Sunk Costs

Occasionally we run into customers who are reluctant to use Disco because they have large sunk costs in the software they currently use.  While people are free to make decisions based on whatever criteria they choose, we think that particular analysis is not correct.

A sunk cost is a cost that has already been incurred and cannot be recovered. Compare that with a prospective cost, which is a cost that might be incurred when a decision is made.  Economic theory states that people should consider prospective costs but should not let sunk costs influence decisions because doing so does not allow for making a decision exclusively on the merits. However, sunk costs greatly affect many people’s decisions because they don’t want to “lose” the money they have already spent. This is the sunk cost fallacy.

The sunk cost fallacy is often seen in a reluctance to abandon projects in which a person has already invested considerable resources. People want a good return on their investments and have a genuine interest in making their efforts worthwhile. People also tend to hold on to an investment despite evidence that the investment will not produce the returns they had hoped.  No one wants to feel that they wasted time or money.  

The key to avoiding the sunk cost fallacy is to remove any past investment from the decision making calculus.  Since the costs have been incurred and cannot be recovered, they should simply not be considered at all in making future decisions.

Let’s take an e-discovery software to illustrate the point.  Let’s say Company X has been marketing review software for one year and is presented the opportunity to sell Disco at the end of that first year.  Chances are that Company X had to pay a yearly license fee for its current review software.  It likely had a yearly fee to license the software it used to process the data for loading into the review software.  It also likely paid 2-3 employees to run the processing software.  Hypothetically, if the review license costs $10,000/year, the processing license costs $15,000/year, and the employees are paid $25,000/year each, Company X had costs of $75,000-$100,000.  The company has also likely invested significant resources hiring salespeople to sell the review software and training people how to use it.  This is in addition to the per gigabyte fee for storage in the review platform.  Those are sunk costs that cannot be recovered.   These same costs are also prospective costs for year two for the incumbent review platform, costs which Disco does not have.

It’s easy to see why Company X might feel tied to the software package it currently offers. However, the correct economic analysis is to ask which software will make the most money in the future (including evaluating prospective costs), and not to think about how much has already been spent.  When the company applies the correct economic analysis, however, we think the correct decision becomes clear. 

Sales Tactics

I am not a sales CEO.

As a law firm partner, I focused on delivering great work and letting the work speak for itself. I was good at closing new matters (when a client was picking between multiple firms, I was good at convincing them to pick us), but I was not good at developing new leads, maintaining relationships, cross-selling new services, golf, or any of the other skills of rainmaking partners. I have great respect for people who are good at sales because I know I’m not.

But sales is critical, especially in a startup like Disco. My approach is “scoreboard”: test as many strategies as possible as cheaply as possible and let sales tactics be guided by the results, defined in terms of revenue numbers.

Some sales tactics can be ruled out at the experiment stage because they’re unsavory. Anything that involves unethical behavior or less than perfect honesty with customers, for example, is out, regardless of whether it drives revenue. Anything that dilutes our focus on building a great product with broad appeal is out — and this includes catering to customer-specific requests or any kind of services offering when one is building a technology company. But anything else that sounds reasonable is in.

At my law firm, for example, we found that press attention was the number one driver of new business (> 50% in dollars); referrals from former clients or former cocounsel or opposing counsel was second (25% in dollars); and everything else (speaking at events, CLEs, cold or warm calls, writing, lunches, general networking, etc.) was a very distant third. This was contrary to our initial expectation. It may have something to do with our particular strengths and weaknesses; golf may be more effective for people better at it than me.

At Disco, we initially tried direct marketing to law firms and legal departments, based on our belief that a 10x faster 10x cheaper product would sell itself. Boy, were we wrong. We also experimented with outsourced direct sales using a sales company in Silicon Valley. That was a complete failure (and a big waste of money) too. Then we tried selling through channel partners who have existing relationships with litigation partners and other end-user decision makers. That was a great success.

With channel partners, in turn, we’ve tried a variety of things. We’ve tried mass emails advertising speed and cost. We’ve tried advertising in bar journals and other publications that lawyers read. We’ve tried personal calls to law firm partners and other decision makers. We’ve tried free trials, flat-fee whole-case pricing, a full-featured online demo with a big data set loaded, and lots of other things. Again, the strategy is try everything that seems reasonable. So what works?

The number one driver of Disco sales, by far, is demo days (> 50% in dollars). In a demo day, we take over a conference room for a day at a firm where we already have one or two initial customers or at least one or two contacts who have seen and like the software. We’re available all day in the conference room to give demos to litigators whenever they have 15 or 30 minutes free. What usually happens is that the one or two people who are already disposed to like it come and see the software, then go tell their friends. The friends in turn come and see the software and go and tell their friends, etc. etc. so that we have a stream of people coming in to see demos throughout the day. The same people wouldn’t take a call from us or a channel partner and probably wouldn’t schedule a meeting; but when we’re at their firm and all they have to do is pop in to a conference room whenever they’re free, they will come and see a demo; and once they see a demo, they love it. In this way, Disco spreads from one or two users at a firm to a large part of the firm’s total business.

The number two driver of Disco sales is RFP participation. We are almost always the low bid by a significant amount (coming in at 50 – 75% of the next bid) and with the fastest software, especially for the large data sets for which RFPs are more common. The number three driver is lunchtime CLE presentations where the entire firm is invited to a catered lunch to hear about Disco in the context of a CLE presentation on “The Future of Legal Technology.” The number four driver is one-off demos to individuals with an existing relationship with a channel partner. Everything else that we’ve tried (and there’s certainly more yet to try) is much less successful than these four tactics.

As a non-sales CEO trying to drive sales in a startup, it’s important to measure what affects revenue and to put your weight behind the sales tactics that the data shows do drive revenue. All the beautiful collateral, all the hard work, and all the good intentions you and your team can have don’t matter in the world of sales if they don’t result in posting numbers on the board. When you’re not naturally good at something, there’s no substitute for data.

We love to share information with our channel partners about what sales tactics work and what don’t. And we actively try to help channel partners pursue those sales tactics that we have found do work. Right now, that’s (1) demo days, (2) RFPs, (3) CLEs, and (4) targeted demos to individual decision makers with an existing prior relationship to the channel partner.

Big Cases

Disco shines on big cases.

The largest single database in Disco right now is 4 TB, ~25,000,000 documents, and ~125,000,000 pages. On a database of this size, search and document navigation remain instant.

Getting speed on big databases requires a combination of our usual speed strategies: (1) doing everything computationally intensive (OCR, threading, duplicates, near duplicates, imaging, indexing etc.) on ingest and in parallel rather than during review; (2) using a fast document-based database (RavenDB) and Lucene for search; and (3) running on excellent noncommodity hardware.

For > 1 TB cases, you should always use an RFP process and you should make sure to send an RFP to a Disco provider. Disco’s bid is typically 50% – 75% of the next-lowest non-Disco bid; includes complete processing, unlimited users, and unlimited digital productions at a single price; and is for software that is both faster and easier to use.

Why do this?

Law is how we say things should be given how they are; law is the projection of a vision of the future on the present.

Law determines what businesses we can structure, what enforceable relationships those businesses can have, what the duties and obligations of each are to each other. Law is not the only way of determining these things; the law lord competes with the lords of culture, media, religion, each influencing the others in a struggle over human action.

Law is important because that’s the field of play; law is about determining human action, guiding it, enabling it, shaping it for the better, for the right, for the just. Law is civilized violence, pain administered in a velvet glove; in the end, people obey law because of the pain the legal system imposes on them if they don’t. That’s the weapon and the responsibility.

It’s crucially important that law be done well. Lawyers are called on to administer the law, to shape it, to apply it, to criticize and cauterize, to explain its glories, minister to its subjects, to implement it. When lawyers fail, when they lose sight of law’s role, of the legal calling, we see injustice and lawlessness; lawyers’ failings too often obscure the fundamental good of the law; and we see far too much of this today.

Our job at Disco — the reason you should work here if you’re a world-class person who wants to make a difference, who’s deciding between this or Google or SpaceX or Goldman or government — is to enable lawyers ultimately to do law better, to better guide human action. If you’re for the basic promise of law — of a force-backed, reasoned vision of the future — and you want to work to make that promise good through the leverage of technology, then this is the place for you.

It’s easy to get lost in the day-to-day details of a particular product, like Disco, or the specifics of competition in the ediscovery industry. But we shouldn’t lose sight of what all that is a stepping stone to, of why we’re doing this and not self-driving cars or populating Mars or getting filthy rich on Wall Street.

That’s the big picture: doing law right.


Market Size

A recent blog post gives some nice graphics and estimates of the size of the ediscovery market today and estimates for its growth into 2017. The post combines estimates from Gartner, Radicati, and other industry analysts.

The 2013 estimates are $1.49B in software and $4.04B in services. Of the software spend, 62% is for “processing and review,” while 38% is for “early case assessment.” The 2013 numbers are $2.7B on software, with 55% of that on processing and review and 45% on early case assessment. This reflects an estimated annual growth rate of 16.78%.

(It’s not clear what goes into the services estimate; it includes a review component, but this may be intended to be limited to third-party reviews (contract reviewers); if it is meant to include law firm attorneys’ fees on document reviews, it is likely a gross underestimate. But we are a software company, so we’ll just ignore that part of the post.)

At Disco, we think early case assessment is not a category of software that should or will exist at all; instead, it’s a response to how bad (slow, difficult to use, and expensive) processing and review software has been in the past. Or, as we put it earlier, early case assessment is obsolete. The win in this industry is going to go not to the company that makes it easier to cope with the unpleasantness of legacy review software, but to the company that delivers better review software in the first place.

Our own conservative market estimate goes like this.

Take the AmLaw 200 (the 200 largest law firms in the United States ranked by revenue). Assume that each AmLaw 200 law firm has 50 “regular” cases per year and that these cases average 200 GB of data each. That’s 200 x 50 x 200 = 2,000,000 GB of data. Assume that each AmLaw 200 law firm additionally has five “large” cases per year and that these cases average 1,000 GB of data each. That’s 200 x 5 x 1,000 GB = 1,000,000 GB of data.

Together, that’s 3,000,000 GB of data per year. Of course, you need pricing to translate that into dollars; assume very conservative (i.e., lower than reality) all-in pricing (processing, hosting, production, etc.) of $20 / GB / month. That is 3,000,000 x $20 = $60,000,000 per month or $720,000,000 per year. So our very conservative software market size estimate is $720M — and growing rapidly.