I have decided I need to explain more clearly why I am so skeptical of the confident claims made for the “health” of Delphi based on the sales performance announced in the press release by Embarcadero earlier this year. Not that I don’t wish it were true, only that I fear that misplaced satisfaction can lead to complacency that I think could be costly.

Rest assured, I shall respond to David I’s invitation to contribute to the list of things that the new Product Manager for Delphi might focus on soon. In the meantime I shall elevate his invitation to the status of a “New Post”, so as to elicit responses from others.

But first, that press release.

The Embarcadero press release of Feb 2012 states that worldwide sales “grew 54% over sales of the previous version” and that this followed 3 years of 15% year-over-year growth, but crucially – if we are to draw any meaningful conclusions from these numbers – it doesn’t say how or on what those percentages are calculated.

What “growth” do these percentage sales increases represent ?

Is it based on comparing sales of new user licenses (predominantly Starter? Pro? Enterprise?) ? Or was it Upgrades sales ? Sales including a purchase of SA ? What ?

At an even more fundamental level, it isn’t even stated whether it is a percentage increase in sales volumes (numbers of customers/licenses) or sales value (raw $’s). Given that there was a big push to move people on to SA and that SA agreements themselves provide a built-in annual increase of up to 5%, if the increase was based on sales value and SA renewals were included then how much is real growth and how much was built-in, mandated “growth” under the terms of SA agreements ?

To add to the muddle, the 54% cited is for the “RAD family .. including RAD Studio, Delphi and C++ Builder”, whilst the 15% figures relate only to “RAD Studio and Delphi”.

As the saying goes (and the title of this post): “Lies damned lies and statistics”.

I do not mean to imply that Embarcadero are lying with these numbers. I don’t doubt for a minute that Embarcadero would have been sure to be entirely accurate with the 54% and 15% figures, but without the underlying numbers or at least a concrete explanation for the derivation of the percentages quoted, it is simply impossible to say what those numbers actually mean.

The one absolute number that does get quoted time and time again, that of “3 million Delphi/C++ Builder users“.

I seem to recall “3 million” being a figure that was widely cited at the time that Embarcadero purchased CodeGear, over 4 years ago. It stuck in my memory because a joke that went around the Delphi users here was that at the bargain price of $30 million, each user could have chipped in just $10 and the ‘community’ could have bought it ourselves!

And I wasn’t the only one. Bret Graffin had heard the same figure at around that time also, as he mentioned in the comments to the blog of a one Marco Cantu (you’ll need to scroll down to the post addressed to “King of All Access”).

In 2007, Michael Swindell quoted a figure of 1.75 million, along with an admission that this included what could only be an estimate, to account for piracy, i.e. unlicensed users.

Currently Embarcadero do cite 3 million, on their AppWave store marketing/PR landing page. But in this case the 3 millions account not just for Delphi, C++ Builder or any other combination of “RAD Family” products, but ALL Embarcadero Technologies customers, including – one must presume – those customers of their wide variety of database tools.

In any event, taking 3 million at face value, 4 years after the acquisition, after a year of 54% growth on top of 2 years (3 years in total, only 2 under Embarcadero as of Feb 2012) of 15% year-over-year growth and still the number of Delphi users is the same “3 million” ?

Something isn’t adding up if everything really is as rosy as some people have concluded, based on the “official numbers”.

And have people already forgotten that 2008 was also the year that saw CodeGear close a deal with the Russian universities for 1 million licenses ?

This 1 million license deal would account for the 3 million total in 2008, since when added to the 1.75 million would yield a total user number of 2.75 million. Add a slightly more generous estimate for unlicensed users and you can easily round this up to 3 million – what’s 250,000 users among friends, eh ? 😉 Though how much allowance was made for the fact that some of those unlicensed users might now be licensed as a result of the Russian deal ?

Maybe we can make sense of things if we work the numbers backwards, from the information in the press release ?

We are told that in 2011, sales grew 54% and that this followed 3 previous years (2010, 2009, 2008) of 15% growth. We also know that in 2008 1 million licenses were added as the result of the deal in Russia. This is an immediate problem for the numbers because 1 million is neither 15% of 1.75 million nor can it possibly represent 15% of the sales in 2007. We’ll just have to ignore this for the time being, yet we do also know that in 2007 the number of Delphi licenses claimed was 1.75 million. So let’s plug those numbers in to a rudimentary calculator:

2007 1.75
2008 = + 15% + 1 million
2009 = + 15%
2010 = + 15%
2011 = + 54%

If you do the math, the total you get is 6.1 million. Clearly this is out of whack with the assertion that there are currently 3 million users.

So let’s take the 1 million license boost in 2008 out completely. But we still get 4.1 million.

Ok, so let’s now assume that the “1” in the “1.75” in 2007 already accounted for/anticipated the 1 million license deal in 2008. This means we start with just 0.75 million, but we still end up with 3.8 million.

If we hack and slash at the numbers, we still cannot get the claimed “growth” to tally with the asserted number of users.

I am being facetious, but this appears to be the interpretation of “growth” that some people are using, despite the fact that the numbers don’t actually fit, however hard you try (so presumably, they simply haven’t tried).

Because of course the percentages in the press release are increases in sales, not increases in total licenses. This makes things a sight more complicated because now you really do need to be able to know whether it is sales volumes or sales values that are the basis for the quoted increases.

Furthermore, without a breakdown of new user licenses vs SA renewals or upgrades, the percentages are quite simply meaningless. More people may be upgrading and renewing SA when compared to previous years, yet the number of overall users might still be in decline.

I don’t mean to say that this is the case, only that the numbers as presented in this press release – as accurate as they may be – do not support any meaningful conclusions, and certainly not the ones that some people are so determined to draw.

Let’s Take This Debate Down the Pub

My response to David I’s assurance that the numbers in the press release were indeed accurate was to say that “percentages can’t be taken to the bank” – a saying that I first became familiar with when I ran a pub for almost two years in the UK (a great way to spend money, incidentally).

In the UK, spirits have to be sold – by law – in measures of either 25ml or 35ml (so a “double” can be either 50ml or 70ml, depending on where you buy it, for example).

As a retailer you have to choose – one or the other. This affects your prices obviously, but more relevantly here, also affects your margin.

If you increase your serving from 25ml to 35ml, then you increase the volume per serve by 40%, but you cannot increase the price by the same amount – nobody would stand those prices. So you have to take a lower margin.

But that doesn’t matter – you are still selling “1 drink” and although the margin is lower as a percentage cost of sale, the actual bankable money in the till is higher. And you still sell the same number of drinks, because people don’t buy alcohol by volume (“No more for me thanks, I’ve had my 70 ml of whisky for the night”) but by the serving (“Yeah, go on, I’ll have another one”).

The point being, that percentages rarely tell the whole story.

The real question is, what are the absolute numbers ? What was the baseline figure for Delphi 2010 sales over which 2011 represented a 54% increase ? Or, getting straight to the point, how many active (in use) licenses are there ? Really ?

As I said before, with the need to activate licenses these days, this should be an easy number to determine.

Whether you would choose to talk about those numbers of course – that’s a different matter.

17 thoughts on “Lies, Damned Lies and Statistics (and the Margin on Spirits)

    1. I don’t think so, Delphi have its punch, unfortunately embarcadero is taking bad decisions. Marco Cantu can make the difference if success, otherwise it will be harder to return, I still have hope in Delphi.

    2. I’ve reached the same conclusion. I upgraded to XE3 to complete a single project that was promised. My loyalty to the brand itself died a few months ago when I realized that a very prominent EMBT member lied to me personally. Now that my project is complete and I have some wiggle room, future development will switch to Lazarus and FPC.

  1. Marketing growth rate is meaningful actually for investors and shareholders not for users or 3rd party vendors. Absolute numbers of licenses and activated licenses of Delphi by edition and year of purchase/activation are numbers that matter for vendors of components and tools to give them enough numbers to make their own decisions. IMHO, Delphi would be as good as dead without the market of 3rd party vendors. Obviously, unlicensed Delphi users may use unlicensed tools and components and would’nt pay for them. So the number of them doesn’t matter.

    1. sure. you can go torrents and use unsold license.

      u also can stop using the license you purchased 5 years ago.

      but given those overboard prices, i can hardly imagine making a fresh purchase of license and instantly stopping using it.

  2. Growth is about money. Allan once commented on this.He said the prices would have increased indeed but quantity must have increased too. Indeed it must have been about XE sales a lot more than XE2 sales – if I remember correctly. Assuming EMB sold 1 Mio licenses to the Russian government of course not at the regular price at the store. Just to give an idea – Trusts buy Software at the price of 10 to 20 seats but count almost every computer. You cannot guess from these numbers.

    In a global market driven by people who believe in neo-classics you can assume that a tool at the price of Delphi can have about 200k to 300k seats but not millions. We can invest an buy EMB’s income statement. I am not sure if I really want to read – guessing is more fun.

    The only thing that is relevant are the sales numbers in money. It is irrelevant how many people are using a certain product from the perspective of being worried about the ‘existence’.

    For me it’s clear – voting, customer surveys and similar things leads to ochlocracy. I believe EMB that their nominal ‘GDP’ is growing but I am doing hard to believe in real ‘GDP’ growth in term of sustainable structures in the customer base. Customers who buy, update once decide for an S.A. and stay happy over the years. Charging higher prices to existing customer base and growing from this comes close to kleptocracy.

  3. When a company talks about growth in sales, they are almost always talking about dollars of revenue, not unit sales. There’s no need for the mathematical gymnastics here. The million units to Russia are completely irrelevant in this argument, and the figures say absolutely nothing about the installed base.

    In my opinion the only thing Embarcadero have said to raise an eyebrow is the switch in terminology from “Rad Studio” to RAD Family” in those annual growth numbers. That may or may not have been a sleight-of-hand to make the figures look a bit better, but it’s not likely to have been a massive misdirection.

    1. That was essentially my point. Some people have interpreted the “Growth of Delphi” as meaning an increase in the user base. The wording of the headline and the introductory statements in the press release seem to reinforce this – most likely incorrect – interpretation.

  4. Selling 1 million copies of delphi in 2009 and another 1 millionn copies in 2010 doesn’t mean you have 2 million delphi users. And that seems to be the basis of your whole post. The same person who bought Delphi 1 and also bought Delphi 2, and then bought Delphi 3, 4, 5, 6, 7, 2007, 2009, 2010, XE, XE2, and XE3 has bought 14 copies of Delphi. (I have purchased 14 copies of Delphi). I am only one user.

    The drop in sales in Delphi occurred around the 2007->2009 period and was due to the “unwillingness to deal with Unicode”. The same existing commercial base that refused to jump at 2009 more or less started buying in slowly again by about the XE timeframe. With XE2, there was a HUGE bump in sales due to everyone who wanted 64 bit and was willing to pay for it, giving some coin for it.

    Estimating the actual live # of delphi users is something that nobody inside of or outside of Embarcadeo can do accurately. However, sales figures could be up 30% per year for up to 3 or 4 years before I’d get suspicious of the figures, because frankly, 2010 might have sold atrociously. XE sold well, I know for sure, and so did XE2. I don’t believe Embarcadero is going to get as much traction with XE3 because the release isn’t as ground-breaking as XE2 was.

    But your logic is flawed, and your math is conflating sales and user base.


    1. No Warren, the majority of the post attempts to explain the numbers in the way that other people have interpreted them, and explodes it’s own logic. That was my entire point – hence “I am being facetious”. The real point, as stated in conclusion, is that the numbers in the press release tell us nothing, without the underlying data, which we don’t have.

  5. if they count al the tools, then that may be 2-tier number.

    Say, i wrote classic client-server tool with Interbase under the hood.
    I sell 10K copies of it – tadam!, Embarcadero can add up 10K new “customers” – but really those are mine customers, not theirs.

  6. Let’s start with a basic premise: Most users do not upgrade every year.

    Another premise: Ranking of license shares, most to least, is probably Pro, Ent,, Arch.

    Thank you, yes, I do realize I have no foundation for those assertions, but bear with me.

    If we take as a guess that 10% of users upgrade every year, and assume that the growth figures cited by EMBT refer to dollar volume, not license count, then where are we?

    1.75 Million * 33% = 580K
    2008: 664K
    2009: 764K
    2010: 868K
    2011: 1.01M

    Now if we further guess the distribution of licenses as Pro = 50%, Ent = 30% and Arch. = 20%, and ignore altogether the sale of new licenses, and use upgrade prices of $400, $1000, and $1700, respectively, and constant for those years (I don’t have the real numbers at hand), we find revenues of:

    2007: $485M
    2008: $558M
    2009: $642M
    2010: $738M
    2011: $848M

    Now further consider that my model is all too simple. No RAD Studio, no BCB, and no new licenses. Still, the revenue figures are impressive. They would also lead us to believe that EMBT is a company with revenues of over a billion dollars a year. Plausible? Before drawing any conclusion, have a look at some old Borland 10K reports.

Comments are closed.