Sometimes being able to see things in fine detail is unpleasant- particularly when it comes in the form of a sudden revelation.

Every now and then you get that kind of shock in PPC data, when the serene progress of an account has to be re-evaluated in an instant because of a new discovery. This is exactly what happened to me when I took a look at the affect of Google’s new Adwords transparency on some of our PPC campaigns.

As you’re probably aware, Google places ads in three places- its own results page, the results page of its search partners (Google Product Search, Google Groups, Earthlink, compuserve, shopping.com, AT&T Worldnet, and search sites such as AOL and Ask.com) and on its content network. Previously, Google bundled together its own results page with the other search sites like Ask.com, allowing you to separate out the data from the content network if you wanted to. The new tool allows you to fully break down the data from the three places Google places your ads.

I don’t often run content network campaigns so I hadn’t felt the need to delve too much into this new breakdown. I assumed that the results from Google and its search partners would be very similar– same adverts, same keywords, same delivery system and surely then similar CTR and conversion rate. This would mean that the best approach for me would be to use the summary view of my accounts.

However, when I did try separating them out the result was very surprising. The figures were astonishingly different; clickthrough was often a fraction of Google’s, conversion rates were much lower and although cost per click was less, cost per conversion was way up. Here’s some data that I think highlights the problem pretty well:

This campaign has been running on Google’s conversion optimiser with CPA set at £12 and yet it has chosen to spend a rather large amount on the Search partners’ share of the clicks when the CPA is way above the required level. At this point I am still unsure as to why there should be such a big difference. Do the lower quality search engines fail to distribute correctly by location? This could explain some high traffic volumes but low rates. Whatever the reason, judging by the clickthrough and conversion rates, the search partners produce traffic that is poorly targeted.

We can’t choose to target certain parts of the partners network, nor can we create ‘partners only’ campaigns which we could then refine the traffic from, bid in accordance with the cost per conversion and maybe use to our advantage. Even if we could separate out ‘partner’ campaigns we can’t drill down to see which of the search partners is more effective. If Google gave these options we could quickly duplicate our campaigns, distribute them to each area of the ‘partners network’ and find out if they could be improved to come into line with the other campaigns.

As it is, I am switching many of my campaigns to Google only and, unless greater data and control becomes available for the search partners, I can’t imagine returning to this area anytime soon.

Hopefully, this post will prompt someone else to take a look at their figures and maybe, with this one change in Adwords, improve their CTR and conversion rate in a stroke. To take a look at your own figures in Adwords, go to your Campaign Summary page and choose ‘Split: Google Search/search partners/content network’. I can assure you that it’s well worth your time and may make a significant difference to your campaign.

Tweet This Post

So, as promised, over the last couple of weeks we have been running a Crazy Egg test on the Distilled site. You may have noticed quite a dramatic change to the home page: we have brought our four main focuses of work out of the main menu into bold call-to-action buttons. This helps people to see more clearly our main areas of expertise without having to read the much smaller text in the menu. We used the Crazy Egg test to monitor 1000 visitors on the Distilled home page, for two separate designs.

The first design (our old home page). The majority of clicks were on the ‘Contact Us’ and ‘Company’ links. A spattering of clicks were evenly distributed between our four main areas of expertise, both within the menu bar and the reputation management feature on the right hand side. Hardly any one clicked on the Distilled logo.

The new design with bold call-to-actions was launched on 20/05/08. It is clear that the call-to-action buttons have been a success. The main body of clicks now seems to be divided between the call-to-actions and the top navigational menu. However, as more users get used to the new design I wonder if the ratio of clicks between the main menu and the call-to-actions will verge more towards the call-to-actions.

It would be useful to monitor whether the number of people who clicked on any of the links increased. This would indicate that a higher percentage of people understood the options available to them. People tend not to spend long on a page so it is imperative that at a glance your users know what their options are. Hopefully the call-to-action buttons will help that. If this is true, we should see a decrease in the bounce rate of people landing on the homepage and immediately leaving it. This will be difficult to gauge in this instance as, during our second test, we had our team building day. The buzz surrounding this caused our click through rate for that day (and a few of the following) to greatly increase above normal figures. The team building day may also mean that people tended to click links that they may not normally have done.

Only time will tell as to whether this new feature will permanently increase our click through rate and indeed consequently boost new custom.

Tweet This Post

The gender balance has got a little testosterone heavy in the office with our new recruits taking the head count to 6 guys and 3 gals. So I thought a thoroughly gal-orientated blog post might help level the hormones.

With the sun’s summer debut last weekend it was time for my ’spring clean’ (as it were) and, what with one thing and another, I realised Reputation Management is much like a beauty regime…

Preventative
Cleanse, tone and moisturise. Rinse and repeat as necessary. Much like those crucial preventative or preparatory steps, reputation management is best achieved when tackled in advance. Building up your positive content can act as a buffer if anything negative happens, just like that skin care mantra can keep you looking resplendent even if you indulge in one too many Cornetto ice creams.

Break-out
Yikes! Break-out! You wake up, look in the mirror and realise something has gone very very wrong. Maybe you’ve actively done something bad (left the hair dye on far too long; distributed exploding laptops), maybe you’ve missed a crucial step of the preventative stage (passed out with your make-up on again; failed to create any positive content), maybe you’ve done everything right and it’s still gone horribly, horribly wrong (hormones; pissed off ex-employee).

What to do? The knee-jerk reaction is to cover-up, but this is often the worst choice of solution. So you’ve dyed your hair pink by mistake? Style it out: re-dying or a hat that covers all may seem appealing but you’ll fool no-one. The same rule applies for reputation management. If you face the problem head on and don’t try to fool anyone, you’re much more likely to come out with at least some of your reputation intact.

Recovery
Ok, the worst of it is under control. The recovery stages are simple: deal with the problem and fix it if you can. Then return to those preventative steps and work at them until all the bad stuff either gets better or disappears.

Now, what have we learnt? No prep + negative stuff = trouble. And, because a flare-up can happen anytime, anywhere, the only constant is the preparation.

Also, if you get yourself sorted, you can afford to take risks- how would you know pink wasn’t your colour if you didn’t try it? But these actions have to be supported with something to fall back on. Great eyebrows and well-applied lipstick could take a lot of the attention away from your hair, just as a kick-ass set of online profiles and some independent positive content can keep your rep lookin’ rosy!

Tweet This Post

Following on from our recent post where we asked various SEOs from around the world questions on geo-location one of our readers Sean Carlos emailed me asking if it was alright to leave quite a lengthy comment which included a few links to some articles he’s written. I thought the content was good enough to be worthy of it’s own blog post so here you go!

Sean Carlos:

Very nice post – oh so many complicated issues when considering locali[s|z]ation issues!

On point 1, when considering accented characters, I try to use numeric html entities to ensure my text is compatible with xml feeds used by blogging and other CMS software. The XML standard only recognizes 5 character entities (", &, ', <, >), one of which, ', is not even part of the HTML standard. More detail can be found in my article Accented Characters in HTML Documents: Considerations for Search Engine Optimization.

Point 7 is rather complicated. There are three main ways to distribute language variants of a site:

1. folder, i.e. apple.com/uk , apple.com/it
2. top level domain, i.e. apple.com, apple.co.uk, apple.it
3. sub domains, i.e. uk.apple.com, fr.apple.com, it.apple.it

Google is very good at dealing with any of these approaches – especially if it is easy to recognize content language. As mentioned in the post, server geographic location and incoming links are additional clues search engines can use.

The “best” solution really depends on the client’s market and future plans. For a site in Italian, the language market is limited to Italy with some “overflow” in neighboring countries. So a .it domain is a no brainer. Yet what about Switzerland? There are four legal languages. Add English to the mix in the case of many multinationals. Three of the legal languages are used extensively in neighboring countries (French, German and Italian).

Once consideration is the need to avoid shooting myself in the foot with duplicate content issues and dispersion of incoming links across multiple domains, i.e. .ch, .de and .at or .ch, .fr, .be and .ca. Yet some solutions may get me multiple listings (= more real estate, good!) in search results. Some top level domains are difficult to register if you don’t have a physical presence in the country (i.e. .fr). So, unfortunately, the answer is “it depends”.

In addition to the points mentioned in the post, I would insure language clues are inserted where appropriate in the html. Search engines are very good at automatically recognizing languages, using clues such as the domain, hosting location and incoming links. Text pattern analysis is probably decisive. Using the “lang” attribute on html tags and specifying a content language http header or meta equivalent can assist this process. More details can be found in my How Search Engines Detect HTML Document Human Language.

I met Sean at the Manchester SEO meetup last year and he was a pleasure to talk to. He works for Antezeta SEO Company in Italy. Thanks for stopping by Sean

Tweet This Post

At the end of last week, we heard the news that the initial bids in the FCC’s auction of 700MHz spectrum reached $2.4 billion. This auction is being run to reallocate spectrum previously used for analogue TV that has been freed up by the move to digital.

I don’t know very much about the technical details of spectrum, but I find it a fascinating commodity and allocation process (OK, I’m a geek). It’s a classic public good – it isn’t at all clear who should own it in the first place in order to sell it. There is a tragedy of the commons problem though if it isn’t allocated somehow – if everyone tries to use it, then no-one gets to (because of interference problems). This isn’t a problem for some low-power, short-range technologies (like bluetooth or ultra wideband (UWB) (not that UWB seems to be a problem anyway, given no-one is using it)), but you need allocation before you get technologies like mobile phones, TV, radio etc.

Historically, governments tended to allocate spectrum to public bodies (e.g. the BBC) or companies by beauty parade (read: cronyism). Back at the end of the 1900s, the economists, mathematicians and game theorists crashed the party and argued that allocation by auction would bring a number of nice benefits:

• Revenue for the Government / taxpayer
• Efficient allocation (i.e. the party prepared to pay the most should be the party who has the ‘best’ use of the spectrum in the sense of the one that people are most prepared to pay for)
• Fairness of allocation – no reliance on ‘contacts’ or cronyism – it becomes all about the business model. The markets sort it all out.

I am actually more qualified to write about this stuff than pretty much anything else you’ll see us blathering on about here on the Distilled blog since auction theory research formed part of my graduate studies. If you care about the background, I’ve included a bit at the bottom of this post.

As well as a bit of academic background, I also used to work for a company called Analysys consulting for industry, financiers and the Government on telecoms and Internet issues. One of the things we studied (which I still can’t really talk about) had to do with re-allocation of spectrum used by the mobile operators for 2G services (which is theoretically being made redundant by the introduction of 3G). Another study that is public knowledge that I was involved with was reallocation options for a range of selected spectrum bands that had become free after use in various public safety, military and emergency services roles (among other uses).

US 700MHz spectrum

The 700MHz band is a particularly attractive band for many commercial users. It is a long enough wavelength that a single ‘cell’ can cover a significant distance and also has good penetration of barriers such as walls and buildings. As wired says:

Generally, the lower a radio signal’s frequency, the farther it can propagate and the more easily it can penetrate obstacles like walls and buildings. Lower frequencies also tend to be more efficient, enabling radios to transmit more bits for each hertz of frequency band. As a result, the 700-MHz band should provide better coverage than current cellular bands, which are between 800 MHz and 1900 MHz. So if you’re frustrated by the lack of reception you get on your mobile phone while in the office, a cell service that uses 700 MHz spectrum could offer some relief.

GigaOM has more on the technical details.

Why SEOs should care

The flippant answer for why SEOs should care about the results of the 700MHz auction is that Google is involved (and we all care about everything Google does, right?).

In fact, I think there is another reason. There is a pretty good chance that the eventual winner of the so-called “C-block” of coveted spectrum is going to roll out a new (US) nation-wide wireless data transmission network. People who know about these kind of things (see GigaOM link above) estimate that using this spectrum in place of the most attractive currently available alternative will half the cost of building the infrastructure for such a network from $4 billion to more like $2 billion. That makes it pretty likely that someone planning to do this kind of thing will be the efficient user of this spectrum.

Especially in rural areas, the properties of the 700MHz spectrum mean a far greater chance of ubiquitous wireless connectivity. Finally, mobile internet comes of age.

More than anything we are currently seeing, this will result in game-changing shifts in power for online marketers. We are going to be writing more over the coming weeks about the differences between ‘regular’ online marketing and online marketing targeting mobile users. But for now, suffice it to say that much of what you currently do is going to need to be re-evaluated as more and more users go mobile.

Going back to the flippant reason: the ‘Google factor’, there has been some speculation that Google would like to win the 700MHz spectrum not only to create ubiquitous wireless data connectivity, but also that they might want to launch free services supported by usage-contextual advertising (link). Scary huh? Opportunities, though…!

Some light relief – past auction comedy

While my overview of auctions above lists loads of good reasons why they are an attractive allocation mechanism, they can also backfire badly (as can anything when governments are involved). From my paper on combinatorial auctions (compiled from J. McMillan, “Selling Spectrum Rights”, Journal of Economic Perspectives, Vol. 8, 145–162, 1994.):

Having been advised by the US/UK consulting firm National Economic Research Associates (NERA), the New Zealand government decided to adopt a second-price sealed-bid auction for their radio, television and cellular telephone spectrum. The scarcity of competition in the small New Zealand markets caused a politically embarrassing situation when winners paid prices far below their bids. Some extreme cases included a firm paying the second highest bid of NZ 6 following a bid of NZ 100,000 and another paying NZ 5,000 after bidding NZ 7m (at the time NZ 1 equaled 55 (US)). Whilst these can perhaps be justified in terms of economic efficiency (if not revenue maximization) another case of a student from Otago University bidding NZ 1 for a television licence for a small city and being awarded it for free after no-one else bid simply makes a mockery of the process.

If the New Zealand government wanted to take advantage of the nice theoretical properties of the second-price auction, they should have anticipated the relatively low competition and imposed reserve prices. Due, though, to the media furore surrounding the fiasco, this route was discarded in favour of first-price sealed bid auctions for future allocations.

Similar difficulties were had in Australian auctions for satellite television services, described by an opposition politician as “one of the world’s great media licence fiascoes”. The licences were won unexpectedly by Hi Vision Ltd. and Ucom Pty. Ltd. (beating favourites including a consortium of Rupert Murdoch, Kerry Packer and Telecom Australia). They did this by exploiting a flaw in the auction design that allowed them to default on bids without paying a penalty. They bid very high: A 212m and A 177m respectively (at a time when A 1 was worth 68 (US)), which was widely hailed as demonstrating that the Australian television industry had come of age.

In fact, they never had any intention of paying these high prices. As they defaulted on their bids, the licences were awarded to the next highest bidders: the same companies. They had placed bids at A 5m intervals right down to A 117m and A 77m respectively (both in fact going to Ucom after Hi Vision defaulted on all their bids). Ucom proceeded to sell both licences at a profit. This auction had both failed to award the licences to their efficient owners, and to generate decent revenue. In addition, the repeated default process had delayed by almost a year the introduction of pay television to a country already behind much of the world. If the cost of defaulting is low, the bidders are effectively bidding on options on the items for sale rather than the items themselves (P. Klemperer, “What Really Matters in Auction Design”, Economics Working Paper, WUSTL, 2000.).

It has been suggested that even the deposits of hundreds of millions of pounds may not have been enough to prevent default in the UK 3G telecom auctions had the desire arisen among the winning bidders – who were paying billions of pounds (P. Klemperer, “The Biggest Auction Ever: the Sale of the British 3G Telecom Licences”, Economic Journal, 2002.).

My background in auction theory

At university, I studied combinatorial auctions (which are often used in spectrum allocation) during my ‘part III’ 4th year. Combinatorial auctions are those where bidders are bidding on multiple lots at once and may have preferences for particular groups (making it inefficient to auction the lots one after another). The difficulty from a mathematical perspective is that depending on how you structure the auction, either the bidders or the auctioneer has a very hard (NP-complete) problem to solve.

I even managed to introduce a small bit of original thinking (I believe) around a particular class of auction whereby the auctioneer distributes the hard calculations across all the bidders and asks them to search for the winner. The problem with this, of course, is that bidders have an incentive to only report improved solutions that benefit them. I wrote about game-theoretic approaches to stopping bidders gaming this stage of the process.

Tweet This Post

I know from our recent PR seminar feedback (by the way, you can now buy the ebook for only £25: how to make PR work online) that many people in the PR industry and especially those working in-house are interested in the best ways of distributing press releases online. Obviously you will want to engage directly with your journalist contacts and send them press releases directly when they are interested, but there is also a need to syndicate most releases and send them out on the newswires. What’s the best way of doing this?

First, the benefits of syndicating your release:

1. it might actually get picked up by journalists directly from the newswires (try to make your headline good to increase the chances)
2. dedicated bloggers often use online news sources to find inspiration for new stories – it can be very hard work consistently coming up with new content when you don’t have the resources of a large publication
3. you can gain exposure through news channels that perform automated scans of the newswires and syndicate relevant content (and this gives you another shot at getting in front of journalists)
4. many of the newswires let you control links and anchor text, giving you some immediate links back to your (client’s) website and increasing the chance of you getting the links you want in any editorial content generated

So how do you syndicate your press release?

I would actually like to get some input here about other services that people have found useful. I believe, like almost everything online, there will be large differences in effectiveness for different sources in different industries and kinds of story and that, like all of these things, the importance of trial and measurement can’t be overstated. First, our experiences:

• We have had the most success with PRWeb who offer a number of pricing models, from the basic $80 option (which allows links, but no control of anchor text), through the $200 option (with control of anchor text) up the all-singing, all-dancing $360 option (with ‘additional social media features’ – essentially social bookmarking links, ‘digg this’ buttons and tagging). We have tried all three of these options and got the best results for the money with the $200 option – controlling the anchor text is very useful, but the additional features of the most expensive option weren’t that useful for us as we did our own social media marketing. If your story is very strong, with a compelling enough headline to help you get it picked up by journalists who contact you for more information, the basic option might well be enough, as any hyperlinks would be included from fresh anyway.
• I have been impressed by the features offered in webitpr’s social media release – from PR blogger’s new company but haven’t had a project opportunity to try it out yet
• Other paid newswire services include PR newswire (which I have heard some good things about but don’t have first-hand experience of) and Business newswire
• There are also free services, such as i-newswire – as with anything though, I think to some degree you get what you pay for

If you have used any of the services above, or indeed any others, and have a case study you can share, please leave us a comment and I’ll include any good ones here in the main article (and if you let us know your website address, you’ll even get a link).

Tweet This Post