Accuracy and Speed in Patent Infringement Searching Need Not Be “One or the Other”

“Faster and better results at lower cost!”

A claim so common it blends in with the hum of office chatter. We’d like to believe that patent searches could be turned around within a day or two, but it typically takes much longer.  We’d like to believe that the work returned to us includes everything we need to know.  But, at some level, we sense there are shortcomings.

This is particularly true in the case of freedom-to-operate investigations (FTO).  Searchers often fail to consider certain product features.  Claims are not always afforded their full scope.  And sometimes key patents are just plain missed.  Despite relying on highly competent and experienced analysts, “faster, better and cheaper” just does not seem a reality.

But a departure from conventional methods can overcome the trade-off between speed and accuracy. We can achieve both.  Clearstone Elements is that departure.

Why is there a speed/accuracy trade-off when using conventional search methods?

Imagine for a moment that you have been asked to carry out an FTO analysis.  Let’s also say that, as is often the case, you must complete this analysis under budget, say $3,500.  Assume a billing rate of $75/hour (you have overhead and a boss) and that you’re able to review a patent every 6 minutes.  That means you only have budget to manually review about 470 patents.

This might seem like a reasonably-sized lot.  However, experienced FTO searchers would likely feel that it’s not nearly enough to guarantee a high level of recall (i.e., the ratio of the number of relevant patents retrieved to the total number of relevant patents), say of about 95%.  This will of course vary case-by-case.  But, for many competitive industries, you’d need a starting set in the range of about 2,500-5,000 patents to achieve a sufficiently high recall.

Why must so many patents be manually reviewed in FTO?

Certainly, products don’t typically infringe thousands of patents.  Else, industries would come to a grinding halt.  Indeed, out of a robust set of patents to be reviewed, a searcher is likely to only uncover a handful of relevant ones.

The short answer to the above question is that it’s very difficult for a searcher to anticipate whether a patent is relevant without actually reading and understanding its legal claims.  Products may be described in countless ways, many of which will go unnoticed to a searcher.  Claims of a patent may be written in much broader terms than used to describe the specific embodiments of the patent’s technical disclosure.  For these and other reasons (see here for more information), conventional search tools are imprecise blunt instruments in the FTO context.  Thus, a broad net must be cast.

It should now be obvious why inaccuracy seeps in.

A searcher will typically approach a search with the best of intentions, considering every angle of attack.  Patent classes will be reviewed, important ones selected and queried.  Key assignees will be queried.  Citation, clustering, and family analysis will be conducted on patents flagged as relevant.  Natural language queries, perhaps using semantic analysis, will also be applied relating to key product features.

These tools are all great at robustly representing technology areas.  However, most likely, this aggregation will far exceed, in number, what could be reviewed under budget in an FTO context.  Accordingly, the searcher will necessarily have to take additional steps to downsize this set.  And this is where it goes awry.  The searcher may do as follows:

  1. limit all results by requiring the inclusion of certain key terms;
  2. exclude unknown or small-scale assignees;
  3. exclude patents based on Title, Abstract and/or Drawings; and/or
  4. limit patents by relevancy score cut-off (if relevancy scores are provided).

These processes are generally arbitrary in nature and bare little correlation with relevance.  They are simply not sound bases for reducing a large aggregation of patents in an FTO search.  Inevitably, they cause omission of critical patents.

How does Clearstone Elements overcome this dilemma?

Clearstone Elements is a fundamentally unique search platform.  In Clearstone Elements, associations between patents and technical attributes are memorialized, forming specialized data files called workspaces.  Notably, these technical attributes are associated with patents on the basis of their claims and using human analysis.  These attributes (or elements) are then presented to a searcher as an interactive taxonomy.  A searcher may than effortlessly eliminate swaths of patents from a robust patent set by selecting elements not embodied by the product undergoing search (more on this process here).  In contrast to the arbitrary procedures discussed above, the Clearstone process is objective, reliable, and deliberate.

In fact, current users of Clearstone Elements are typically able to reduce a robust patent set by 90-95%.  This means that, once a workspace is put in place, provided the same constraints as in the above example, a searcher using Clearstone Elements is actually able to objectively consider upwards of 4,700-9,400 patents.  We are essentially empowering searchers to cast significantly broader nets with significantly reduced manual effort.  By replacing the conventional tools that are blunt and arbitrary with Clearstone Element’s logic-based reduction process, “faster, cheaper and more accurate” is no longer an empty promise.