Addressing the Proportionality Challenge in Discovery Starts at the Beginning

Written by Doug Austin, Editor of eDiscovery Today

I’m sure your response to the title is “no kidding, Captain Obvious!” So, let me explain what I mean by that.

On this blog, I have previously discussed how Big Data is impacting eDiscovery workflows and what that means for you.

In today’s blog I will discuss another sign of how Big Data is impacting eDiscovery workflows – the rise in proportionality disputes.

The Dramatic Increase in Proportionality Disputes

According to the 2020 eDiscovery Case Law Year in Review Report released earlier this year by eDiscovery Assistant and my blog eDiscovery Today (and available for download from here), the number of eDiscovery case law rulings that involved proportionality disputes has risen dramatically in recent years, from 90 in 2012 to 889 in 2020! That’s almost a 10-fold increase in eight years!

The number of proportionality disputes rose from 476 in 2019, reflecting an increase of 86.8% in a single year.

There were even more proportionality disputes than there were sanctions disputes last year (851), so proportionality disputes have become quite common as organizations try to apply a proportional approach to discovery to address the Big Data challenge while they encounter push back from requesting parties.

Addressing the Proportionality Challenge Downstream

Not surprisingly, the first place that organizations attempted to address the proportionality challenge was in document review, which has been historically the largest percentage of eDiscovery costs – by far. According to Rob Robinson’s ComplexDiscovery site, “[s]pending on review-related software and services is estimated to constitute approximately 68% of worldwide eDiscovery software and services spending in 2020”. And that’s actually down from 2014, when it was 73%.

So, the original focus to make eDiscovery more proportional logically began in review, and the use of artificial intelligence (AI) and machine learning technologies to perform predictive coding to streamline the review process and reduce review costs has become a common approach to make eDiscovery more proportional for organizations.

And it works – to keep review costs more manageable. But, while I’m certainly a big proponent of predictive coding approaches and technologies, predictive coding falls short when it comes to the reduction of expenses in other phases of discovery, such as preservation, collection and processing.

The approach for a lot of organizations that have established predictive coding workflows has continued to be preserve, collect and process broadly, then apply the technology to reduce review expenses. Predictive coding applied to downstream review does nothing to reduce expenses for upstream processes.

Addressing the Proportionality Challenge Upstream

But what if you apply the same technologies and workflows upstream and review earlier – before preservation and collection?

Then, the process of streamlining starts earlier and extends to preservation, collection and processing, as well as review.

By applying AI and machine learning technologies further upstream, organizations can avoid the “custodian dump” of data into the discovery process, where an entire custodian’s corpus of data is preserved, collected and processed – just to determine that a large majority of that data is non-responsive to the case.

Why make that determination after you’ve preserved, collected and processed it all when you can do it sooner and only push the data downstream that is potentially responsive? Applying AI and machine learning technologies further upstream streamlines and saves costs throughout the entire EDRM life cycle, not just review.


So, when I use a seemingly obvious blog post title like “Addressing the Proportionality Challenge in Discovery Starts at the Beginning” to illustrate the best place to stem the dramatic rise in proportionality disputes, the title is not as obvious as it seems – at least historically. Most eDiscovery project teams have not started at the beginning to address proportionality; instead, they’ve started in the middle. That is no longer soon enough.

As Sarai Schubert, COO of IPRO, states in this blog post: “Using AI to validate these upstream review processes can have a hugely beneficial effect on downstream eDiscovery processes.”

AI and machine learning can provide benefits in multiple phases of the EDRM life cycle, but it provides the most benefits the further upstream you go. Think upstream! For more educational topics from me related to eDiscovery, cybersecurity and data privacy, feel free to follow my blog, eDiscovery Today