Apples, not Turtles

This article was triggered by unreliable internet connections, especially after upgrading to Apple’s latest Mac OS X release, Leopard. It’s a reflection on Brand, the Pareto Principle, and Perception, which is why it ties, strangely, to Internet Marketing.

There’s an almost certainly apocryphal story that someone was presenting cosmology to a non-scientific audience. At the end, a (according to most versions that I’ve heard - my guess is that the attribution just adds character) “little old lady” told the presenter that he was wrong, the earth was flat, so what he’d just said was nonsense. Talking the doubter afterwards about what supported the flat earth, she said that the earth stood on the back of giant turtles. In an effort to try and get the flat-earther to think about the implications, the speaker asked “So what do the turtles stand on?”. The reply?

You can’t fool me, young man, it’s Turtles all the way down.

On becoming a MacHead

I’ve gradually migrated to become a MacHead. Initially it was just because I didn’t want to spend time on administering the desktop. I loathed the time I wasted on Windows reboots and efforts wasted trying to diagnose the causes of problems that resulted in annoyances like a pointless uninformational popup window with one option - “OK”; or of losing a network connection and seeing a web page fail to refresh, and two popups to let me know about the already visible problem (popups that refused to accept any settings telling them to shut up and stay out of the way). I’ve used Linux since 1994, and still use Linux on Servers. But for my desktop I’ve gradually grown to rely on my Macs.

My small home network and office networks mostly have Macs. But the router that connects to the internet is different. At the office, I use an Airport Express to the buildings’ shared ADSL connection. The Express has more than enough range (covers the office and the shared presentation rooms) and I don’t need the features of the Airport Extreme base station. At home, I have a choice of no fewer than five ADSL router modems - mostly because every so often my connection will become unreliable, and BT Wholesale will claim the problem is my router. I usually use a NetGear DG834PN.

After upgrading to Leopard I started another bout of unreliable internet connections. As usual, I assumed that the problem was the ADSL reliability and did the exercise of switching around routers, and reboots, etc. Nothing much helped and I resigned myself to losing the connection, typically for a few minutes to an hour or so, but sometimes for several days at a time.

The first serious clue that at least some of the problem was Leopard, was an older iBook. While my newer MacBook Pro would mostly connect and get traffic, the iBook would sometimes spend 24 hours unable to connect to the home network, unless rebooted. That was a new symptom and made me suspicious. I eased the problem by tweaking a config file that extended the DHCP lease interval and allowed a longer period to renegotiate the DHCP lease.

I even made the DHCP lease edits to a MacBook Pro, despite that usually connecting. But there was still this nagging disconnection problem. While looking at a web page, or using an AJAX application with autorefresh, the WiFi connection would just up and vanish or fail to pass traffic to the internet.

This is where we get into the marketing…

Segmentation

NETGEAR didn’t mention any problems with Macs - but then they’ll do most testing on Windows. There’s a rule, useful in marketing, called the “Pareto principle“, after the guy that invented it. However, the rule is usually more abused than correctly observed. Basically the rule is an 80:20 rule. However, the 80% and the 20% are from different domains. That is, you might say “80% of the customers arrive at the web site from 20% of the marketing communication effort” or “80% of the sales come from 20% of the customers”. It’s *not* that 80% is one thing and 20% is the remainder.

Marketeers frequently abuse the Pareto Principle. For example, a classic technique in marketing is to rank all the vendors in a market, and sum up their percentages until you reach 80%. Anything falling under that line is ignored - and the claim is made that this is a Pareto optimisation. However, if the nature of the market has been misunderstood, then it isn’t a Pareto optimisation, it is just misguided.

Take the current problems I’m wrestling with. If I want a new WiFi Router, I don’t care whether it works with Windows. My main concern is my Macs. If it isn’t tested with a recent version of the Mac OS, I simply don’t care about the product. Similarly, in mortgages, there’ll be people with specific interests - in the UK, this may be a specific type of mortgage known as a “Buy To Let” mortgage. Ranking lenders because they do other types of mortgage, is irrelevant, when I only care about a BTL mortgage. This kind of special limited interest in a market is quite common, in my experience - so lumping together dissimilar markets, and then claiming a Pareto optimisation fails. In the case of WiFi, if the top 80% of vendors fail to test with the Mac, I’m simply not going to buy one of them. I’m not switching from Macs to Windows, just because of a router problem.

However, within the router companies, they probably rank Macs in with Windows systems. The result will be that Macs don’t get (much, if any) testing, or column inches on the website about compatibility. From my perspective, as a customer, this means that I can expect that very few vendors will have checked that their stuff works properly for the Mac - unless they strategically address that market.

Breaking up the audience into parts that should hear your message, and disentangling which precise message should be delivered to someone, is segmentation. It’s a lot more important to get right than you might think, at first - because it affects costs, and conversions.

Let’s assume that your business is in Michigan, and you service racing lawn mowers. You need to address racing lawn mower owners, but only if they live in Michigan. If they live in other states, and your advert reaches them, then the value of your advert - to the consumer - is greatly diminished. Indeed, if someone in Arizona clicks on your advert, they’ll probably be annoyed that you wasted their time, and you’ll have spent money to reach someone who doesn’t care about your service.

So you need adverts to reach only those people most likely to buy. The right audience usually contains lots of the segment that you want to reach.

If your message helps people to recognise that they are being addressed, it can help drive up the CTR, even if the audience is already the right segment. Note that if the audience you reach is significantly larger than the segment that you want, you might *decrease* response with the right advert - but increase the overall conversion rate. In the racing mower example, if we advertised nationally but included the Michigan location in the copy or Display URL, we could decrease responses from people who felt that Michigan was a long way to go for a mower repair - a lower CTR than if we failed to mention Michigan, but a likely higher conversion rate. Part of the trick of effective search advertising is to work out where that balance lies - spread to a wider audience but risk a lower conversion rate, or target a smaller and more focused audience and have a higher conversion rate.

My segment

Back to the networking… I started looking for a solution to my networking blues. I need a WiFi hub, or better a WiFi ADSL router that is tested for Macs, and particularly for Leopard. I knew from my Linux experience that there is a specialised Linux search - probably because Google uses Linux in house. I searched for, and failed to find, a specialised Mac search, and then just guessed at the URL. Still no joy. There is no specialised Mac search - so I have to see a results page that pretty much doesn’t address my needs - guaranteed. Because the optimisation of search results will mean that most answers should be about the dominant OS in this space - Windows - my needs will be relegated to a tiny fraction of answers, many pages in. And when marketeers fail to mention my tiny segment in their copy (what, waste *any* time on something below the magic 80%?) then search engines can’t help - because there’s no clue in the web site copy.

Worse, the adverts don’t offer me what I want, even if I add “mac”, “apple”, “macbook”, etc.

Intriguingly, the Linux specialised search will also gleefully address Windows in adverts. For example, Skype is ported at different rates to Windows, Mac and Linux. The current Windows revision is 3.6. The Mac is at 2.6 and Linux is at a rev of 1.4. Using the Linux search, what do we see?

Yup - we see the Skype advert for a Windows version number. So if I’m a Linux user, and I click, I’ll be disappointed… because I’ll only be given version 1.4 for Linux, not the offered 3.6.

Conclusions for Linux

Well, if I’m a Linux user and interested in buying - the adverts are going to be mostly annoying. They can’t possibly be better… because there’s no specific clue from Google that the user is in the Linux segment of the audience. Potentially, this should be easy to solve. Implicitly, an additional keyword could be added for searches from the specialised indexes. This would allow advertisers to address specific adverts to those users. This would increase satisfaction amongst specialised markets with the advertising, increasing revenues for Google from those markets and justifying more specialised vertical markets.

Back to Mac

So, my problem is still unresolved. How do I use Google to find a Mac specific WiFi router? There’s no specialised search. I’ve already got a bunch of WiFi router/hubs that appear to have problems with Leopard and no updates - so I don’t trust any of these vendors to properly support a Mac. Time spent checking whether a new WiFi/ADSL router has problems is time wasted - unless I write an article about it, it has no beneficial effect on my business :)

Because vendors usually want the largest impression counts, the keywords in paid search will include “wifi” but not “-mac”, even when they don’t have a Mac solution - so my viewable adverts are stuffed with the vendors that I already don’t trust. And the same goes for the organic search results. Adding “mac” to the search query generates organic results that are little different from established Windows oriented vendors that I don’t trust to provide adequately tested Mac solutions - doubtless because the page rank weight they’ve built for “wifi router” overwhelms my interest in seeing “Mac”.

So what did I do? I went Apples all the way, or as far as I can. I relegated all my ADSL/WiFi hubs to being an ADSL router to ethernet, and added an Apple Airport Extreme (though if I’d waited a month, it’d have been an Apple Time Capsule). The Extreme works just fine. No WiFi drops. It did, however, have DHCP problems, over Ethernet, with the ADSL router. So the newest Airport Extremes also have a problem, not just over the air, but also over the wire. So I’ve now made the Extreme my DHCP server and the ADSL router is only that. This is a bit anoying, as I have to copy my ISP’s DNS servers info, rather than have that automatically updated - but they change their DNS servers IP addresses very slowly.

I didn’t bother with third party solutions from the major vendors, because their likely use of the Pareto principle means that I know I’m relegated to a second class citizen. And Google doesn’t help, because it doesn’t have a specialised search for Mac - and even if it did, adverts wouldn’t be properly addressed to a specialised audience.

Other Observations

This is fundamentally about Brand relationship. Although I now self identify as a Machead, I didn’t immediately reach for an Apple solution. Why not? Because search is my friend, my tool of choice. It was only when search failed to deliver an answer where I was confident that my Mac specific problem would be addressed, that I reverted to using Brand to provide a selection.

Related to this is what happens to users when they look for a brand. A substantial fraction of users appear unable or unwilling to differentiate between the URL bar at the top of the page, and any sufficiently wide input field near the top and middle of a page. Fully formed brand URLs are often typed into search fields, when they’d have worked perfectly in a URL bar. So brand owners should be closely considering what is shown on searches for their URL as well as the raw brand name.

This ties back to the observations in “Blown Apart” (Evans and Wurster, Harvard Business Press, 2000 - often referenced in my blog). As consumers are faced with increasing choice, the safest options are to use two indicators - perceived popularity and recognisable brand name. When you can’t face learning the differences between superficially similar items, picking the popular one is a fairly safe choice - your risk is reduced by making that decision. Without a clear indication of popularity, Brand becomes a powerful force. Indeed, if my relationship with Apple was perceived, by me, to be tighter, I might simply look for a solution straight on the Apple Store and bypass search - that is, after all, the holy grail of Apple marketeers for my affection.

Conclusions

The ability to segment is often crucial to advertisers, and can be only partially addressed by advertisers using AdWords.

Search and Brand are to a great degree complementary - when you can’t navigate with search, consumers will use Brand. But that doesn’t always apply the other way - if a Brand is known, and there’s a brand relationship, the power of search means that even looking up the web site of a well known brand may involve the use of search.

Google has assumed responsibility for segmentation - advertisers are not allowed to select “Linux users” or any other identifiable characteristic. Google has asserted that a history of search is a better indicator of intent than any other identifiable user characteristic. There’s a Google blog article about the targeting changes that Google was making in August 2007 - I can’t find the article again, now, or I’d link to it; sorry ’bout that. However, it looks to me that specialised markets could be identified by Google and made available to advertisers who specifically want that segment.

One solution might be that specialised results pages - vertical searches - potentially could be identified by special keywords, so that they could be used with standard AdWords campaigns.

Google may be optimising to the Pareto rule, but they are ignoring specialised markets that could yield higher revenues for specialist vendors and improve user satisfaction - if Google can deliver a generalised mechanism to address verticals. This may be complicated the effects of personal search - where results could be skewed to favour certain types of product or site that can’t be represented as search query contents.

But to solve my needs, so far, it’s Apples all the way down.