Yesterday at SMX London was full of information and handy tips on Paid Search and Analytics. There were four sessions:
Here’s our write-up of them.
Get the basics right.
Leverage new ad formats
Fine tune once you’ve experimented. There’s always room to test and fine tune. Use a continuous testing methodology.
We have 70 characters in which to grab attention of our prospects.
Identify and target your prospects, then write ads that target prospects type.
Types of buyer:
We choose based on emotion and justify with logic. Emotional part of brain sends 10 times more info than rational side. Evoke emotional response and give facts to justify it with. We seek
Factors that influence our decisions. No one wants responsibility of making a bad decision. Avoid ‘DRED':
There are four key emotional drivers, the four Ps:
Target prospects based on their point in purchase cycle
Expedia is Google’s 3rd largest client, after Amazon and eBay
Travel is a high engagement vertical. Shoppers spend around 2 hours researching, viewing up to 20 websites across 6 sessions. Transaction latency can be anything up to a month.
Travel grows by 25% a year. 85% of leisure bookers use internet as main research aid
Users are device agnostic, using whatever’s available to them. Customers have never had so much choice about when, where, how to shop – so brands have to be everywhere.
Research and understand your customers’ needs, not what you think they want. Use search query reports – they should form the backbone of everything you do. See what searches trigger your ads and identify negatives.
Think carefully about match type strategies – Expedia run on broad match for a week then add negatives and start using more precise match types
Make ad copy as relevant as possible, and make landing page as relevant as possible to the ads
Leverage opportunities to take up more space on the page – eg sitelinks
25% of consumers scan URL for relevance, so include terms as subdomain in the display URL.
Align SEM and SEO strategies
Winning the click is just one part of it: you have to demonstrate how one click affects another. 60-70% of conversions are on brand terms, but people hear about that brand earlier in buying cycle. So consider how you spread attribution.
Most businesses use last-click, so they switch off head terms, which kills the traffic that goes into funnel. If you’re knocked out at start you won’t be there at end. Try different attribution models, eg ‘prefer first click’.
Remember search marketing small part of very big mix. Consider other channels like TV, radio, print. Try using dynamic phone numbers to track where phone calls come from.
How do you imply prestige? Pamela uses words like exotic, luxurious, rich, high quality, 5 star. Also, the brand may connotate luxury already.
Guy Levine suggested looking at magazines within your niche for inspiration, as they have similarly little space to grab your attention. Look outside the ‘bun fight’ that is the search results.
How do you check competitor ads and keep your ads different? Pamela makes seasonal ads: this lets the audience know she’s paying attention to the times, and keeps ads fresh.
Ben suggests having a good spread of ad copy for biggest spending ad groups
How do you measure attribution and use that to allocate budgets? Ed said that attribution is largely academic, and that whatever model you use isn’t going to be right for everybody. Foster a testing culture.
Ben says there are technologies out there to track what assists each keyword gives: eg Marin, Kenshoo, Acquisio, Google Analytics Premium and Adobe.
Does the increase in CTR from sitelinks lead to an increase in conversions? Ben said you’ll only feel the benefit when your ads are already in the three top positions, so it tends to be the keywords with the highest conversion rates that get sitelinks and the increase in traffic.
Pages that look like blatant advertisements don’t work. Don’t say “Buy my stuff! Now!!!”
Don’t violate design conventions.
Answer user questions: normally when people are searching they’re looking for an answer to a question and they’ll leave if they don’t find the answer.
Simple and low cost products can have simple and easy designs with simple messages, eg Mail Chimp.
Complex or more expensive products (eg financial or property) need more information – people won’t convert from one sentence.
Graphically rich pages work well for branding, eg LimeTree.
Biggest challenge for CRO is the cookie law.
The home page is not a great place to send PPC traffic.
Every landing page should have a purpose and a defined ‘most required response’. The most required response should match up the stage in the buying cycle – eg someone with searching on a generic is too early in cycle to buy, so the required response shouldn’t be buying.
Use ecommerce tracking on form fills, so you can get ‘days until purchase’ info.
Think above the fold – that’s the first step you’ve got to make someone decide to stay. Repeat your messages – like a small child, you have to say the same thing over and over, but in different ways.
Restrict navigation. Don’t give people too many options of where to go if there’s a specific thing you want them to do.
Build trust. Video is good for this. Use convincers like trade membership logos and partnerships.
Not everyone is in buy mode – use a 2 step sell. Get info like email address, then follow up to get a purchase. Some people abort if pushed to buy too hard too soon.
Use forms scientifically. If you want to qualify use longer forms, as those who fill them in will be more important. eg They had a training company with a form to fill in to book a free session; as the form is long, people who are bothered to fill in the whole thing are more likely to turn up.
Don’t just think about keywords and landing pages – think about people and their needs
PPC testing comes on the back of a good campaign set up. Otherwise traffic is out of context. Use ‘peel and stick’ – peel out good keywords and stick them into their own ad groups.
Have a tight correlation between keyword and ad copy.
Track conversions (including call tracking)
A/B testing versus multivariate testing – A/B works well if you’re starting out or have less traffic to work with. Multivariate is good if you’re advanced.
Testing is about overcoming people’s objections – people have reasons they won’t buy what you want to sell, so you have to overcome them.
Choose an experiment, don’t just guess – have a hypothesis, eg ‘people don’t buy because they can’t find the Buy button’
Do big tests first. Then refine.
Site needs to say 3 things:
Brian went through some top landing page mistakes. It takes two tenths of a second to form a first impression of a website, which will bias them for the rest of the visit – so your page has to be good.
Mistake 1 – Visual Bullying.
Trying to use overly aggressive visual elements to make people buy, eg really large ‘Buy Now’ button. No one buys just because the button is big. The ‘Buy’ button should be prominent but shouldn’t be aggressive.
Too many font treatments – serif v sans serif, drop shadow, bold, italic, coloured… creates clutter and makes the page hard to read. Use fonts to organise copy not to call attention to too many different things.
Distractions, eg rotating banners –distracts attention and slows loading of page. Can’t digest and understand when something is changing every few seconds. You can’t just slow the banner down to the right speed, as then it will be too slow for some people. Can put little boxes to flick between banners but people won’t click them. Brian was less opposed when you’re using pictures to create emotional appeal.
“You’re gonna have to read it all” – too much text on one page. Use tabs. Make pages scanable.
Overuse of colours and contrast. Eyes go from dark to light, and notice high contrast first. Too much colour and contrast clutters the page.
Mistake 2 – Ignoring Context
Why are visitors on the site and where are they?
We like to think that everybody visiting our site is ready to buy: but they aren’t.
Get inside peoples’ minds with use cases. Define the most typical roles and tasks of people who come to your site. Who are they, what’s important to them, what level of knowledge, location in buying cycle, where they’ve come from, motivations, beliefs, desires.
Remember the ‘pre-research’ phase – “is your product even right for me? I don’t even know what questions to ask yet!”
Mistake 3 – Not Establishing Trust
Types of trust:
Say if you have thousands of customers or have been in business for 35 years.
Relevance – What a user has seen just before they come to landing page. Landing page has to look like and be what they’re looking for.
Try to work out what people have seen – snippet in organic search, ad for PPC, etc.
Attention – Grab attention. Use the headline and image. Link back to the brand and values you want to represent. The appeal should reflect your value proposition.
Motivation – Need to motivate people to keep using the page.
Show, don’t tell: don’t say “this is the best product”, give proof. eg LifeLock – CEO gives social security number on website to demonstrate his trust in his identify theft protection product.
Make your appeal sticky – read Made to Stick by Chip Heath and Dan Heath.
Don‘t need all of these factors, but you should try to get as many as possible:
Orientation – don’t lose the call to action. Guide peoples’ navigation so they look at things in the right order.
Action – people need to know what to do next and you need to push them to do it.
You can’t skip the other steps and just have a CTA. Can’t just say “15% OFF!” without saying why they should trust you and what your product appeal is.
Bonus tip – steal! Do research and find out customer concerns and language, and steal ideas from that. Do consumer research with surveygizmo, surveymonkey, crazyegg, kissinsights, silverback, usertesting, going into Starbucks and interviewing people, etc.
Use Evernote religiously – it lets you organise screenshots of websites, emails, etc. Tag things according to the principles they use so you have examples. If you have an account you have an Evernote email address, so you can sign up for all competitor newsletters and forward them automatically into Evernote to build up a portfolio of email marketing.
Stephen – Any consultant should go after biggest returns first, so there are diminishing returns on doing lots of testing. He’s seen a lot of Website Optimiser accounts with around 20 tests in, none of which have statistical significance – make sure you have a strategy.
Malcolm – User testing can make more drastic improvements than CRO.
Guy uses Visual Website Optimiser, Google Website Optimiser, Optimisely. Biggest lift comes from knowing what tests to do rather than which platform you choose.
Malcolm – ClickTale gets good data (records every mouse click) but can slow page loading.
Stephen – New ClickTale users can get overloading with the info – need purpose for it
Brian uses roles instead of personas – personas need more details (what type of car they drive, favourite sport etc), while you can manageably have more roles.
Stephen – In an ideal world, you would divide up website between user groups and place in buying cycle, but takes that significant development time.
Stephen – The process for different languages or cultures are the same, although results may be different. For example people in Germany took at terms and conditions for poker sites more than in UK or USA.
Malcolm – Some countries have less credit card penetration, so ecommerce sites can’t be set up the same way.
Different countries are implementing the European cookie directive differently. Germany hasn’t actually implemented anything yet. Italy hasn’t really required the opt in. France has a strict opt-in law.
The UK’s implementation has been “rather messy”. They said they’ll be tough, but then one of the ICO websites said that it’s “highly unlikely” they’ll act on first party analytics cookies that have a “low level of intrusiveness”. But they will look more harshly on people who’ve done nothing than people who’ve done something even if it’s a bit wrong.
The ICO website use a banner at the top of the page for you to opt in to cookies. 90% of visitors chose not to.
What should you do? Guidance says:
Demonstrate you have taken action on cookie law, then you’re much safer than you would be otherwise.
Don’t put your head in the sand.
Do a site cookie audit – should do this about once a year
“we know there are too many tags when we have tags to manage the number of tags”
Call the other departments in your company – SEM/SEO/IT/Compliance/Legal/Everyone – everyone likely to be affected
Watch what other people are doing. Watch the giants – Google, Microsoft, Omniture, etc. Watch your direct competitors – the more prominent you are the more you should worry.
Don’t be ‘excessively’ open about your modifications – users won’t respond well to splash pages and pop ups. You can still browse the ICO site while ignoring banner at top.
Allaboutcookies.org – give pop up with choice of permanence. If you say ‘no’ the page says ‘we need ads to support us, please change your cookie settings’
The ICO are not saying exactly how they’ll police laws.
Look for new tools to emerge eg Wolf software http://cookies.dev.wolf-software.com
Intent Marketing – intention tracking has been going on in direct marketing for decades. You can get unbelievable amount of info from credit card use, and it’s acceptable to sell this info. You can also get data from store cards – are you diabetic, alcoholic, a parent? This is mined and used for circulars and promotions.
These same techniques are used for digital, to create relatively precise advertising.
Why is the standard different in digital? Why do people expect privacy that isn’t there in a physical shop?
An Econsultancy study says that 82% of digital marketers think cookie law is bad for consumers, but 80% of consumers think it’s a good idea. But most consumers hadn’t heard of law before, and associated cookies with malware.
Keywords are good indicators of intent – but they stay out of the privacy debate as consumer is giving them away. But keywords are not the end state.
Google think their main competitor is Facebook rather than Microsoft – because Facebook has more info on the users, so they could answer users’ questions more precisely.
What does this mean for SEO? Personalisation and social are becoming more important for SEO. There’s going to be a huge amount of flux.
What does this mean for PPC? Better info on users’ location, social networks, device, demographics. There will be additional signals to help target intent.
But the expectation of anonymity online is a key consumer issue.
Andy – There is now a call in Spain for the right to be forgotten; people are now realising that digital data are permanent. There isn’t the same direct link between credit card data and the direct mail people receive as a result of credit card data mining.
Craig – The current conversation is technologically naïve. The cookie law is responding to last year’s issue. How do you market the advantage to the consumer? No advertising is not an option: the options are untargeted advertising or targeted advertising.
Andy – People think remarketing is neat initially but then it becomes spooky. Things won’t go as far as Minority Report because if advertising goes too far consumers will complain.
Anthony – In online space a lot of what we’re seeing is a reaction to what marketers took to excess, eg adware. There is general mistrust of what we’ll do with information. Governments have slowly caught up with this because of flak from constituents. So we have to explain that remarketing can be a good thing. We’ve put ourselves in the position where we have to explain ourselves.
Andy – The problem is partly that advertisers are not managing their retargeting correctly. Advertisers have to use it right so that consumers appreciate it.
Craig – After people drop out of the conversion process and get retargeted to, most conversions happen within 4 hours. There’s not much point retargeting
What are dangerous cookies? Andy said that the fear with cookies is that we’re taking more data than we need. There are rare instances of people misusing cookies.
Could the cookie law make smaller companies more dependent on larger players, like Google and Facebook, who have access to personal data? Anthony said it does depend how large you are how much info you can collect. Andy pointed out that big companies have the advantage of more data but also have harder time complying; small companies can be more modest in complying with cookie law.
Andy – Google is signing people in automatically. You could say that ‘signed in’ is the new ‘cookie’. The danger for the industry is that people push that too far and that need it needs more legislation to regulate. The industry should self-regulate before that happens.
What’s your analytics plan? Build a plan around marketing calendar, sales funnel, key initiatives, or channel/discipline.
It’s easy to ignore dashboards when you get used to them. But they can be useful.
Develop key performance indicators over time – not a long laundry list of metrics, they should be absolutely key. Not seen any that are not measured over time.
Then plan ad hoc exploration.
Despite its ‘pretty appearance’ Google Analytics has robust dashboards and pretty powerful reporting.
They have a ‘demand gen KPI’ dashboard. Look at 5 or 4 week period and compare with the previous
Tip – if you’re looking at short periods of time, eg 5 weeks, then if one period starts with a Monday then compare it to one that also starts with a Monday so the days of the week align.
Transfer dashboards between profiles – be careful as conversion metrics may not be lined up.
Use custom reporting and then add segments.
Look for meaningful spikes when there’s some segmentation. Eg spot above average conversion rate for paid search / non-brand traffic – did you run a newsletter? The spikes are actionable data.
Use standard multi channel funnels for PPC – use ‘ad group’ as primary metric and ‘AdWords keyword path’ as secondary metric. You can then work out which keywords are working best.
Set up conversion values for soft conversions, eg sign ups for webinars or email list subscriptions – consider how much you’d spend to get someone to convert and use that as the value. Then you’ll be able to get relative value of campaigns.
Use customised channel groupings to separate brand versus non-brand keywords in organic and in paid.
Use custom multi channel funnels to find the most effective campaigns or keywords.
Sampling – watch out for the ugly yellow bar that tells you the data are sampled. The more data you have the more likely this is. You can press the button (“looks almost like a bingo card or something”) and drag the slider to maximum accuracy, but it’s still going to be sampled.
To avoid sampling, you can add ‘keyword contains’ as part of the dashboard, so you don’t need segments – this limits you to the top 10 results, but it might get you some long tail clues.
The conversion optimisation overview is great for CMOs – it shows the combined effect of multiple channels.
Custom variables are the most important feature of Google Analytics by far.
Define additional segments to your visitors – eg if they’re subscribed to your newsletter, you don’t want to show them the subscription form again.
It doesn’t come out of the box – you have to add some code to use this feature.
There is a hierarchy of custom variables. Women have:
Similarly on Analytics, there are:
An example of a visitor level variable would be if they are registered or have made a previous purchase. You could also use demographic info from forms visitors have previously filled in, like gender or age.
Custom variables can be used for campaign attribution, if you have the original campaign that led them to the site as a variable.
You can use custom variable to see which promotion banner on your homepage someone clicked, to see how effective the promotion was.
Segment or die – you can’t provide better experience on website without it.
For more information see http://Onbe.co/CustomVar
You can see Anna’s slides here.
Social reporting was added to Google Analytics recently. It shows you:
It works on social platforms that Google has access to, ie places they own: Google+, digg, delicious, reddit, etc. The list links as ‘Trackbacks’.
It’s like multi channel funnels in that you can see what assists conversion. You can see where social comes into the user journey, and how can you improve it, by using secondary dimensions like “Keyword (or source/medium) Path” to see more detail.
There are many analytics tools, like Google Analytics, comScore, Yahoo Analytics, Omniture, etc. These tools are focused on data collection, not visual representations. All have export functionality.
Using Thomas H Davenport’s ‘Maturity Model’ you can see what level your analytics is at. Some clients on stage 2, ‘localised analytics’ – they’re focused on the ROI of individual marketing channels. They need regular reporting, which is probably automated. At stage 3 you’re beginning efforts to see integrated data and web trends.
People don’t have clear understanding of what some metrics are. To make real sense you need segment or context.
For example duration of visit might be used when there isn’t a proper conversion (eg for informational website when you want people to read your pages). But duration of visit doesn’t have the length of time you spent on last page. And for info websites it’s likely you spent a small amount of time navigating to article then a long time reading that article, so ‘duration of visit’ is unlikely to be accurate.
Need event tracking or custom variables for insights into shopping basket behaviour.
You can have events to see if people click on footer or header links, to see how people navigate, but this can slow the page down.
Use video measurements to monitor your video adverts – how do visitors interact with adverts, what’s ideal length or the best placement? You can see the viewing behaviour of videos, how many completed the video. You can see who watched from different sources to help promote viral marketing.