Why there's nothing special about the business of software

Tweet

Software businesses should be infinitely scalable, right? You've done the hard work. You've built your product, your money making machine. All that's left is to turn its frictionless handle and churn out uncountable quantities of dollar bills. After all, the cost of shipping your creation's sweat-filled bits and bytes to your next customer, and the customer after that, is zero.

But that's not what happens in real life. Take a look at this graph of revenue per employee*, generated from the 2008 Software 500 data:

The median annual revenue per employee is around $160,000. Almost all (90%) of software businesses generate under $300,000 / employee. By comparison, General Motors has revenue of $600,000 per employee; Walmart $200,000; Intel $450,000; Exxon $5,000,000. It turns out that shipping electrons is no easier than building cars, selling cereal, building chips or drilling for oil. There's nothing special about software.

How come?

In 1980, Theodore Levitt - Harvard Business School superstar professor - wrote that there is no such thing as a commodity (or more accurately, there need not be such a thing as a commodity). In 1986, Bill Davidow - erstwhile product crusader for Intel's 8080 and 8086, now venture capitalist - expanded this theme beyond commodities and wrote about the concept of 'device' vs 'product'. In Davidow's terminology, a device is the good that, at first glance, you sell. It's the coffee beans, the silicon chips or the bits and bytes of your software. But the product is what you really sell:

• Trivially, bits and bytes
• Reassurance that you won't rip off your customers and that they're doing the right thing ('nobody ever got fired for buying IBM')
• Reassurance that your software will work as advertised, and that you will be there for customers if they get stuck (Rackspace's fanatical support)
• Reassurance that there is a roadmap, that you will continue fixing bugs, refining the product and releasing new versions (Intel's chipset)
• A statement about your customers (anything from Apple)
• The chance to belong (ditto)

Of these, only the first is scalable and easy to supply. The rest require people to deal with customers, communicate with the market, investigate new opportunities, build brands, grow communities, write documentation, create a company culture and so on. Those are the activities required to decommodify the bits and bytes. They're not cheap and they don't scale.

But if you want to grow, they're the most important things you need to do**.

I'd like to hear what you think. Post your comments here, or carry on the conversation on Twitter (I'm @neildavidson).

(* Yes, I know that revenue per employee isn't as important as profit, but they're the numbers I've got. Also, these numbers are from the Software 500 so may not be truly representative. But they illustrate my point that software businesses don't, on the whole, scale).

(** Assuming you've done the hard - and it is extremely hard - work of building the bits and bytes already).

Tweet

Subscribe to the RSS feed

Comments

[alex]

image is broken

[Mark Dalgarno]

Can you break down the figures further?

I'd be interested to see how companies that derive most of their business from software consultancy fare compared to product-based companies. ISTM that the former are even less scalable.

[Neil Davidson]

Alex,

What browser are you using? Is the image still broken (it might have been a temporary problem)?

Neil

[Neil Moodley]

Really illuminating post, Neil. Thanks for writing it.

I have often felt that the production of a good or service is, comparatively speaking, the straight-forward part, especially in software. Sure it needs smart people to do it well, but that's not the end game. It's the 'profitable business' bit where most come undone.

Why should that be? Reasons vary, naturally, but I think much of it is to do with a degree of snobbery on the part of the producers - the programmers and designers. I have found in over 15 years of employment in the software industry that a disproportionate number of technical people have enormous egos of the "if I build it, they will come" variety. These egoists do not generally want to sully their vision with the grubby concerns of, fundamentally, money.

Finding a market, understanding a market, reaching a market, competing in a market, being heard in a market are all really, really hard and often incredibly nuanced things to do. But until you find someone willing to fork over cold, hard cash (or something else of preferably intrinsic value), the producer's 'vision' is worth literally less than zero.

I talk a lot about the nuts and bolts of starting, running and growing successful businesses over at my blog - http://www.fourthirds.com. Feel free to check it out and comment accordingly!

Look forward to reading more of your work!

Neil

[Debby M]

We are nine years into a profitable Internet business based on Windows applications positioned as free. As I watched the owners grow their business, one of the reasons they were successful was that “they followed the money” even if it meant changing their business model with the times. W3i’s revenue comes from our proprietary Windows installer that promotes relevant software during the installation when the user is already installing so are more receptive to other software offers. Unusual revenue source but it has been supporting several new start-up businesses, including Digsby. Having a quality, valued app is not enough to keep the lights on, you need to determine how it is going to make money.

[Allan Martinson]

I think the comparison with automotive or chip industry is quite flawed. Many, if not most industries have higher revenues per employee than software industry. But the analysis makes sense only if we compare VALUE ADD per employee, not revenue. Most industries have many production inputs they must purchase (cost of sales). The price of a car consists of many components most of which has been produced outside automotive industry (metals, plastic, rubber etc). Same with chips, or even oil.

The software industry, however, produces the product mostly from thin air and almost all of their revenues = added value.

In other words, to make correct comparisons you shall divide the revenue of automotive industry to ALL people involved down the value chain (subcontractors, metal and plastic producers etc). I am sure you will get totally different number.

[Kevin Dangoor]

I agree with Allan above, but rather than going with value add and tracking down the value chain, I'd say that it would be interesting to look at *gross profit* per employee (or even net profit if you want to be daring, but there's a lot more involved in net profit).

As you point out, when you ship software the marginal cost of software is very low. It takes much lower revenue per employee to break even.

[C R Patiño]

I kind of agree with the last two comments, in the sense that it is not a fair comparison between the software and automobile industries. However, I think the critics, as well as the original article miss an important point. The most important differences between them is not revenue per employee, but return of investment and barriers of entry.

I think this article starts with a flawed assumption. That most of the people in the ISV community are in for the huge amounts of money they expect to get out of their effort.

In the ISV scene, a non trivial (if not major) percentage of upfront investment is directly related to the founder's cost of living for the first 2 or 3 years of the venture. Basic computer equipment and Internet access is affordable. Development tools can be free, thanks to the open source movement. Delivery infrastructure over the web can be cheap also, at least with starters packages. In other words, labor, domain knowledge and relationships are the most relevant inputs for the process.

Contrast that with the automobile industry, or whatever high-value-added industrial enterprises (automobile, pharmaceutical, semiconductors, etc). Their process will require the upfront purchase of very expensive machinery. Any of this equipments have the capacity to produce high volumes, and it does not make economic sense to produce in small scale once you have to amortize that equipment in your quarterly reports. You are bound to operate them as close as full capacity as possible.

Now, if you have your factory you will need lots of inputs and operators to keep it going. The supplies will need storage space and more people to manage vendor relationships and inventories itself. The end product will need some sort of distribution channel to reach the customers.

The stakes of the game will be so high that you will need an army of lawyers and PR professionals to scare off the barbarians wandering around the borders of your raising empire. And a praetorian guard^H^H^H^H^H^H^H^H^H HR team to keep all the rest together.

So, your chances of starting up another big car company are as good as your chances of your last name being Rockefeller. Even starting a little niche company in this industry (lets say, a golf carts firm) will require you to have significant resources; financial, political and otherwise. You will be messing up with your and other people's money, so a perfect business plan will be required before anything else.

If you are going to compare software companies with currently successful corporations, I would go Wallmart. To take on Wallmart will be foolish from day one, but so would be to take on Microsoft. A single founder can use personal middle class resources to start up small, learn as he goes and grow from there (at least until he pisses off either Wallmart or Microsoft).

[Thomas Attila Lewis]

What isn't addressed here is the cost of marketing these products by industry. The amount of advertising and positioning and the skillfulness of that marketing has a huge impact on the ending revenue. It's not all about production. As Peter Drucker said, the purpose of an organization is to innovate and market. They are equal players.

[Marco Fontebasso]

There is a flaw here:
> The median annual revenue per employee is around $160,000.
> Almost all (90%) of software businesses generate under
> $300,000 / employee. By comparison, General Motors has
> revenue of $600,000 per employee; Walmart $200,000; Intel
> $450,000; Exxon $5,000,000

You can't simply put the total revenue ignoring the gross margin associated with the revenues which is really different between Automotive (industrial cost) and Software.