tag:blog.kedrosky.com,2014:/feedPaul Kedrosky2015-04-14T12:42:07-07:00Paul Kedroskyhttp://blog.kedrosky.comSvbtle.comtag:blog.kedrosky.com,2014:Post/googles-microsoft-moment-when2015-04-14T12:42:07-07:002015-04-14T12:42:07-07:00Google, Microsoft, Stall Points, and Growth<p>April 7, 2015</p>
<p>When might Google have its Microsoft moment? That is, when will it begin its inexorable decline – as most aging tech companies do when their growth stalls, and as Microsoft did – largely unbeknownst to (and even denied by) most observers? How might we know?</p>
<p>I have a few things I look for when thinking about inflection points in aging technology companies. Not all of these conditions need to be satisfied, but at least some of them do. </p>
<p>There are five major parts of the argument:</p>
<p><strong>1. Stock compensation loses its luster</strong></p>
<p><strong>2. Declining growth</strong></p>
<p><strong>3. No longer hiring for innovation</strong></p>
<p><strong>4. Price & margin pressure in core business</strong></p>
<p><strong>5. Government action</strong></p>
<p><strong>——————————————————————————–</strong></p>
<p>Let’s start:</p>
<p><strong>1) Stock compensation loses its luster</strong></p>
<p>Generally speaking, technology companies retain their best employees with equity compensation more so than with salary. And when that equity compensation – stock options and/or equity grants (usually restricted stock units) – stops paying out, the best employees (especially engineers) often start looking elsewhere. </p>
<p>One way of judging this to look at when Google’s stock stopped paying off for new employees. For the sake of argument, let’s say that’s when the company’s share price stopped appreciating in a meaningful way. You can see that for over a year now – the longest time in Google public history – it’s stock has flat-lined, after more than a decade of producing huge gains for employees. </p>
<p><a href="https://svbtleusercontent.com/2ihthwrngo9lfq.jpg"><img src="https://svbtleusercontent.com/2ihthwrngo9lfq_small.jpg" alt="Google’s Microsoft Moment_ When?.jpg"></a></p>
<p>Something similar happened to Microsoft in 2001, and it was years before equity compensation there partially regained some allure. In the interim, many good employees left, whether because they had made their money and were done, or because they knew they wouldn’t in future. At the same time, Microsoft couldn’t use its equity currency as effectively any more to buy employees, which raised the cost and difficulty of hiring.</p>
<p>There is a twist here. Restricted stock units (RSUs, which Google calls GSUs – ick) are currently Google’s main variety of non-executive employee equity compensation,and they aren’t lottery tickets, like options. They are worth something at any combination of grant and post-vesting trading price. That means Google’s RSU grants are still worth something, no matter how poorly the stock performs. </p>
<p>Doesn’t the above mean that Google’s equity compensation, insofar as it’s now skewed to RSUs, is still highly valuable? Yes and no. Yes, because RSUs are, by definition (and unlike options), always worth something. But also No, because the payoffs, while more predictable, are also lower. And that has an effect, at least at the margin, in terms of the kinds of employees it attracts. More on this in 3) below. </p>
<p><strong>2) Declining growth</strong></p>
<p>Growth companies are magic things, right up until they aren’t. There is data out there showing that, with rare exceptions (like Apple), when a technology company slows down below 10% or so, it is rare to re-accelerate again. Investors notice, and they revalue accordingly. </p>
<p>The following graphic is from from a <a href="https://hbr.org/2008/03/when-growth-stalls">2008 article</a> in HBR. It looked at “stall points” in Fortune 100 and Global 100 companies during the period 1953 to 2006. Broadly, once stalled, a company stays stalled. Further, it, on average, loses 76% of its market value once investors realize the growth genie is gone.</p>
<p><a href="https://svbtleusercontent.com/lzhcaezos0720w.jpg"><img src="https://svbtleusercontent.com/lzhcaezos0720w_small.jpg" alt="R0803C_A.gif (585×138).jpg"></a> </p>
<p>When does a growth company cease being a growth company? When investor say it does, of course. But it’s worth noting when annual revenue growth rates fall back from historical levels to ones more in line with other technology companies deemed to be mature, like Microsoft.</p>
<p>As the following chart shows, that is roughly what has happened with Google. Its year-over-year revenue growth rate is now roughly in line with Microsoft, which is hardly the peer group to which Google would like to belong. </p>
<p><a href="https://svbtleusercontent.com/qmxlyopaxgjp9q.jpg"><img src="https://svbtleusercontent.com/qmxlyopaxgjp9q_small.jpg" alt="2-BLOOMBERG-1.jpg"></a></p>
<p><strong>3) Hiring for risk management, not innovation</strong></p>
<p>Something that happens in technology companies as they grow is they often stop innovating. That is, in part, because they’ve run out of ideas. but it’s also because they’re reluctant to hire innovators. Those people are a pain in the ass at any time, and more so than ever in large companies with many rules and procedures. </p>
<p>One indicator – and it’s far from definitive – is how well a company hires for innovation. While Google is rightly lauded for hiring smart engineers, and for having internal programs that encourage them to experiment and try new things, it’s not clear how well that is working anymore. The <a href="http://tech.slashdot.org/story/15/03/06/1837225/the-abandoned-google-project-memorial-page">list of botched & abandoned Google projects</a> grows ever-longer, and the list of business segments contributing meaningfully to revenue hasn’t changed much since inception. </p>
<p><a href="https://svbtleusercontent.com/tiioeavwrrkgcw.jpg"><img src="https://svbtleusercontent.com/tiioeavwrrkgcw_small.jpg" alt="2-BLOOMBERG-5.jpg"></a></p>
<p>There is an interesting, related discussion <a href="http://www.quora.com/Why-have-so-few-successful-startups-come-out-of-Google">on Quora</a> about this subject. It is about the risk aversion in hiring at Google, and how that has affected innovation at the company, especially with respect to how many startups it produces. </p>
<p>Here is a year-old quote from that discussion, from venture investor and entrepreneur Keith Rabois:</p>
<blockquote>
<p>“… after Google became successful, the type of candidate who applied and was hired shifted from the entrepreneurial to the smart yet homogeneous type. (Shift was pronounced by 2005.) As I have observed previously, only disruptive people create disruptive companies. (Stated differently, great entrepreneurs do not tolerate rules and constraints very well). Google has screened out personalities of this sort since at least 2004 and maybe since 2002.”</p>
</blockquote>
<p>This has been going on for some time, of course, and it has hardly destroyed the company. It is, however, another data point worth considering. After all something similar happened at Microsoft in the late-1990s, ably chronicled after the fact by sites like Mini MSFT in <a href="http://minimsft.blogspot.com/2004/07/microsoft-stack-rank-as-popularity.html">posts like this one</a> about stacked ranking.</p>
<p>It is worth pointing out here that Google continues to spend aggressively on R&D, as the following graphic of R&D as a percent of sales show. It has hardly ceded the innovation field, at least speaking from the point of view of research expenditures. </p>
<p><a href="https://svbtleusercontent.com/belvaceazd5bjq.jpg"><img src="https://svbtleusercontent.com/belvaceazd5bjq_small.jpg" alt="2-BLOOMBERG-6.jpg"></a></p>
<p>Then again, Microsoft’s R&D as % of sales peaked in 2004, just before it finally became obvious it was no longer a growth company. Its share price then flat-lined for most of the 2000s.</p>
<p><strong>4) Price & margin pressure in core business</strong></p>
<p>When technology companies cease to innovate, and face increasingly competitive pressure in their core business, it is inevitable margins come under pressure. We saw that happen with Microsoft more than a decade ago, and we see it to some degree now at Google. The advertising business is increasingly moving away from the desktop, where Google has a de facto monopoly built around search, to mobile and apps, where Google is just one of many (even if its Android is formidable). </p>
<p>We can see some of what that has caused in the following graph of Google’s operating margin over the last decade. It’s hardly collapsed, but it isn’t expanding either – it’s fallen significantly from its low-30s peak of the 2000s. </p>
<p><a href="https://svbtleusercontent.com/iakaoec4zmudia.jpg"><img src="https://svbtleusercontent.com/iakaoec4zmudia_small.jpg" alt="2-BLOOMBERG.jpg"></a></p>
<p>IBM and, later, Microsoft, responded to this margin pressure by moving into services and enterprise business. You can expect Google, a highly consumer-centric company, to increasingly do the same. Of course, like Microsoft and IBM, it may provide a floor for margins, but, as history shows, it almost certainly won’t turn the company back into a growth firm again.</p>
<p><strong>5) Government action</strong></p>
<p>There is nothing that spoils a good growth story party for a quasi-monopolist quite like antitrust action. Rather than revenue growing, it is the headcount of in-house lawyers that grows, as does risk aversion and commensurate wariness about acquisitions. </p>
<p>People will have forgotten, but Microsoft’s adventures in antitrust began in March of 1998, and was decided in June of 2001. Microsoft was never really the same afterwards, with it increasingly obvious that it was nervous about looking aggressive in its markets, as well as about doing any sort of large acquisitions. As a result, a generation of companies could thrive relatively unimpeded, like Google, Apple, VMware, and others. By the time Microsoft got more aggressive again, years later, its growth mojo was long lost. </p>
<p>Something similar is arguably beginning to happen at Google. It faces <a href="http://www.wsj.com/articles/eu-to-file-antitrust-charges-against-google-1429039881?mod=djemalertTECH">antitrust action in the EU</a>, which could presage more action in the U.S. While it hasn’t stopped the company in its tracks, as eventually happened with Microsoft, it is almost certainly a new factor in the company’s decision-making. We can’t know how Google’s legal staff has expanded, or how big it is, but it has become <a href="http://time.com/3677301/google-lobbying-comcast/">the largest lobbyist</a> among technology firms, and its <a href="https://www.google.com/about/careers/search#t=sq&q=j&jl=Mountain%2520View%2CCA&jc=LEGAL">current legal requisitions</a> aren’t just business process related. </p>
<p><strong>——————————————————————————–</strong></p>
<p>It’s worth considering that we are at some sort of inflection point in Google’s business. Is it a “Microsoft Moment”? I have no idea. The parallels are there, even if they aren’t and could never be 1:1. Nevertheless, it’s at least interesting the pressures on Google.</p>
<p>Of course, if and when it’s all clear, it’s too late. Like many things in economic life, decline happens at first slowly, and then all at once – and is usually missed until it’s already happened, and then, of course, it is declared obvious.</p>
tag:blog.kedrosky.com,2014:Post/playing-with-pi2015-04-03T19:53:57-07:002015-04-03T19:53:57-07:00Pi Play: Tracking Home Broadband Speed<p>I’ve been messing about with our <a href="http://www.raspberrypi.org/raspberry-pi-2-on-sale/">Raspberry Pi 2</a> here in recent weeks. My first project was an SSH honeypot, built around the open-source honeypot <a href="https://github.com/desaster/kippo">kippo</a>. I’ll post something on it in the next week or so. It was … entertaining.</p>
<p>The second, ongoing project is tracking and analyzing home broadband speed. I pay for Time Warner Ultimate Max XL Something Or Another, which is supposed to be 30 mbps down, and 5 mbps up. About 70% of the time that seems to be the case when I check, but at least 30% of the time it feel like it tests lower.</p>
<p>I generally use <a href="http://www.speedtest.net/">Ookla</a> to test speeds, but that’s only when I remember, so I have no good data. Is it as bad as I think? Or am I just testing at bad times? Or am I only remembering the poor speeds? I really have no sense how bandwidth trends over the course of the day here – Time-Warner sure doesn’t provide it. </p>
<p>Enter my Raspberry Pi. I had long ago written a script to track home bandwidth, but it was a hack that ran on my router. (Trust me: Don’t do that.) This time I decided to host the scripts on the Pi, and store the data there too. I’m a Python guy, (insofar as I have any programming skills whatsoever, which I don’t) so that part was easy. </p>
<p>After some futzing about, I came up with this simple approach. I would use the <a href="https://github.com/sivel/speedtest-cli">speedtest-cli</a> open source tool to grab hourly broadband data. I would store that in a text file. I would then use Python to take the data from the file and store it in <a href="https://www.sqlite.org/">sqlite</a>.</p>
<p>The script for grabbing the data & running the tests isn’t mine. I just called the speedtest script (installed as above) from a cron process on the Pi. Here is the line from cron:</p>
<pre><code class="prettyprint">0 */1 * * * /usr/bin/python /usr/local/lib/python2.7/dist-packages/speedtest_cli.py --simple > /home/paul/speed.txt
</code></pre>
<p>In cron lingo, I’m calling the script every hour on the hour, and I’m using the “simple” flag for speedtest-cli . That mean it only returns three things: ping time, download speed, and upload speed. The data is then “piped” to the file “speed.txt” in my home folder. Its contents are overwritten every hour, & they look like this:</p>
<pre><code class="prettyprint">Ping: 24.903 ms
Download: 31.74 Mbit/s
Upload: 5.39 Mbit/s
</code></pre>
<p>I then created a second Python script to scrape that data file with a regular expression. It extracts the numbers, stores them in a database, and then updates a graph. I’ll give a brief overview of each of those steps in turn. Note first that this second script runs five minutes after the above one. </p>
<p>Here is the cron entry:</p>
<pre><code class="prettyprint">5 */1 * * * /usr/bin/python /home/paul/st.py >>/home/paul/st.log 2>&1
</code></pre>
<p>1) <strong>Extract the data</strong>. It’s a simple regular expression. I didn’t even bother to put in a loop – it just wasn’t worth it. I had to import the relevant Python packages first:</p>
<pre><code class="prettyprint lang-python">import re
import sqlite3 as lite
import sys
import time
import sqlite3 as lite
import plotly.plotly as py
import plotly
from plotly.graph_objs import *
import pandas as pd
</code></pre>
<p>And now extract the data:</p>
<pre><code class="prettyprint lang-python">lines = open('/home/paul/speed.txt').read().splitlines()
p = re.match("Ping: (.*?) ms", lines[0])
d = re.match("Download: (.*?) Mbit/s", lines[1])
u = re.match("Upload: (.*?) Mbit/s", lines[2])
</code></pre>
<p>2) <strong>Store the data in a database</strong>. I used sqlite because the data was so simple, and there was no need for concurrent users, etc. I won’t bore you with creating the data tables, etc. In addition to the three data fields, I have an auto-incrementing “id” field, plus a “sqltime” field that automatically stores the time (in GMT) of the latest update. </p>
<pre><code class="prettyprint lang-python">con = lite.connect('/home/paul/tw-speed.db')
with con:
cur = con.cursor()
cur.execute('''INSERT INTO data(ping,download,upload) VALUES(?,?,?)''',(p.group(1),d.group(1),u.group(1)))
</code></pre>
<p>3) <strong>Update the graph</strong>. You could stop at the preceding step. After all, you’ve got the data, and you’re updating it with more data hourly. But it’s not much fun being unable to see the trends, so … a graph is required, at least for me. I’m a big fan of <a href="https://plot.ly/">Plotly</a>, a hosted graphing service, so I used it to host my graph. </p>
<p>First, extract the data back out of the sqlite database and put it in a Pandas dataframe:</p>
<pre><code class="prettyprint lang-python">with con:
cur = con.cursor()
cur.execute('SELECT id,ping,download,upload,datetime(sqltime, "localtime") FROM data')
rows = cur.fetchall()
df = pd.DataFrame( [[ij for ij in i] for i in rows] )
df.rename(columns={0: 'id', 1: 'Ping', 2: 'Download', 3: 'Upload', 4:'Date'}, inplace=True);
df = df.sort(['Date'], ascending=[1]);
</code></pre>
<p>Next, set up the look and structure of the chart. This is all Plotly syntax, and it’s well-documented on the Plotly site:</p>
<pre><code class="prettyprint lang-python">trace1 = Scatter(
x=df['Date'],
y=df['Download'],
name='Download',
)
trace2 = Scatter(
x=df['Date'],
y=df['Upload'],
name='Upload',
yaxis='y2'
)
layout = Layout(
title="Time Warner Broadband Speed",
xaxis=XAxis( title='Date' ),
yaxis=YAxis(
title='Download speed (mb/s)',
range=[0,35]
),
yaxis2=YAxis(
title='Upload speed (mb/s)',
range=[0,6],
titlefont=Font(
color='rgb(148, 103, 189)'
),
tickfont=Font(
color='rgb(148, 103, 189)'
),
overlaying='y',
side='right'
)
)
</code></pre>
<p>Finally, update the graph with the latest data. Make sure you enter your Plotly username and key. There is one issue here though. Unless you make your graph public, only you will be able to see the graph you create, if you embed it somewhere other than Plotly’s site. (I have mine set private, and so it can only be seen on our internal network.) </p>
<pre><code class="prettyprint lang-python">py.sign_in('username', 'key')
data = Data([trace1,trace2])
fig = Figure(data=data, layout=layout)
py.plot(fig, filename='broadband speeds',world_readable=False)
</code></pre>
<p>And that’s about it. Here is my graph, and it gets updated hourly by the scripts. You can see interesting trends, like how bandwidth declines every evening as my neighbors go mad with Netflix and other streaming services. You can also see, more importantly, that my recent bandwidth, while varying, has been acceptable. Curse you Time-Warner for providing what you said you would – but I won’t stop watching you.</p>
<p><a href="https://svbtleusercontent.com/zgnhqybxro2mdq.jpg"><img src="https://svbtleusercontent.com/zgnhqybxro2mdq_small.jpg" alt="obelix.local_bw.html.jpg"></a></p>
<p>Have fun. Enjoy tracking your bandwidth.</p>
tag:blog.kedrosky.com,2014:Post/a-thought-experiment2015-03-03T10:43:29-08:002015-03-03T10:43:29-08:00Tech Bubbles, FOMO & Revenge Effects<p>What are we missing in all the chatter about fear of missing out (FOMO), scarcity, and the current technology bubble among late-stage private companies? I’ll posit a few things, some obvious, some less so, and a couple deranged and hypothetical:</p>
<ol>
<li>Lots of people took this 2001 bumper sticker very seriously: <a href="https://svbtleusercontent.com/pc5zzxwixiimcg.jpg"><img src="https://svbtleusercontent.com/pc5zzxwixiimcg_small.jpg" alt="please give me one more bubble bumper sticker - Google Search.jpg"></a> Knowing that most of the money is made by being early to prominent tech IPOs, they did the rational thing: They bought earlier. In this case, not just early, but pre-IPO – in late-stage, private markets. One reason we are seeing this incredible rush to buy overpriced companies in private markets is fear of missing out on the appreciation when the companies come public. As a result, the post-IPO price appreciation is happening pre-IPO, making private companies back-door public – except, as <a href="http://abovethecrowd.com/2015/02/25/investors-beware/">Bill Gurley has written</a>, in the ways that matter most, like liquidity, quality of financials, oversight, etc. You know, little things.</li>
<li>A bubble in equities is generally less destructive than a bubble in debt – there are fewer claims on real assets, and thus less collateral to be sold – but a bubble in private equity markets is also very different from a bubble in public equity markets. There is less liquidity, so price declines tend to be sharper, more surprising, and less continuous. <a href="https://svbtleusercontent.com/yad0ovkmhtghgg.jpg"><img src="https://svbtleusercontent.com/yad0ovkmhtghgg_small.jpg" alt="wile coyote running in air - Google Search.jpg"></a> We should not confuse the new ability to sometimes buy and sell private company shares with the existence of a liquid and well-functioning market in those shares. Liquidity disappears from even the most liquid markets when it’s most needed. What happens when liquidity was never really there in the first place?</li>
<li>The related distortions make it difficult to anticipate knock-on effects. For example, consider the growth in securities lending, especially in so-called specialty names with high scarcity. Markit says there are now $15.3 trillion of securities available for lending globally, with more than $2.0 trillion on loan on a typical day. Low interest rates in developed economies has an increasing number of institutions trying to cover shortfalls on the fixed income side of their portfolio by building large securities lending businesses, especially in securities with high scarcity premiums. That, in part, explains why some funds have used early positions in pre-IPO companies to drive a highly profitable post-IPO securities lending business. They use the companies’ small floats to earn high rebates for lending “specialty” equities out again – look at the 100% negative rebates out there – and then reinvest the proceeds. But how well are the proceeds invested? What are the default risks at the borrower in event of a jump change in price? There are some very strange things going on in securities lending, with negative rebates abounding. Who knows what defaults might look like in future, or whether lending proceeds have been reinvested intelligently?</li>
<li>Revenge effects in technology markets can be very strong, as Edward Tenner has <a href="http://www.amazon.com/Why-Things-Bite-Back-Consequences/dp/0679747567">written</a>, but revenge effects also exist in technology <em>capital</em> markets. The interplay between capital markets and technology creates myriad new opportunities for risk creation <em>and</em> risk obfuscation, much of which is justified, at least in part, by the wondrousness of technologies and the jobs created. This breeds a species of hucksterism, like the oft-repeated claim that “We overestimate technology’s short-run effects, but under-estimate its long-run effects”. This is, of course, just true enough to be dangerous, as ever. </li>
<li>Will this end badly? It doesn’t need to, and it may not, but it would be surprising if it didn’t. Like prior bubbles, the catalyst will be something largely anodyne, perhaps a modest public market repricing causing collateral issues as illiquidity drives a sharper late-stage private market repricing. There are some superficial similarities here with what happened in the late 2000s in illiquid collateralized credit instruments (even if this is equity, not debt), where the repricings rippled through primary and then second markets, causing collateral sales, thus forcing more securities sales at sharply lower levels, and so on.</li>
<li>Another problem with these pre-IPO, post-IPO (PPIPO) companies is the inevitable accumulation and calcification of preferences & related terms in their capitalization table. Each round has its own preferences, sitting on top of the round before it, generally with terms uniquely tied to that round’s participants. The ordinary way to unwind this mess is to, in the process of going public, collapse all the preferred shares into a single class of common shares, wiping out the preferences. In the absence of doing that, it’s a mess, one that encourages predatory types, like private equity firms, to show up and <a href="http://www.bloombergview.com/articles/2015-03-03/can-private-equity-rescue-silicon-valley-s-startups-">offer silly & unnecessary methods</a> by which they can <del>screw people over &</del> solve the problem. This would, of course, in most cases lead to much lower prices, if you could make it happen at all given the absence of cash flow at most pre-IPO, post-IPO tech companies.</li>
<li>8 o'clock, 8:30 in Newfoundland. </li>
</ol>
tag:blog.kedrosky.com,2014:Post/like-seemingly-happened-with-many-people-the-following-tweet-got-me-thinking2015-02-12T09:19:18-08:002015-02-12T09:19:18-08:00Music and Innovation and Teams <p>The following tweet after the Grammy music awards got me thinking:</p>
<p><a href="https://svbtleusercontent.com/ralib5mtadhnna.jpg"><img src="https://svbtleusercontent.com/ralib5mtadhnna_small.jpg" alt="Producers, writers, and teams"></a></p>
<p>Granted, I know zero about music. And I know even less about the Grammys. But the idea that albums up for record of the year had such dramatic differences in the number of people behind them was striking. </p>
<p>Leaving aside the sniping about whether larger producer/writing teams is somehow discrediting – some argued that Beck was a more legitimate winner because he was the auteur, unlike with Beyoncé – I was curious whether this was new. Innovation has become a team activity in many areas of economic and artistic life, so is that also the case with music? </p>
<p>I decided to look at the annual list of top-selling albums back to the 1960s. How many writers & producers were behind them? Sadly, it wasn’t easy to find how many songwriters were involved, but it was fairly straightforward to scrape <a href="http://en.wikipedia.org/wiki/List_of_best-selling_albums_by_year_in_the_United_States">Wikipedia data</a> on the number of producers.<sup id="fnref1"><a href="#fn1">1</a></sup></p>
<p>Here is a graph of producers per bestselling album back to the 1960s. It’s a bit messy, so I’ve added a five-year rolling average as well.<sup id="fnref2"><a href="#fn2">2</a></sup></p>
<p><a href="https://svbtleusercontent.com/9zut66sdtjxe9w.jpg"><img src="https://svbtleusercontent.com/9zut66sdtjxe9w_small.jpg" alt="Microsoft Excel.jpg"></a></p>
<p>The number of producing credits per bestselling album has roughly tripled over the last four decades. Granted, some of that may just be compensating for undercounting in the past (see footnote 1 below), and some seems also to be an artifact of the way producing credits are handed out like bon-bons on rap records, but it’s still intriguing. At least superficially it suggests that music, like so many innovative activities, is becoming increasingly a group exercise, even at the highest levels of achievement<sup id="fnref3"><a href="#fn3">3</a></sup>. </p>
<div class="footnotes">
<hr>
<ol>
<li id="fn1">
<p>Counting producers can be more than slightly dodgy. People are sometimes given music producing credits for doing very little; and sometimes de facto producers are denied a producing credit because the artist’s ego doesn’t allow it. In other words, this isn’t physics. <a href="#fnref1">↩</a></p>
</li>
<li id="fn2">
<p>The bizarro 1993 spike is Whitney Houston’s soundtrack to “The Bodyguard”; the 2008 spike is Lil Wayne’s “The Carter III”. Make of either what you will. <a href="#fnref2">↩</a></p>
</li>
<li id="fn3">
<p>It is a legitimate question, of course, whether an album’s bestselling status marks it as among the “highest levels of achievement” in music. Ask any college radio DJ. They’ll tell you. Over and over and over. And over again. <a href="#fnref3">↩</a></p>
</li>
</ol>
</div>