter experts will understand that some minor differences in the invention are obvious, but that doesn't assist anyone in a search for midazolam by increasing the number of search results by tens of thousands.

Even if you have a keyword that will be useful and restrict your results to a manageable level, you still have to be aware of variations on that keyword. The English language is very colourful and allows for a wide range of potential descriptions for a single object.

Consider a surfboard. It's a well known object and you would be hard pressed to find someone who described it otherwise when asked. I've had a quick look and I can find reference to a surfing board, a watercraft board, a water sports board, a surfing device, and of course, a surf board (with the space between the words). In terms of how much that restricts your search, 'surfboard' provides 15 thousand results while 'surf board' provides 80 thousand.

What's even more useful than a single keyword is a combination of two or more. We can take an object (the surfboard) and combine it with an essential feature (and apologies to anyone whose hopes and dreams I'm just about to destroy, a battery). Combining these two sets of keywords so that we are locating them essentially within the same paragraph in a patent specification, i.e. (surfboard OR surf board) NEAR (battery) we can whittle down the 100 million patent applications to just 300 or so, which is eminently searchable.

The last method of reducing the number of search results you have to consider relates to the type of search you are undertaking. It's really a novelty search or a validity search that has to consider all 100 million patent applications as relevant at the start of a search. An infringement search or freedom to operate search has features that automatically reduce the overall numbers to consider, and that's before you start to think about patent classifications or keywords. These searches are restricted to a single country, so with the pharmaceutical class A61K31 above, we have already reduced that number by 75%. These searches are also restricted by filing date to patent applications that have a live status, which in most cases is within the last 20 years, or the last 25 years for pharmaceuticals. To continue our pharmaceutical example we have removed a further 50%, and that's before we even consider what the actual status is. If you were to look through Australian patent applications in the broad pharmaceutical class A61K31 that have a live status, you would be looking at just over 20 thousand records, which is a lot less than the 3 million we found above.

With most of these examples the numbers of search results are still enormous and I'm not saying you would search them but it shows you how the application of a few parameters, alone or in combination, can substantially reduce them to a manageable level, and make that 'quick' search, well, quick.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.

Shelston IP ranked one of Australia's leading Intellectual Property firms in 2015.