E-disclosure Headaches: Case Study in Pain Relief

November 21, 2013

Referrals between barristers are commonplace whether for logistical reasons (where one barrister cannot make a hearing) or for tactical reasons (eg where leading counsel is brought in to provide additional experience and skill).  As e-disclosure continues to grow in importance, referrals from one barrister (who is briefed on the substantive litigation) to another barrister (for specialist e-disclosure skills) offer the client (both professional and lay) an opportunity for an enhanced overall service.

This case study explores, within the context of a barrister-to-barrister referral, the tactical advantage that early case assessment of electronic documents combined with deep understanding of the legal issues can bring to commercial litigation. 

Case Study Outline

A client becomes embroiled in litigation.  There is an injunction followed by meaty particulars met with firm denials and, because the claim is on the multi-track, there is going to be an order for e-disclosure.  This was the background to a case that came my way because the barrister dealing with the substantive aspects of the claim felt that I could provide specific expertise on e-disclosure.  He remained the ‘lead’ counsel throughout; I had responsibility for the e-disclosure.

Case Management Conference

After getting to grips with the legal issues in the claim using existing advices, information from both solicitor and barrister and the statements of case, I discussed tactics with the other lawyers in the team.  We decided that I would be briefed for the case management conference because the e-disclosure element was potentially significant.  I drafted a ‘vanilla’ e-disclosure order; it simply provided for discussions followed by a meeting so that the parties could seek to agree their respective e-disclosure proposals in relation to custodians, date ranges and keywords.

In fact, the e-disclosure order was, with minor variations, agreed.  I still attended the CMC but the focus was on various other contested applications.

Early Case Assessment

I sketched out a preliminary list of custodians (the keepers of pertinent electronic documents).  This was discussed with the client and with certain modifications the process of collecting the data from the custodians began.  The client had its own IT department which meant that I could concentrate on the next major task – identifying a tool that I could use to analyse the data once it arrived.

I chose e-Discovery Tools (now known as EDT); it gave me powerful mobile tools installed on my custom-built i7 laptop to analyse the data at an early stage and the support was terrific.  I was able to start crunching the data received from the client (which I call ‘the harvest’) as and when it arrived.

The harvest comprised 90Gb of data (450,000 documents).  At 50 documents per hour that meant 9,000 review hours.  A plan was needed to define the scope and methodology for an e-disclosure exercise that would, in terms of cost, be proportionate to the size of the case (which was roughly £250,000) but also have a sound chance of achieving a ‘reasonable search’ within the meaning of CPR 31.5.

There were certain fairly obvious ways to reduce the data which the e-disclosure order expressly identified such as date ranges and keywords.  Before even applying these processes, we could analyse the impact of de-duplication (ie removing exact duplicates from the document set).  After de-duplication in EDT the volume reduced to 360,000 documents.  In other words, about 20% of the documents were duplicates (from previous experience this was around the percentage I expected).

Whilst it was tempting to discard these duplicates forever (on the basis that they could never be worth reviewing), at this stage I was not seeking to take any permanent steps, I was seeking to understand the document set and the impact of each process that could be applied to reduce the volume.

Meanwhile correspondence was in full swing about the scope of the e-disclosure exercise.  As is often the case, the parties had little dispute about 80% of the exercise (whether that was in terms of the custodians or the keywords or the date range) but attitudes were hardening about the remaining 20%.  Should this custodian be in when she had (on one view not shared with the other side) no connection with the case?  Why should the date range stretch back to 2008 when the breach (on one view not shared with the other side) was two years later?

However, this early correspondence was based not on statistics (eg how many documents might be disclosed if an additional custodian was added) but based on the lawyers’ perceptions (guided by their clients) of where relevant documents might be held; the arguments were initially being put forward in a statistical vacuum.  I felt it was vital to fill that vacuum as soon as possible.

The client’s instinct was firm from the outset: ‘we’ve got hardly anything of relevance; we’re the ones chasing them for their wrongdoing’.  Would the documents back up this instinct?  I was mildly sceptical; long experience had told me that litigants generally underestimate the number of relevant documents often because the concept of disclosing adverse documents runs contrary to general commercial principles (where revealing a weak hand any earlier than absolutely necessary makes little sense).

Correspondence continued, the date range was close to being agreed and applying the likely date range drastically reduced the volume of documents to 120,000.

Now I needed to test the keywords against the 120,000 documents.  My general experience has been that between 10-20% of documents (where the document set has been trimmed by a good set of keywords) are disclosed after review.  My sampling suggested nothing of the sort and supported my client’s instinct; no matter what set of sensible keywords I applied (and I settled on a set that returned 25,000 out of the 120,000 documents), the predicted volume of disclosable documents (using statistical sampling techniques) remained startlingly low, at 0.2%.  In other words, to find one disclosable document, 499 irrelevant documents would also need to be reviewed.  Was this good or bad news?  How could it best be turned to the client’s advantage?

e-disclosure Negotiations

The key next procedural step was a meeting to discuss e-disclosure with the other side.  For the purposes of the meeting with the other side, I produced a comprehensive report on my client’s documents based on my extensive sampling.  That report was written to be as objective and impartial as possible, recognising that the report might form the basis of a witness statement should the parties end up unable to resolve their disagreements without judicial aid.  At the meeting we presented the report to the other side and asked for the equivalent information from them.  We left the meeting quietly confident we could control the e-disclosure battlefield.  The key now was to keep the pressure on whilst ensuring we would not miss the deadline for disclosure.

Reviewing the Documents

We were still faced with 25,000 documents to review.  In a case where the statistics suggested such a low hit-rate, would technology-assisted review offer a way forward?

I had experience of technology-assisted review.  In a previous case we had agreed with the other side that technology-assisted review would be used.  The documents would be ranked by the system based on its ‘learning’ acquired by its processing of the decisions of expert reviewers on a sample (or ‘training’) set of documents.  The ranked documents would then be top-sliced and that top slice (being the 20% that the system judged most likely to be disclosable) would then be reviewed.  The results had been positive in the sense that, from the top slice, roughly 30% of the reviewed documents ended up being disclosed (suggesting that the system was effectively making decisions on the likely relevance of the documents).

With the relatively low volume of documents I was uncertain whether or not technology-assisted review would prove cost-effective because a fairly substantial percentage would need to be reviewed anyway in order to train the system.  Before I explored the value question further, I turned to a simpler solution to reflect the client’s unusual position; we would propose to carry out no further review of electronic documents at all.  We would accompany that proposal by indicating a willingness to listen constructively to requests for specific disclosure.  This model, effectively dispensing with e-disclosure for the claimant, is not as far-fetched as it might seem.  For example, Practice Note SC Eq 11 – Disclosure in the Equity Division sets out the procedure for the Equity Division of the Supreme Court of New South Wales where disclosure in the Equity Division has been curtailed so that:

·        no order for disclosure is made until the parties have served their evidence save in exceptional circumstances;

·        even after service of evidence, no order for disclosure will be made unless it is necessary to resolve real issues in dispute in the proceedings.

In other words, disclosure is no longer the default; it has to be justified before a court will make such an order.

On reflection (concerned that no disclosure at all could be seen as too radical), we would temper our proposal.  Backed up by a report detailing the statistical analysis I had undertaken, we would offer only to review 5,000 documents selected at random from the 25,000 that had been identified (or ‘hit’) with keywords.  The proof of the statistical pudding would be in the eating.  If our sampling proved reliable then on reviewing our random selection of 5,000 documents we would expect to manually find roughly 10 disclosable documents.  If we were wrong in any statistically significant sense, we would consider reviewing more documents.

Then the case settled.


My cautiously firm view is that it was the upfront investment in properly understanding our data, an investment made in good time, which was the catalyst for settlement and that cases like this will become increasingly common.  After all, the Jackson Reforms (capitalisation is surely appropriate) will encourage litigants to be creative in order to control costs while still constructing a reasonable search on solid foundations.

Why should it be judged unreasonable to offer to provide limited e-disclosure when all statistical analysis indicates any review will produce minimal fruits for maximal expense?  It might be said that more disclosable documents equates to a better chance of the case being dealt with justly since the more relevant information that the court has, the better the chance it will come up with a just result.  However, such a statement fails to take account of the need for justice to be carried out at a proportionate cost.  With proportionality in mind and with the statistical analysis having been carried out thoroughly and summarised in a proper, expert-style report, there is nothing unreasonable about putting forward a novel and (on the face of it) contentious proposal for the disclosure of electronic documents.

Whilst on the particular facts I faced, it will remain forever unknown what decision a judge or master might have made presented with such a radical proposal, I am convinced that there has never been a better time to take a calculated gamble and present the court with a well-researched and fresh idea as to how the electronic documents should be disclosed in a case. 

Damian Murphy is a commercial / chancery barrister at Enterprise Chambers.  For over three years his practice has largely comprised the provision of advice and project management in e-disclosure.  In January 2014 he is launching Indicium Chambers, the first chambers to specialise in e-disclosure.