L’Open source commercial : au-delà des utilisateurs pionniers
Par Bernard Golden, Navica Software (en anglais)
Publié le 22 juillet 2007
Commercial Open Source: Crossing the Second Chasm
If you’ve spent any time at all in the technology industry, you’ve heard the phrase “crossing the chasm.” This phrase, made famous in Geoffrey Moore’s book of the same name, refers to a phenomenon of technology markets.
His theory breaks technology customers into a number of categories: early adopter, laggard, and so on. You can see a chart representing customer types and the location of the chasm in Figure 1. If you examine the Figure, you’ll note that the chasm separates early stage customers from mainstream customers, with the former representing around 15% of a market, with the latter representing around 85%, thereby illustrating why crossing the chasm is so vital; most of the growth potential and profits lies on the far side of the chasm. Moore’s primary message is that there is a discontinuity between early stage customers and mainstream customers: what worked for early customers isn’t sufficient for later stage customers, and satisfying the latter isn’t just a matter of doing a bit more incrementally; in fact, there’s a chasm of expectations between early stage and later stage customers.
Later stage customers expect far more than early stage customers, and they expect a different type of offering from their providers. While early customers are willing to live with incomplete and non-fully featured products, across the chasm customers demand fully functional products – but that’s just the beginning; beyond the product itself, these customers need great support, wide professional services availability, easy integration with other products, and so on. In other words, across the chasm customers need products easy to consume. Moore calls them “whole products,” because they have all the characteristics to succeed with mainstream customers.
Moore posits that getting a technology ready to address these customer requirements is an enormous challenge – as one might expect, given that a chasm must be crossed. The crux of his argument is adding all these additional product characteristics is a huge amount of work for small vendors and requires different abilities than what got those vendors to the chasm jumping-off point. While quick responses and immediate customer satisfaction may have been enough to get going, building out a professional services network and the myriad of other tasks is a job of a different order. Not every company successfully jumps the chasm.
Moore’s theory is extremely influential in the technology industry. “Crossing the chasm” is as widely-bandied about a phrase as there is. In fact, I used it in the Open Source Maturity Model, noting that, for open source, finding and evaluating the different resources that comprise a whole product (which I called a mature product) is a different type of task, given that many open source products emanate from non-commercial communities or from tiny startups without the resources to create whole products.
As you might expect, Moore’s theory is widely applied to open source software markets. You often hear people asking “has open source crossed the chasm?” I have heard the question answered in the affirmative each of the past four years, thereby indicating that open source has had the most protracted crossing in history!
However, in presenting this notion, the people asserting so confidently that open source has crossed the chasm have overlooked a critical thing: it isn’t enough for commercial open source companies that their products be adopted by mainstream users; they have a second chasm to cross – and that chasm is a lot tougher to get across.
That chasm is the chasm of payment; in other words, getting a user to become a customer. If you look at Figure 2, you’ll see that the location of this second chasm is far to the right of the first one. Depending upon the individual product, most open source companies estimate that approximately one in 1,000 users become customers; that is, the paying customer base is only those companies lying beyond three standard deviations from the mean.
Seen from the perspective of the traditional “crossing the chasm” model, this situation is unheard of. The traditional model has mainstream companies unwilling to buy a technology until it’s a whole product; once the additional characteristics are available, mainstream companies, according to the model, turn into very willing buyers.
The situation with commercial open source appears to be vastly different. In fact, it directly contradicts the traditional model. In an upending of Moore’s formulation, it seems that mainstream companies are adopting open source without relying on a vendor - or, at least, without needing to purchase the full range of capabilities from a vendor. Instead, it appears that the leopards have changed their spots, and mainstream companies are downloading open source software, putting it into production, yet fleeing a financial relationship with the provider.
What can account for this puzzling change in the behavior of mainstream companies? Why are they confounding us with their inconsistent comportment, their contradictory willingness to overturn the time-honored truths of Geoffrey Moore? How can we understand the reality of commercial open source and better predict the future of open source software?
Definitive answers are not possible, but here are some thoughts about what the second chasm might mean:
Reason #1: Open source hasn’t actually crossed the first chasm
The prediction of Crossing the Chasm is that once a product has all the secondary elements of a whole product, mainstream customers will line up to purchase. Yet, despite the efforts of commercial open source companies to provide all those elements, very few companies actually buy.
Maybe it’s because open source really hasn’t crossed the chasm and mainstream companies are not actually using open source; that is, maybe open source is still the province of early adopters, few of whom feel compelled to buy because they are, generally speaking, self-reliant.
While this reason would account for why only 1 in 1,000 users would buy, I’m not sure I buy this reason. After all, some of the announced customers of commercial open source companies don’t seem like risk-seeking early adopters. MySQL claims the US Census Bureau, Avery Dennison, and Macy’s as customers, which doesn’t sound like a risk-seeking group to me.
Therefore, it seems that open source has crossed Moore’s chasm. It’s just that those companies lying on the far side of it aren’t behaving the way the model predicts. It’s awful when reality doesn’t conform to hypothesis!
Reason #2: We’re using the wrong denominator
In this view, maybe that 1,000, of which only 1 becomes a buyer, is a wrong number. Perhaps a large percentage of that 1,000 is not really made up of real product users, so we’ve been using the wrong number as the denominator of the fraction of users to buyers. Maybe most of that 1,000 is really irrelevant, nothing more than students, experimenters, and idle riff-raff. If you get rid of these, maybe you’re down to a one in 100 or one in 50 ratio.
There’s obviously some truth in this reason. Everyone acknowledges that downloads are not a very accurate measurement of true interest in an open source product. Of course, that hasn’t stopped anyone from using downloads as the metric of open source success. Venture capitalists, in particular, seem wedded to downloads as a key factor in funding decisions.
Even if the denominator is changed significantly, however, this reason seems insufficient. Taking it to one in 100 or one in 50 still seems radically inconsistent with how companies should behave when a product crosses the chasm.
Reason #3: Maybe we’re in a transitional period of customer behavior
It might be that we’re in an interregnum, partway between the old order of proprietary software and the new order of open source. While resting in this purgatory, the long-term behavior of mainstream companies is masked due to confusion or temporary giddiness at the new state of affairs in which the formerly forbidden fruit of free software is now brazenly on offer. Once free software is no longer novel and tempting, mainstream customers will come to their senses and start behaving like they used to; their conservative instincts will reassert themselves in the future, and they’ll all become paying customers.
While it’s emotionally tempting to view the one in 1,000 (or one in 100, if you accept the rationale advanced in Reason #2) situation as a brief bacchanal, a temporary loosening of quotidian morality, shortly to be superseded by a return to stolid mainstream customer behavior, it’s hard to see today’s perplexing refusal to buy as a short-lived phenomenon.
It’s a cliché of the self-help movement (and its IT counterpart, re-engineering) that people resist change. If that’s the case, why would the mere business model change represented by open source software be sufficient to overturn long-established expectations and processes?
Frankly, it’s hard to see. Whatever the reason so few mainstream companies buy from commercial open source providers, it’s hard to imagine that it’s short-term. It seems more likely that the current ratio will be consistent into the future, and that some other reason must account for it.
Reason #4: Non-buyers are free riders looking at signals
It might be that the elements that pragmatic buyers traditionally demand accompany new technologies are really necessary. Unless and until the technology’s owner creates viable support, training, professional services, etc., pragmatists may indeed refuse to adopt the technology. And, once those elements are available, pragmatists will eagerly embrace the technology.
However, nothing in the preceding paragraph implies that pragmatists will buy the technology, just that pragmatists will adopt the technology once those necessary elements are in place – and adopting is different than buying!
It may be that, in the absence of paid license coercion, pragmatists interpret the availability of the whole product elements as sufficient for adoption; put another way, as soon as those elements exist and someone is paying for them, pragmatists will be satisfied about the product’s maturity. In evaluating a product’s suitability, pragmatists look for signals in others’ paying decisions; once someone else has taken the (financial) plunge, the cheapskates reap free software. In this view, the vast majority of pragmatists are free riders, taking advantage of the company creating the whole (in Moore’s sense of whole) open source product. In essence, these non-paying pragmatists are gaming the system, allowing others to absorb all the costs of crossing the chasm and siphoning off 99.9% of the benefits.
There’s quite a lot to recommend this view, as it aligns with the observed behavior of non-paying pragmatists that, at first glance, seems so puzzling. On the other hand, opting out of a paying relationship means that these pragmatists have no “throat to choke,” no support source providing the alleged all-important SLA. If one accepts that pragmatists are risk-avoiding organizations, taking advantage of others’ efforts and payments seems out of character. After all, one would hardly expect to see a pragmatist drive without insurance, despite a lack of mandatory purchase.
Reason #5: The behavior of pragmatists really has changed
Eliminate the impossible, said Sherlock Holmes, and whatever is left, however, improbable, is the truth.
Having discarded all the other reasons for the quizzical refusal by pragmatists to pay for open source software, we must conclude that, for whatever reason, their behavior with respect to open source truly is different than might be expected from Moore’s theory.
It may be that the signaling reason advanced above does hold – that pragmatists wait for someone to take the actions that indicate product wholeness and then they feel comfortable enough to adopt the product.
However, the assumption that the nature of pragmatists is such that they would be impelled to purchase from the technology supplier is wrong. Instead, they see the whole product as indicating sufficient maturity but are individually willing to rely on community support.
This is a troubling state of affairs – and not just because these trustworthy pragmatists are showing an unexpected willingness to drive sans insurance.
What’s more troubling is the question of how one segments the pragmatist bloc to identify that portion of it that will adhere to the traditional expectations about the bloc: demanding of a whole product and willing to pay for it, too. In other words, how do you identify users who will cross over the second chasm and become customers? That will be the subject of next month’s newsletter.