Being a judge looks easy, and most trial lawyers secretly believe they could do the judge’s job as well as His Honor.  But, being a good judge is harder than it looks. Trial judges must be gifted generalists.  They handle disputes in contract law, tort, patent, trademark and copyright law, eminent domain, employment law, criminal law, domestic relations, administrative law, environmental law—you name it.   Judges have to understand procedure and the process of protecting (or muddying) a record better than the average practitioner.  Judges manage bigger dockets than most lawyers with less help.  Trust me: that bozo on the bench is a lot smarter than he looks, and few imagine how much he has to endure from the advocates that come before him.  Sometimes, district court is just traffic court with better shoes.

I offer all that as preface for judging a judge who was just trying to make justice work with precious little help from the lawyers.  I speak of the judge in Armstrong Pump, Inc. v. Hartman, 2014 WL 6908867, No. 10-cv-446S (W.D.N.Y. Dec. 9, 2014).  I’ve never met U.S. Magistrate Judge Hugh B. Scott, but I know that he has twenty years of distinguished service on the federal bench in Buffalo; so, you can be confident he has a well-honed judicial demeanor and has seen it all before.  But, he may suffer from the same debilitating condition that afflicts me:  He was born at an early age and grew up before computers ruled our lives.  He likely learned discovery when everything was paper, and while he may have evolved into judex electronicus (the wired judge), few of his generation of judges have.  It’s asking a lot of judges to keep up with all they must do and become well-schooled in electronic search and retrieval.  That may explain why his order for relief in Armstrong Pump seems destined to fail.  Unhappily so, because when he cries, “Enough” in the opinion, I am with him wholeheartedly.

Defendants Optimum and Hartman licensed rights under several patents to Plaintiff Armstrong, and the lawsuit is a dispute concerning payments pursuant to that license. The events leading up to the order were characterized by delay and sniping of a nasty stripe. The Court wanted discovery conducted in a manner that minimized “piecemeal production.”  It wasn’t.  The Court’s frustration seems amply justified; it’s remedy not so much.

In frustration, the Court took a no-nonsense tack:

In response, the Court now will fashion a new and simpler approach to discovery that keeps the core of Optimum’s counterclaims in mind. In the various discovery documents attached to the motion papers, the Court has noticed that certain phrases appear that inevitably refer to or hint at the Hartman LOOP Technology:

• Hartman
• Hartman Loop
• Hartman Loop Technology
• Loop Technology
• Integrated plant control
• Integrated operation
• Integrated mechanical control systems
• IPC
• IPC controller
• Variable speed
• Variable frequency drives
• Natural curve
• Natural curve sequencing

These phrases open the door to a more objective discovery process that leaves Armstrong no room for gamesmanship. For a period starting from January 1, 2004 through the present time, Armstrong must search ALL corporate documents, files, communications, and recordings for EACH of the above phrases. Armstrong will maintain a list of every server, computer, file room, or other place searched, and a list of all positive search results. For each positive result, Armstrong will procure a full copy of the document in question. …. When the search is complete, a representative of Armstrong and all of Armstrong’s counsel of record will file a sworn statement confirming that Armstrong made a good-faith effort to identify sources of documents; that a complete search of those sources for each of the above phrases occurred; and that the search results have been furnished to Optimum. All of this must occur on or before April 1, 2015, with absolutely no exceptions or extensions. Failure to comply will lead to sanctions under Rule 37(b)(2)(A).”

E-discovery denizens, where do we begin to find a path where success here is remotely possible?  How can Armstrong genuinely accomplish the task ordered and honestly furnish the requisite certifications of a “complete search?”  If you were Armstrong’s counsel, would you be willing to swear that a complete search was made and that all responsive items were produced?

To start, the scope is “ALL corporate documents, files, communications, and recordings… every server, computer, file room, or other place.”  I’m pretty sure His Honor capitalized every letter in “ALL” with purpose.  The judge wasn’t cap-happy with “every;” but, I expect he also chose that word with care.  The searches must encompass the last ten years of information, and it all has to be done “with absolutely no exceptions or extensions” by April Fool’s Day.

If I am correct in thinking that Armstrong Pump, Inc. is the entity also known as Armstrong Fluid Technology, it’s a multinational entity with seven manufacturing facilities spanning three continents. It’s been in business for eighty years; so, though the temporal scope is ‘only’ a decade, the collection from which the ten-year interval must be culled will be much larger.

Armstrong Fluid Technology employs over a thousand people worldwide.  As a manufacturing entity, I imagine some Armstrong employees don’t use computers, phones or tablets in their work; but, just about everyone pictured on the Armstrong Facebook page (except the puppies, awwww) is holding a phone, tablet or laptop and touting the importance of digital connectivity.  They design their products using computers (CAD).  They fabricate their products using computer-assisted machinery (CAM).  Their demos at trade shows are computerized.  Flagship items in their product line are managed by mobile devices.  Sure-as-shooting, a LOT of their employees do their work on computers and mobile devices (the happy teddy bears pictured here had two PCs and an iPhone on their desk).

You can be sure Armstrong has physical and virtual servers galore.  And legacy devices.  And cloud repositories. And backups.  And voice mail recordings. Probably about two zillion thumb drives and lots and lots of paper records…in English and French. It’s a Canadian company, n’cest pas?  Did the Judge intend they search for documents that say systèmes mécaniques intégrés de contrôle?  How does the search tool handle diacritics?

Like most people, Judge Scott may hold to the Myth of the Enterprise Search; the misbegotten notion that companies have the ability to run keyword searches against their myriad mail systems, archives, local drives, file shares, portable devices, removable media and databases.  Adherents imagine that finding responsive ESI hinges on the ability to come up with the right keywords, and that these just, sort of, like, go into a hopper somewhere in IT and, you know, “hit” on responsive documents wherever they reside.  Right?

But, enterprise data is siloed, encoded, compressed and encrypted.  Some is stored as images without searchable text until subjected to OCR.  Searching data completely requires processing it completely…and properly.  Before you can search, you must collect, filter and ingest the data.  Then you must tokenize it to extract text and index that text.  Now, you can search.

Except, you’re going to miss a lot if that’s all you do.  Here’s another picture from the Armstrong Facebook page.  See where it says, “variable speed?”  That’s one of the search terms the Court specified.  Would this JPG image have been processed for optical character recognition?  Will OCR work on such skewed text?  Must they OCR all the thousands of faxes, TIFFs and PDFs that have no searchable text layer?

Yes, they’ve been ordered to do it.  ALL and every.

How will they resolve exceptions?  That is, all those files that can’t be searched without first decrypting password-protected content or resolving a bevy of routine glitches that keep text from getting into that oh-so-crucial index.

No doubt Armstrong employs many databases.  As the contents of those databases are unlikely to be ingested by a review tool in a conventional e-discovery effort, how will they be searched per the Court’s directive?  Though databases share common networks, they rarely share common search capabilities or syntaxes. Searching databases entails specialized query languages and user privileges. Moreover, different tools extract text and index extractions in different ways, with the upshot being that a document found on one system may not be found on another using the same query.

How will the “recordings” be searched?  Will they be manually or machine transcribed, or will a phonetic search tool be deployed?  Will ALL paper documents be scanned and then searched electronically, or will some poor souls have to read every scrap of paper over the next three months?

Finally, even if ALL documents, files, communications, and recordings can be made amenable to search—a gargantuan task–how well will the searches meet the Court’s expectations when thrown against the index?

I ran “IPC” against a random e-mail dataset in my lab and got over a thousand hits.  None of the data was remotely related to pumps or HVAC, and I wasn’t searching forensically-significant sources.  Will a search for IPC also find I.P.C.?  Would IPC or I.P.C. even make it into the index created by many search tools?

Finally, what will Armstrong’s lawyers be bound to produce?  What constitutes “a full copy of the document in question?”  If a database has a hit, do they supply the contents of the field, the record or the database?  What metadata fields must be preserved and produced?  What forms of production employed?

Dare we ask what it will ALL cost?

I called this post “Destined to Fail” for a reason.  It’s not that it can’t succeed.  Heck, we can land a probe on a comet.  It’s just that it won’t succeed.  Unless the order is revisited with a better sense of ESI and search, Armstrong’s lawyers must attest to something that competence would belie or decline to sign and face the consequences.

And you thought YOU were having a bad day!