Tag Archives: Google

Retargeting: Hooray, it works. Let’s do so much of it we piss people off

There was a really interesting article on digiday this week  about retargeting and how the industry enthusiasm for it is risking the health of retargeting, and increasing resentment for online advertising as a whole.

To give some background, the display ad industry has suffered from a reduction in the perceived value of their ad inventory ever since the industry’s inception. Many cost models now exist (CPC – cost per click, CPA – cost per action) but the original pricing model for online inventory was the CPM (cost per mille, yes I know it’s french but it means cost per thousand to the ordinary person). By cost per thousand they mean cost per thousand impressions, and an impression is counted each time one ad is served (each ad may have multiple images within a loop, a drop down or other engagement and run multiple times, but until the ad itself is clicked, or another link on the page, that still counts as only one impression).

When  I first started in online at MSN in *cough* 1999, we asked for, and (sometimes) received, £26 CPMs for untargeted Hotmail inventory, and £29 for targeted inventory which then was only available based on content, not user demographics or any other criteria. Due to limited availability the MSN business channel was always sold out at a rate card of £29 CPM, while we’d do deals for, say £15 for Run of Hotmail, and run a LOT of house and charity ads.

That was during the dotcom boom, however, and things had to change, if only because all the 19 year olds with random website ideas (and funding) either became gazillionnaires or failed miserably. What remained after the chaos was real businesses doing real things and having to use a proper calculation of profitability to justify their share value. That coincided with added trackability of online, past the point of click and all the way to sales or leads that could be attributed back to a media campaign.

Trackability allowed a realistic value assessment (which became a stick with which to beat the display sector), and the huge growth of consumer usage of the internet meant that billions of available impressions were being added to the inventory of websites each month. Within a couple of years we also had networks brokering deals for massive volumes of blind remnant inventory that the publishers just wanted to make *something* from, and the CPMs kept tumbling.

From a few pounds in the mid 2000s, CPMs kept tumbling to sub £1 levels for large buys, meaning display publishers were constantly struggling to make money from their users, and the “pile it high, sell it cheap” approach was a self fulfilling disaster for advertisers as the vast majority of the inventory was wasted.

As users gradually became blind to the banner (the dominant 468×60 pixel image that held sway for a while), click through rates also fell from single figure %s to less than 0.25% meaning every advertiser and publisher was scrabbling to buy as cheap as possible as this was the only way to make the performance figures add up.

It didn’t help that all this was happening at the same time as the growth of search advertising, which with a CPC (cost per click) pricing model was a lot less risky than display, and also fits into the user journey much closer to the final sale, the death knell of display seemed almost palpable.

and then along came retargeting….

Retargeting hinges on cookies. Basically once you have interacted with an ad (clicked on it) or visited an advertisers’ website – anything that can result in a cookie on your PC, you are potentially identifiable (not personally but as a string of text/numbers) as someone who has done this action. This means that advertisers can either a) serve you a specific ad or b) in some instances decide not to advertise to you at all depending on the actions you’ve taken.

Here’s an example:

Yesterday I went on the BHS website to look at lighting in their sale.
Today when I visit the Huffington Post website, this is the ad I see – featuring the exact products I looked at yesterday:

Image

This is a retargeting ad, and is possible because the cookie on my PC captured data of what I was looking at, and BHS have bought a retargeting campaign, probably from a large provider like Criteo or Struq, stating that they want to target people who’ve been to their site and abandoned items in their shopping baskets, to persuade them to come back and complete the purchase.

Brilliant stuff. This works like a DREAM compared to standard display ads.

Target market – BOOM.

In purchase mode – BOOM.

Relevant products – BOOM.

Happy users – BOOM.

No ad wastage – BOOM.

Inventory gains value – BOOM.

Comparative click through and purchase rates went through the roof, ads are capped to only show 4 times per user to avoid annoying them, suddenly display gets back on the media plan, and everyone’s happy, right?

In theory yes, except like pop ups in the mid 2000s, the most successful ad type gets used to the point of saturation. Many advertisers will now only use display for retargeting, and in their enthusiasm for it, they use multiple providers to serve the same campaign. The result of this is that the frequency capping becomes 4 per network or exchange (and many advertisers use several) so the user can see the ad 20 times+. Plus the complexity of managing this many campaigns means that it’s also unlikely they’re  pulling client 1st party data through to show whether the sale was actually made or not – so, as in my case I’VE ALREADY BOUGHT THE DAMNED THING!

How likely is it that customers enjoy this experience and are encouraged to return another time?

Easy answer. It’s not.

<rant>

Given the amount of times that I see an ad after I’ve bought the product, or an insane number of times you’d think that it was difficult to prevent this happening, but it isn’t at all.

Any agency or advertiser using retargeting who doesn’t manage cross campaign frequency is risking not just their own customers, but increasing resentment for the whole online ad sector. And for crying out loud, drop a cookie when people buy so they don’t keep seeing it ad infinitum.

</rant>.

To brand search or not to brand search

Last week eBay threw its gauntlet down by publishing its internal study which found paid brand search advertising “ineffective” (to the point of negative ROI) and non brand search to have a very small impact, and even that only on the least valuable customers.

I wouldn’t be overly concerned for Google’s share price. This is just the latest sabre rattling in the SEM market between one of the world’s biggest search advertisers and the dominant supplier (Google) who given their 80%+ share in most markets, and their resulting ability to ‘make or break’ a business are seen as a “frenemy” by many clients and agencies.

Proving ROI for the wide area of non-brand search and the rest of digital is a complicated issue, which I discuss in more detail in my post on digital measurement and attribution, so I’ll save my comments today for the issue of “brand” search and how to tackle it.

The term “brand” search normally refers to the name of the site concerned – often its domain, derivatives of this plus misspellings. To take an example of a well known UK retailer, terms that would qualify as “brand” search for Marks and Spencer would be (I won’t go into match types here & now):

Once upon a time search terms like these were offered some protection by Google’s Trademark policy, meaning that trademark owners could register as such with Google and prevent anyone other than themselves (and their chosen list of resellers/affiliates) appearing in the search results.

Most weeks at agencies we’d spot a cheeky affiliate bidding outside office hours, trying not to be caught; or receive a furious call from a client who’d seen a competitor seemingly bidding on a brand term. We’d dutifully report the instance and send a screenshot to Google and then wait impatiently while they were hopefully removed. It was a constant round of chasing and frustration, and took up a lot of our time.

Google had an entire team of people dedicated to checking and implementing the restrictions, although the offences weren’t always as clear cut as they may have seemed. An advertiser whose brand name included a “generic” term could find that competitors ads would appear against their results through a process called “broad matching” – so “Toys R Us” may trademark the whole search string, but they couldn’t stop competitors appearing who had bid on the term “toys”. Lots of work for all involved, with the added issue for Google that if advertisers could protect their own trademarks, they wouldn’t have to bid on them, so Google made less money. So guess what Google did in 2007?

You’re right.

They removed the trademark protection policy for keywords (it still applies for ad text – so you can’t say ‘We’re better than competitor x’ or ‘We sell iPods’ if you don’t), which made everyone panic and assume there’d be a free-for-all as soon as all advertisers could bid on each others’ brand names.

What had been underestimated was whether it was going to be commercially viable to bid on competitor terms or not. Given that the more relevant a keyword is, the lower the cost (I’ll expand on Quality Score also at a later date) competitor terms tend to have a higher CPC than your own, so each advertiser has its own ROI metrics to decide if competitor bidding made sense for them or not. Not quite the panic that was predicted (Y2K anyone?) but even without competitors muddying the water, whether or not to bid on your own brand is just about the most frequent question that still occurs across paid search.

After much consternation at the time of the change, various factors led to the current status quo, which is that “brand” keywords such as those above make up a large proportion of many PPC advertisers’ spend, sometimes the vast majority. It’s not surprising then that marketers will often ask the question:

If they’re looking for me anyway will they come to my site even if I don’t appear in the paid search results?

and thus

Can I turn off my brand search terms and either save the money or invest it in other keywords or channels that will provide me with incremental traffic and sales?

The answer is …. yes occasionally. But this applies to only a very few advertisers …. and it depends on various things.

Natural search results:

The single most important factor in making this decision is where your listings appear in the natural (organic) results. 70% or more of all clicks on a results page are on the natural search results, so your potential traffic is limited if you’re not appearing.

Tips:

  • Page 1 in natural? Test brand search on/off across similar seasonality and track combined search plus direct-to-site traffic levels to see the impact. It can go either way, and it can differ depending on your offline marketing activity/seasonality.
  • Page 2 or later in natural? You may as well be invisible. Use PPC to ensure your potential customers find you.

To show an example for Marks and Spencer:

marks and spencer brand search screenshot

Marks and Spencer brand search screenshot

Their site takes up the vast majority of the results page because they’ve used all the methods available to dominate the space

  • they are the first result in natural search
  • they have natural site links (the sub links beneath the main result)
  • they have a Google+ page
  • they have their store locations registered with Google places so the nearest store shows up on the local results too

This dominance serves both a practical and a branding purpose. The searcher is offered very few other choices other than to click through to the M&S site; plus they are reassured that they have come to the right place with stimulating social content and useful location information that should help both their online conversion and possibly drive store visits too.

Competitor activity:

As we’ve seen above, there’s nothing but cost to stop your competitors bidding on your brand name, so the levels of aggressiveness in your marketplace will have a huge impact. Some clients will pursue a disruptive strategy to buy market share even if it isn’t profitable at first, in order to either gain future sales from those customers.

I had a large auto client once say to me;

“I’m pretty sure that if someone can afford to spend £50K on a car, they can find my website.”

My answer then is my answer now;

“Yes, I’m sure they can. And they can equally as easily find your direct competitor sites.”

The issue is that with online, everything is easy. It’s not like the offline world where a customer walks into a showroom and then has all the glossy cars and the smell of leather to seduce them, or the fact that it’s a faff to get the kids in the car and drive to the other luxury car showroom to make them stay. All they have to do is click a different link, and the competitor has just as high a chance of sending them a brochure/developing a relationship with them as you do.

Of course people develop emotional relationships with car brands, as they do other high involvement purchases, but there will be many who have a shortlist of more than one brand – and do you really want to take that chance?

An example for Go Compare, the price comparison site:

gocompare brand screenshot

Gocompare brand screenshot

The UK insurance market is cluttered and competitive, with many well known brand name insurers, plus the price comparison sites – many of which are now brands in themselves.

The search above is for the brand name derivative “gocompare” – and in this instance you can see that the competitors Confused.com and Compare The Market are bidding on the brand term, which dilutes the dominance that they would otherwise gain from their successful natural search results, and the Google+ page.

Confused and Compare The Market won’t do this for fun – they do it for hard nosed economic reasons, and it must be worth their while to pay the higher CPCs their competitor terms will demand. To me it shows that where the products are homogeneous (and/or price elastic) then searchers are highly likely to be tempted elsewhere even if they have expressed a preference in their initial search.

Tips:

  • Where your market is aggressively competitive, always bid on your brand term, and consider using ad text stating “official site” to give people reassurance and prevent last minute leakage. 
  • Consider a joint search strategy with your resellers and affiliates to help you to dominate the brand term space. You will lose some margin but prevent the click going to an outright competitor.

Branding, business as usual and campaigns:

One of the main differences between paid search and SEO activities is the speed with which you can see results, and for this reason you may consider continuing to use paid search for campaign driven activity, even if an “always on” approach doesn’t pay dividends.

Natural search results are something that take a long time to build. They are based on the authority, readability and quality of the content on your site and as such, shouldn’t be messed around with for a short term gain. The last thing you want is your natural search results saying “Sale On Now!” when it in fact ended weeks ago, as this is a bad user experience and will mean customers do not believe you the next time.

For sales, new product launches – any communication that differs from business as usual, then PPC is the ideal channel.

Example of Thomas Cook:

Thomas Cook screenshot

Thomas Cook screenshot

In this example, Thomas Cook have great natural search positions, location results, Google+ and comprehensive natural sitelinks, but still choose to run paid search ads to promote their  summer holiday deals, and to capture email addresses for future eCRM activity.

Tips:

  • If it’s worth running a campaign in offline/other media, then it’s worth repeating in PPC. 
  • Increasing dual and triple screen usage (TV, tablet and phone) means that TV spots times are immediately mirrored by brand search trends. Spending millions on TV only for people to get lost whilst looking for that promotion online is a leaky bucket directly to your competitors. Don’t stimulate demand and then fail to scoop it up.

Phone Book versus Yellow Pages.

The examples above have been for e-commerce sites and concentrate mostly on acquiring new customers, but one enormous factor that causes resentment amongst marketers is having existing customers click on the paid search link, when they are visiting purely to log into their account, interact with customer services or heaven forbid – make a complaint!

Most advertisers that have long term customers avoid using brand PPC for exactly this reason (have a look at British Gas, O2, Tesco, Barclays Bank for instance) and as long as their natural search is healthy, this isn’t the end of the world. It is a bit like having the free listing in the phone book or Yellow Pages, you’re there, you’re findable but aren’t going to get the customer particularly excited.

One thing to bear in mind for this approach is that your homepage will need to be easily editable with campaign information, to ensure that any campaigns that do get run elsewhere can easily be followed through rather than lost in the usual clutter. If you have 6 month code release windows, you may wish to run a PPC campaign rather than fight with IT.

A special mention should be made here for those who continue to advertise heavily in brand PPC with a standard message even with strong page dominance, such as Ladbrokes, Sky TV and Expedia.

You can look at their approach in two ways, the purist commercial or the customer centric way:

  • they are blending their (cheap) brand search results with their (more expensive) lower performing parts of their campaign, and this is artificially inflating the incremental value of their brand search;

or

  • they’re genuinely trying to just make it easier to find them, like having a freephone number

Either way, they have their reasons and as in most marketing decisions, it’ll be a blend of hard numbers, branding, a bit of finger in-the-air and “the boss likes it”.

Facebook and Atlas

The first thing I knew about Facebook’s purchase of the Atlas ad serving suite from Microsoft last week was a trail of disbelieving social media commentary by colleagues saying “What? Are they insane?”. By “they” here they mean Facebook, as the overwhelming view amongst past and current digital media operators is that Atlas is a “pile of s***” and that Microsoft must be glad to be shot of it, even at the rock bottom price of $100million.

Facebook has an enormous job to do to regain the trust and support of the online marketing community for Atlas or anything else that they build based on its skeleton. I have spent the best part of the last 4 years complaining about Atlas to the tech support teams, to the product leads and vainly shouting and throwing objects at my screen when another huge spreadsheet of thousands of bespoke tracking URLs is lost in the ether at midnight when the campaign needs to go live by 8am.

When all pleading made no difference I made a very vocal point of persuading clients to move their advertising to other solutions wherever possible, and I am not the only one.

It wasn’t always thus – 5 or 6 years ago Atlas was the defacto ad server for every digital agency I knew, so much so that Google seemed to have played a slightly inferior hand with its 2007 purchase of Doubleclick, and Microsoft’s decision to buy Atlas seemed to make perfect sense.  MS did make a classic 2nd mover mistake of desperately throwing money at the problem however, and the purchase raised gasps throughout the market with its $6.3billion cost (Yes, $6.3 billion) – more than double what Google had paid for its acquisition.

Granted that $6.3billion cost (Yes, $6.3 billion) was for the whole aQuantive group, which also included Razorfish (which they later sold for $530 million) and Drive PM, which was absorbed into the Microsoft Media Network (and arguably did add value to the offering). This didn’t hugely impact the end result however, which was still that Microsoft wrote down $6.2billion in 2012, mostly due to the aQuantive purchase, for which Atlas is the biggest culprit.

So how did Atlas go from the dominant ad server in a growth market, to losing the entire GDP of a small country (such as Liechstenstein) in 5 years?

Without seeing internal figures its impossible to say how much Microsoft has invested in Atlas technology since its purchase, but as a user it’s clear that the fundamentals haven’t changed since the mid 2000s. Every team I have worked on has been shouting in frustration at the Atlas team for years about usability and how they literally *hate* using it. The only tech in advertising that stimulates more frustration is DDS (cue bloodcurdling scream).

In the 5 years since Atlas was bought there have been multiple changes in the digital advertising space along with the rest of the technology world, most if not all of which expose Atlas as inferior to more nimble competitors:

The growth of paid search bid management software 

Atlas search offers a very basic click tracking function rather than any operational help to the search operator. This means that any serious campaign will need another software such as Marin on top to ease the thousands of optimisation tasks and data analysis, which just adds cost to the technology stack, and yet more cookies with more potential data discrepancies into the mix.

Cross digital measurement attribution

Atlas “Engagement mapping” garnered a bit of support for about a minute until we realised that many competitor tags couldn’t be placed in the UAT (Universal Action Tag, their container tag solution). What’s the point in running an advertising campaign where your media choices are dictated by your tag provider, not their performance? Any cross-media tracking that exists needs to at least include all paid media (and preferably direct & organic driven traffic too), so again Atlas offers only a partial solution at best

Facebook & other social network launches

Facebook has obviously been a market changer for the advertising world, and its growth has also created a market for Facebook campaign management software. Good Facebook campaigns need hundreds of targeting clusters created to maximise creative performance, and this scale of tracking and the speed with which it needs to be done is impossible with something like Atlas.

Across all digital media many widely used technologies were developed to manage one type of media, but the market direction is towards bundling search, social and display management capabilities into the same technology. If that technology is already being used, such as the ad server (Doubleclick) or search bid management software (Marin and Ignition One) then the consolidation of technology saves an enormous amount of time and complexity.

The consolidation daddies of digital marketing technology are now Google and Adobe.

Google has an ad network, an exchange, an affiliate network an ad server and a free analytics suite sitting on the same technology stack as AdWords, the worlds largest global advertising platform. In the time it’s taken Atlas to lose all its customers, Google has totally rebuilt Doubleclick, and Doubleclick search, is adding new functions almost weekly.

Meanwhile Adobe’s frenzy of acquisitions in the last 2 years have added a Data Management Platform (DMP) and search management to its “Marketing Cloud” which now offers end to end creation, tracking and optimisation for video, social, search, site analytics and landing page testing.

While I understand that Facebook want to have access to technology that helps them to prove that FB advertising works, they’d have been better buying a small, nimble FB specialist plus a small ad serving company and building their own solution. Either way without site side analytics and/or search management their offering will not cover the full picture that a digital marketeer now wants to see (preferably with one login).

Add to this that unpicking someone elses’ tech to rebuild it and re-gaining the support of a busy, cynical group of people (with lots of shiny other options) is not a small task. Facebook will do well to learn from Microsoft’s mistakes that being the dominant force is not an unassailable position to be in. Watch out Facebook –

Only the paranoid survive.

Andy Grove

Is there really a need for SEO?

SEO sort of snuck up on me. With a background in paid and performance media – first display, affiliate and then paid search; it took a while for me to feel comfortable with the often unpredictable effects of SEO activities.

With display or paid search, you have an obvious cause and effect. Pay for the media, it’s shown to targeted users (or self selecting in the case of PPC) and they respond, or not – based on the offer, the competitive market and how well your site and product mix satisfies their information or purchase needs. You can then judge the efficacy of each part, and choose to do more or less of certain elements of your campaign.

SEO seems to be the discipline most shrouded in secrecy – at least for those whose background didn’t start there. It reminds me of the accepted stereotype in 90s IT publishing years, where the clever, geeky (well before geeky became cool) slightly socially inept bloke from  ‘systems’ would come down the corridor bearing a stack of green & white striped listing paper with reams of data on it, to confuse the sales, marketing and financial management with acronyms and talk in seemingly an entirely different language.

Things have moved on, not just because of ITs ubiquity and accessibility on the desktop but I still suspect that many people, wittingly or not, over complicate technology in order to maintain an air of superiority (and possibly also their high consultancy fees). SEO seems a lot like this as words like “robots” “metadescriptions” and “algorithm” get bandied around and confuse the marketing team into paying for some magic fairy dust that will convince the mighty Google to bend over and put your site where you think it ought to be.

The reality is that many marketers pay large amounts of their budget on SEO but the results are not immediately obvious. Sometimes work can take weeks or months to make an impact, and sometimes is justified in terms of the opportunity cost of not doing it “your rankings could have slipped to here if we hadn’t been doing x or y”.

In an increasingly KPI driven world, it’s hard to compare the cost of driving/maintaining traffic from a fee-based SEO service with the cost of incremental campaign traffic from your performance campaigns. Often it’s run by different teams in a client company, and definitely by different operational teams in an agency, which means that too often the integration benefits are overlooked in favour of a competitive us-and-them attitude.

There are of course core fundamentals to good SEO, at which point I humbly refer to the bible of search, Search Engine Land, who have a fantastic infographic showing the factors that influence a site’s ranking, that any SEO worth their salt will live and breathe.

The Periodic Table Of SEO Ranking Factors

The Periodic Table Of SEO Ranking Factors

TL;DR:

  • Say honestly what the site is about, don’t pretend to be something you’re not (even if that subject is more popular than you)
  • Label titles, images and links well. It helps search engine spiders find you and distinguish how each page is different
  • Have plenty of relevant, in depth, unique and fresh content
  • Try to link to and from other good relevant sites that will help people to find more useful information.

Clue: if a user would like it, Google probably will too.

This will also mean that the user will share it which will give Google the social signals that are increasingly impacting how pages are ranked (in both natural and paid search).

As all of the above things are what your website team should be doing anyway, then doesn’t that mean that there’s no need for SEO at all?

The answer sadly, is no.

Website design and development is a long, iterative process, and it’s rarely possible or even advisable to throw an entire site away and start from scratch in an SEO friendly way. Apart from anything else, even if your site isn’t the best one, the length of time it’s been there will give it some authority so there is always a risk with changes even on a long standing domain.

The way that the Google algorithm ranks pages is also constantly updated, so what works well at one time may be less relevant in a few months time, or you may find there’s something new that is being taken into account. For years using a flash site was seen as SEO hell, and in some cases it’s still not a good idea, but there are now ways to ensure that the text content can be read and Hey Presto, users like it, and so does Google.

Some months ago I was lucky enough to attend a presentation at the shiny Google London office from Amit Singhal (@theamitsinghal), one of the original architects of the Google algorithm. I was impressed by his passion that users should be able to find the most relevant content, quickly and efficiently (with seemingly no care for whether Google could monetise it). True to this belief, the updates that happen to the Google algorithm are always meant to improve the quality of the overall search results. You may for instance have heard of Panda, which started in 2011 and is a rolling update meant to remove “thin” content from the results pages (hence preferring sites with substantial deep content).

Quick wins in SEO

The best way to maximise your chances of appearing in the search results page is to apply best practise SEO principles for the site structure, linking architecture and labelling conventions in the first place. If you haven’t done this already then any good SEO can tell you which are the key things to edit that will make a difference.

One the structural issues are dealt with, the best way to build and maintain rankings is to continually improve your site, and make the content deeper and richer – thereby more valuable to users who will visit it, share it and link to it more, which will all help you too.

And a final reminder to avoid any of the “get high quick” schemes. Any SEO that guarantees rankings or makes an unexpected big jump in a short space of time is probably doing something that Google will find out about and penalise. Global brands like BMW and Interflora have been de-listed entirely from Google before for violating their policies, so it is just NOT worth thinking you can swerve the big G.