Introduction

If you search the SEC’s EDGAR database – the largest publicly available contract database in the world and a common source of inspiration for many drafters – for the keyword “confidentiality”, you are met with hundreds of thousands of clauses. If you do the same in your organisation’s document management system, you’ll likely come across at least a few hundred. 

With no way to search for the specific type of confidentiality clause (in whose favour is it drafted? For what domain of law? How aggressive or lenient are the obligations?), lawyers only have one option: start at the top and keep crawling through the list of hits until they find something that resembles what they need, then start tweaking.

The keyword search dilemma

In a way, Google has spoiled us. We are so accustomed to searching for something and then finding it immediately, that we have come to expect it from any exercise that involves keyword searches. Just consider: when was the last time you visited the second page of your Google search results?

Unfortunately, this standard cannot be maintained in a domain as nuanced and as diverse as legal drafting. Keyword searches are always going to bring a lot of noise. But more importantly, when you search for “confidentiality clause”, you are likely not looking for any old confidentiality clause. In reality, you probably need something like “a mutual confidentiality clause for a share purchase agreement where the parties agree on (1) the definition of “confidential information”, (2) obligations of the parties, (3) exceptions to those obligations, (4) damages in case of breach, (5) duration of the confidentiality obligations,…”.

Capturing that kind of nuance in a few keywords is tricky indeed.

For keyword searches to actually help you find relevant material, a keyword is ideally found multiple times within the same clause, but at the same time not too many times (relatively speaking) in all the other documents. This is the reason why a search for "escrow" or "bankruptcy", or combinations such as "garden leave" or "material adverse effect" will work fairly well. Each of them will occur multiple times in your ideal clause, but not too frequently in all the other clauses. Conversely, this is also the reason why a search for "confidentiality", “intellectual property”, “liability”, etc. will lead to very poor results.

In other words, unlike typical web pages where there an enormous variety of words are being used, contract clauses have the disadvantage that the same set of roughly 200 keywords are being used over and over again.

Don’t get me wrong, though. Lawyers need this. Without terminological consistency, contracts would be pure chaos. But for information retrieval purposes, this setup makes it very difficult for search engines to immediately offer useful material.

Cosmetic vs legal nuance

More importantly, though, the hundreds (of thousands) of clauses you find with a simple keyword search imply that there must be hundreds of different ways to draft such a clause. Here, we stumble upon an important distinction: cosmetic nuance vs legal nuance.

Naturally, there are millions of ways to write a confidentiality clause from a cosmetic perspective. This has everything to do with word placement, word choice, sentence structure, terminology, etc. From a legal perspective, however, most of those clauses will say the same thing but in a different way. Research suggests that lawyers are notoriously bad at seeing the difference.

In all the time that we have spent building clause libraries (for ourselves when we were still lawyers, and for our clients as vendors), we have found that 3-10 versions of a specific clause (not counting any translations) is the average sweet spot for legal nuance. Obviously, reality is always going to be more unpredictable than you can ever dream up in theory. Those 3-10 versions will never account for the entire range of possible permutations that can come up in negotiations, conversations with clients, etc. Nevertheless, our findings show that those 3-10 clauses provide much more “starting point value” than 3.000-10.000 clauses.

Optimising for findability

Knowing that the focus should be on quality over quantity, we can start to look at how you can manage these clauses in a way that makes it easy for everyone in your team to find the right material instantly (i.e.: not just the people who draft the clauses).

Without going into too much detail (we already do that here), you’ll want to augment these clauses in a way that provides the kind of legal nuance you are looking for, which we described above. Popular examples include:

  • A folder structure: placing your clauses in a folder structure provides an entirely new way of looking at them. They give you the opportunity to zoom out and survey the state of the library. To see what is there and then to zoom back into the relevant location. This can already drastically limit your search results in a way that makes it much easier to find relevant content.
  • Naming clauses: the name of a clause doesn’t (necessarily) equate to the title of that clause. Instead, it is a separate “headline” and gives you an opportunity to convey crucial information right off the bat. The title of a confidentiality clause may be something like “Confidentiality” but the name of that same clause could be (“Confidentiality – short and mutual”)
  • Tags: as more clauses get stored in the same folder, it gets trickier and trickier to find the right clause for the right situation. File names can help because they offer some insight into the content of the clauses without having to read through the entire text. But even that has its limits. Once you start writing file names the size of a tweet or once your library grows to house hundreds of clauses, you should consider “tagging” clauses. These tags contain information on legal nuance that differentiates them from their contemporaries.


For example: In-house legal teams will want to tag clauses with information like “company standard clause” or “fallback version”. Law firms typically identify which party is favoured in a particular clause (e.g. in a share purchase agreement, you will want to clearly indicate whether a clause is purchaser-friendly or seller-friendly). Other popular tags include the length of a clause (typically expressed in a value ranging from 1-5), the favoured party, etc.

The final frontier – dynamic clauses

As legal teams grow in building and maintaining a clause library, it will inevitably become more difficult to manage many versions of the same type of clauses. Those 3-10 clauses may cut it for a while but, as stated earlier, reality is always more unpredictable than theory can dream up. Once the same clause is used a few times as a “starting off point” to perform the same tweaking exercise, lawyers are – rightfully – going to want to add it to the library as a new nuance.

For many types of clauses, it’s perfectly manageable to add a few legally nuanced alternatives here and there. A library that contains 30 versions of a force majeure clause can still be very easy to navigate if the appropriate augmentations as described above are added to the clauses (though it should be asked how many of these versions are legally nuanced and how many are just cosmetically nuanced). 

Inevitably, though, an organisation’s expertise is going to shine through on a select few clauses. Any experienced drafter can point you to the types of clauses that are usually contested in their line of work and therefore require creative, legally nuanced versions. Over time, those versions are then added to the library and distinguished from the content that is already there through the means set out above. But even that approach can break. It differs from situation to situation, but at some point, this burgeoning collection of static clauses becomes less and less manageable. That’s when you need dynamic clauses.

Dynamic clause example – dispute resolution clause

Dynamic clauses are clauses that have been automated using specialized technology (e.g.: Clause9). In a single, flexible clause, they provide all the different kinds of legal nuances you would want from a collection of static clauses.

Take a dispute resolution clause as an example. There are many ways to draft such a clause. For the sake of argument, however, let’s say that there are two main ways in a commercial contract to settle a dispute: litigation and arbitration. In the clause library, these could already be two separate clauses (although it is more likely that there will be multiple versions of each one).

The parties can then also choose to apply or not apply a “cooling-off” period that requires them to find a way to resolve the dispute amongst themselves before moving ahead with litigation or arbitration. This typically takes the form of a mandatory negotiation or meditation process. The parties could even state that they need to go through both negotiation and mediation to arrive finally at a more traditional dispute resolution procedure.   

Finally, in multi-lingual jurisdictions, you may also need two language versions of that same clause.

As a lawyer, it’s easy to assume that all of these combinations put together are fairly limited because you typically tweak and adjust on the spot as necessary. In reality, all of the abovementioned nuances amount to 16 different clauses. For more complex clauses, we have seen this run up to 45+ variations.

dynamic dispute resolution clause
Example of a dynamic dispute resolution clause

Closing thoughts on the quality vs quantity conundrum

We get it. Quality is hard. Quantity is easy.

We added the Clause Hunt to our MS Word add-in ClauseBuddy specifically because it offers a quick and easy way to access your team’s collective drafting history directly from within MS Word. But that kind of access is still akin to searching for a needle in a haystack. Sometimes you get lucky; usually you don’t.

Legal teams that have successfully leveraged clause libraries to improve their drafting process unanimously agree: a large but chaotic pile of clauses may require less work and offer value faster, but it doesn’t offer a whole lot of value. Vice versa: smaller, but more well-curated libraries take a while to construct, but offer a great deal more value.