Country Analysis of Magnatune SalesBelow, you'll find a country analysis of Magnatune's sales.
Methodology: I started logging TCP/IP addresses with each purchase in September, so unfortunately, I only have 4 month's worth of data, but that's still enough to be interesting. I purchased a "TCP/IP to country" database from IP2Location and wrote a script to determine which country purchasers come from. I was fairly impressed with IP2Location's database, as Usenet is abuzz with the complexities of converting TCP/IP addresses to countries. For example, you can check the .country on the reverse dns name, grep whois entry of the domain that owns it, or grep the whois of the ISP controlling the TCP/IP address. When I hand checked a few, the database was correct (based on the techniques Usenet said to try). Perhaps more importantly, I found few mistakes when comparing (customer entered) CD shipping addresses with the country derived from their TCP/IP address.
Notes on this chart:
* 63% of our sales are to the USA, 37% outside the USA
* The Christmas season has seen an increase in the gap, likely fueled by Christmas gift purchasing. Otherwise, the 63%/37% split is fairly stable.
Notes on this chart:
* This chart shows how the top 4 countries in our non-USA sales are doing.
* The percentages in the chart are the percentage of non-USA sales that each country is responsible for (i.e.: the UK being responsible for 17% of non-USA sales).
* We've seen a big rise in the past 4 months in Canadian sales, from 10% to 18% of the non-USA total.
* France has seen a modest increase (4% to 7%) while Germany has had a modest decrease (10% to 6%). We think this is due to more press interest in France, and less press in Germany.
Notes on this chart:
* this chart shows how much, on average, each country volunteers to pay for an album of music on Magnatune.
* Italy ($9.12), Austria ($9.05), UK ($8.78), Switzerland ($8.64), France ($8.62), Germany ($8.49) pay the most.
* Mexico ($5.60), Israel ($6.57), Singapore ($7.30) and Spain ($7.34) pay the least.
* Only the top 25 countries were considered for this chart, as the countries which purchase very little on Magnatune have not enough purchases to be statistically significant.
Notes on this chart:
* the USA is responsible for 64% of our sales (CDs and downloads combined).
* the UK, Canada, Germany and France are the most significant non-USA countries.
* The other countries account for 17% of our sales.
* this chart is all our music sales, and includes both downloads and CDs.
Notes on this chart:
* downloads are slightly more country-diverse, with the USA being responsible for 61% of Magnatune's sales.
* More non-USA countries are significant (39%) in our download market.
Notes on this chart:
* 85% of CD Sales go the USA
* Very few countries are significant sources of CD sales.
* Perhaps CD sales are lower to Europe (and Canada) because of higher price on Magnatune ($4 more) to ship CDs there. This would imply that we could double non-USA CDs sales (to download-like levels) if we could ship them at USA postal rates (perhaps by manufacturing them in Europe).
Gender Analysis of Magnatune SalesBelow you'll find two charts analyzing Magnatune's sales along gender lines.
Magnatune's checkout system doesn't actually ask for the gender of a buyer, so I had to manufacture the gender data from the name people give when buying.
The US Census web site provides flat text files of the most common men's and women's first names. I wrote a program to look up each Magnatune purchaser's name against the US Census first names list, and thus assign a gender. It's suprising how many names appear in both genders (ie, Leslie, Kelly). Oddly, some clearly male names (John, Steven) appear as infrequent female names. I assumed that these were errors in the census data, and corrected those by hand. Other conflicts, I resolved by hand. The remaining 100 or so names that were in both lists I left in the 'male' category as a visual spot-check seemed to confirm they were mostly male. In retrospect, I should have used the "frequency" percent which the census provides to decide what gender a conflicting name was mostly likely to fall into, but I think my method is mostly accurate. Given that the US is fairly multi-cultural, as is Magnatune's customer base, the Census names cover a wide range of cultures (many Asian first names appear) and is likely relevant in this case.
How much data did I have to work with? 94% of Magnatune purchasers give their name, and of those, the Census files identified 80% of them as belonging to a specific gender. In other words, 75% of purchases were successfully identified with a gender.
Without further ado, here is the first chart:
Notes on the chart:
* Men account for 89% of Magnatune sales, women account for 11%
* the gender sales gap is narrowing, from 91% male/9% female a year ago to 84% male/16% female this month (November 2004)
* since November, the gender gap has narrowed quicker (perhaps because of the Christmas season, or Magnatune introducing physical CDs)
I also wanted to see if men or women paid more for music (since Magnatune allows buyers to choose how much they want to pay, from $5 to $18), and so I charted average purchase price by gender:
Notes on the chart:
* there is a slight difference, with an average of $8.23 paid by men, versus $8.14 by women.
* the "gender pricing gap" is widening, although this widening is entirely due to the past 3 months, which may be Christmas or the physical-CD-introduction at Magnatune.
* With the past 3 months removed, the price women pay was increasing significantly, as was the price men pay, though at a less steep curve. We will have to see if the past 3 months are a permanent or seasonal trend.
Notes on how I make my chartsI've received a few email inquiries asking about the software I use to make these charts.
* I've recently switched to Apple's Keynote software, which makes very tasteful, attractive charts.
* I do all my number-crunching in Excel (including using Excel's very nice "trendline" feature) though usually some initial massaging occurs in Microsoft Access (I haven't found a good Access replacement on OSX, for talking to MySQL (ie, something which helps me write SQL queries and outputs spreadsheets that can go into Excel).
* I use a freeware program called "XL2K" to automatically move my Excel data into Keynote.
* I recently purchase the complete set of design "themes" from Keynote Theme Park and am using their Da Vinci theme as the base for my charts. However, I remove the colored-paper background from the themes, attractive as they are, so that people can more easily print my charts on laser printers.
* When appropriate, I change the colors on the lines to reflect cultural norms, such as using blue lines for male and pink for female.
* The Average and Trend lines are hand drawn based on what Excel is showing me.
* Annoying, when Keynote exports its charts to other formats it does so without drop shadows (perhaps so they print better), so I am screen capturing the Keynote charts (with Command-Shift-4) at their largest screen size, resizing in Photoshop with "magic wand/invert selection/crop/resize-to-450-wide".
Paypal versus credit card sales: how do people pay?The chart below shows what percentage of sales at Magnatune come in from credit card payments, versus paypal payments.
Some notes on this data:
* Paypal is responsible for 28% of sales, credit cards are 72%
* the percentages do vary +/- 10%, which is more variation than I would have expected
* In the past 2 months, credit card sales have increased almost 15%. I am not sure if this is due to our selling compact discs now (and thus drawing a less internet-insider crowd) or if this is due to christmas purchasing.
* Paypal tells me that in a typical business they see 10% of sales going to paypal. They don't understand either, why Magnatune's Paypal percentage is almost 3X the norm. I suspect it's because music downloading is a more cutting edge type of e-commerce, and thus tends to attract a more internet-insider (and thus paypal-savvy) audience.
Magnatune PR News* I've hired UK PR Firm Chocolate Communications to represent Magnatune in the UK, and if everything goes well, the rest of Europe as well. As a freelancer, my account manager Gavin Spicer was involved in the iTunes UK launch, and the significant other of the principal (Jill Coomber) is in a big-name UK electronica band, so she knows the music industry (and it's evil nature) quite well. I'm happy to have a *clueful* PR agency here in London, so I don't have to have conversions like "Oh gee, the music industry doesn't pay musicians, I didn't know that." (FYI, in the USA, I'm represented by Landis PR)
* for some reason, we're getting more press attention lately (it's been pretty dry for 6 months). Landis PR hasn't started their work yet (they start in January, when Chocolate starts too) so I can't really explain the surge in interest, though obviously I welcome it. Below is some of the stuff going on right now:
* Sound on Sound magazine mentions us in this month's issue, and are working on a larger writeup
* the Washington Post is running a story on the Creative Commons, with us as (I believe) the main feature
* a freelance journalist who's written for Wired (among others) is writing a feature article about Magnatune, and has interviewed The Wretch, Artemis, and Asteria for the feature.
* KRON TV (a San Francisco station) is filming a segment on us.
As a public web site, we live and die by what the world thinks of us (and if they know we exist at all), so the trend is good, and hopefully with UK and USA PR firms kicking in, we'll see the interest increase quite a bit.
"Pay what you want" results analyzed at MagnatuneMagnatune allows buyers to set the price they pay for album downloads, from $5 to $18 with a recommended price of $8. When first exposed to this idea, people often ask "why would anyone pay more than $5 ?" Here are some thoughts I have on the Magnatune variable-pricing scheme:
- because 50% goes to the artist, people pay more for music they enjoy to ensure that the musician can continue to afford making the music.
- people want to feel good about their interaction with the commercial world (i.e. The Body Shop, Ben & Jerry's Ice Cream), and with the music business being one of the most exploitative, wish to reward a business who isn't evil.
- offering to let the price go lower shows that you trust the buyer, and they may respond to that trust by paying more. If the price range were $8 to $18 (i.e. no low-cost option), I suspect that fewer people would move the price up, or that fewer purchases would ensure, though I recognisize that point is debatable and I'm tempted to try an experiment sometime to see.
- as it happens, the average price people pay is $8.20, even though they could pay less. The people who pay less than we recommend are more than balanced out by those who pay more.
I wanted to see if the average price paid per album fluctuated much week-to-week, so I charted both downloads and physical CD purchases (CD purchases started in september 2004, so I only have a few month's data for them).
Here is what the chart looks like:
Two facts stand out from this chart:
1) the average price doesn't fluctuate much, no more than 10%, around the $8.20 average
2) people elect to pay almost $1 more (on average) when they're buying music for a CD. Note that the order process at Magnatune separates the price of the CD ($4.97 USA or $7.97 elsewhere) from the music price (recommended at $8). I'm not sure why people pay more for music when they're buying a CD from us.
The other question I often get asked is "how many people pay $5 ?"
I made two graphs, the first showing the number of album purchases at each price point:
The chart shows that 14% of albums sold are at $5, while suprisingly, 14% are at $10, while the majority (56%) sell at the recommended price of $8.
The chart below shows the same data, but instead of number of purchases, it shows total revenue at each price point.
This is important, because 1 purchase at $10 is equivalent in revenue to 2 purchases at $5. As a virtual business, I'm most interested in maximizing revenue rather than units sold (in fact, as a busines I want to minimize units [as each unit has a labor cost] as much as possible while still maximizing revenue), so this chart is perhaps more important.
What's interesting about this chart is that it shows:
1) Almost twice as much total revenue comes from people paying $10 (17%) as people paying $5 (9%).
2) 31% of Magnatune's total revenue comes from people paying more than $8, vs. just 16% of revenue for purchases less than $8. I think this makes a case for accepting the minor reduced revenue from allowing people to go below $8 as those who pay more make up for that loss.
3) Purchases center around the round numbers of $5, $8 and $10. Evidently, these numbers "feel good" to people.
Here is what the raw data look like (click to download a CSV file)
Summary of Magnatune sales so far - sales by artists and albumsAs part of my end-of-the-year "How is Magnatune doing" analysis, below you'll find sales totals, summed per-artist and per-album and some analysis of "what it all means".
Sales by artist, since Magnatune started 18 months ago:
Sales by album, since Magnatune started 18 months ago:
A few things the data tell me:
1) Magnatune's sales don't follow the 80/20 rule (where 80% of the revenue comes from 20% of the artists) - our sales are more evenly distributed. Maybe because this is because we have no "star" artists but I like to think it's because all our artists are on an level playing field, with visitors equally able to listen to any artist, and all the music being randomly shuffled in the playlists.
2) We have a "long tail" -- Wired Magazine's important article The Long Tail argues that there is wealth to be made with a catalog of wide breadth. While our catalog is small in comparison to Amazon's we still see a long-tail effect, with 42% of album sales revenue coming from the bottom 60% in sales rank, and only 20% of sales revenue from our top 10 albums.
3) The top grossing artists are usually those having multiple albums. Notice that in the per-artist chart, revenue is more top-heavy, with the top 10 artists taking 27% of revenue (vs 20% for the top 10 albums).
4) If this data were graphed as a bell curve, the curve should be short and wide. I looked up How to do a bell curve in Excel but it's beyond my skills. If someone feels like crunching the raw numbers and supplying me with additional charges, I'd be much obliged.
FYI, music licenses show a similar pattern, with no single artist having more than 4 commercial use music license deals yet, but I don't have enough data with that yet to provide meaningful charts.
Free Download offer: way more successful than we expectedIn September 2004, Magnatune offered all its customers a free album download of their choice.
- the email went to 5670 people (all the customers from 18 months for whom we had valid email addresses).
- 31% (1802 people) choose and downloaded an album. This is an amazing response rate, almost unheard of in the direct mail business.
- 270 albums were chosen at least once for download (out of 304 albums, so 89% of the catalog was downloaded)
- the top choice accounted for 2.6% of the total, while the top 10 choices accounted for only 17.4% and the top 20 was 28.5%. This indicates that our customer base has a wide variety of tastes and Magnatune's current success is due to the breadth of its catalog rather than one or two standout artists.
- the top 10 "free download" choices were: Lara St John, Ehren Starks, Falling You, Drop Trio, Da Camera, Jami Sieber, Magnatune New Age Compilation, Beth Quist, Magnatune Electronica Compilation. Compare this to the top 10 purchases this month at http://magnatune.com/info/stats/best_selling_this_month - notice the similarities, but also the differences, such as more compilations downloaded for free.
- sales increased by about 50% for the month following the free offer,
as you can see here (week 2004:39)
(note: the dip in sales the week of 2004:41 was due to a Denial-of-Service attack against us which took a few days for us to stop, after which sales spiked up again)
- a spreadsheet showing the detailed choices from the free album download is available here: http://magnatune.com/info/images/freedl.csv
Sales summary for 2004: CD sales set to overtake downloads, licensing up, consumer sales flatAs you can see here:
CD sales continue to increase as a percentage of total sales, and now represent 45% of Magnatune's total consumer sales. CDs are set to overtake downloads in the coming weeks. This is good, as it indicates that our customers prefer CDs, so they should be happier customers now that we offer them physical CDs to purchase.
However, the increase in CD sales comes at the expense of downloads,
and so overall consumer sales (cd+downloads) has not grown, as you can
The good news is that licensing has more than doubled recently, as you
can see here:
We're doing a large number of music licenses to indie films (10 to 20 films per month), at the bargain rate of $44 for a festival license. The upside is if any of these films showing at film festivals get "picked up" for worldwide distribution, additional license revenue comes in, and with the number of films we're licensing to, odds are good this will start kicking in over the next 2 years. And obviously, films with our music exposes the music to ever larger audiences.
Academic analyzes Magnatune's variable-pricing schemeTobias Regner, an economics researcher at Imperial College London, is researching what occurs when people are offered to control how much to pay for a product, which is what Magnatune offers. We shared our purchase data with him (anonymized, of course) and he came up with these preliminary insights:
- The average purchase amount is $8.20
- Customers who buy an album, but then never again, tend to spend much on that: $8.29 On the contrary returning customers have spent only $7.64 on their first purchase on average. Some people might have just tried out the service and spent little. Once they got convinced with it they started to spend more. Some first time customers might have overspent and now stay away.
- The average amount increases with the number of purchase of a customer. First purchase averages $8.06 vs Purchases number 7+ average $8.66.
- The average amount decreases with the total of purchases a customer made. Average amount of one-time purchases is $8.29, which means returning customers' first purchase averages $7.64
- 2 genre are significant in affecting average price: Genre 'Ambient' has a negative impact on amount, while 'Rock' has a positive impact.
- Customers who leave an email tend to spend more on a purchase. With email: $8.24 vs without email: $7.40
These are initial results. Tobias plans on having further results once he's conducted further analysis.