Searcher Behavior Research Updates
And we’re back from the fastest lunch I’ve ever eaten. Where does the time go? Moderating this session is Bill Muller (iProspect) and our panelists are John Marshall (Market Motive), Pavan Lee (Microsoft), Dr. Larry Cornett (Yahoo! Search) and Bill Barnes (Enquiro Search). I have to confess I just love the research sessions. Hard data just makes my little heart sing. Come on. You can’t tell me you aren’t excited about this one too.
I know you are because Bill is telling us about how every year this session totally fills the room it’s in. Why? Because if you know more about the way searchers behave, you’re going to be a better marketer.
John Marshall starts us off.
The interesting thing about search behavior is that it’s not that difficult to get good data. The question is on Monday morning, do you understand search behavior? Most people turn to the keywords report in your analytics tool. That’s a reasonable place to start but it’s an extremely narrow view of the activity on the Web. You’re only look at the keywords that brought people to your site. You only see the search results that brought people to your site. You’re running into sample bias.
How can we really see the intent of people, not just the people who made it to your Web site. You don’t want the whole forest view. You’re probably not going to get the whole forest view anyway unless you have a Hitwise account or something. What you can look at is the single tree of your site.
The trick is to use the site search on your Web site. If you don’t have site search, implement it. Even if it doesn’t work, it’s a great source of user intent. Search engine keywords only give you the people who came to your site. Site search gives you the intent of your users, conversion rate information. A lot of people ignore this data because it’s free. Free data is often ignored. If you pay for something, you value it more.
Things that can go wrong:
- Mixed Case — Google Analytics doesn’t automatically change case for you so your data gets scattered across case. You need to convert it.
- Multiple results pages — Some site search pages for ‘no results found’ don’t get tracked by analytics. Make sure all the pages contain your analytics tracking.
- Injected terms — Most Web sites that have site search, they use it as a cheap landing page creation system. You have to filter that out of your data if you’re doing that because it’s not real data. No one is typing it in.
By using site search you’re answering the question: what’s the true intent of the users when they’re on the site.
Site search data cannot replace competitive analysis but it’s the cheapest way to get good data fast.
Pavan Lee is up next. She’s from Shanghai.
Background from the New York SES: They’ve discovered that search listings have a branded value. Paid search listings have a stronger branding impact than organic search. There is a positive branding effect for both. They’re trying to measure the brand lift.
They studied five brands in five spaces.
Methodology: Eyetracking and post-search survey.
Key findings: Search display and content ads are effective for branding stand alone but more effective together.
They asked “did you remember seeing an ad” 21% lift content 30% display. 38% both.
I can’t see her slides at all.
In all cases with all questions, including lift in purchase intent, there was a brand lift and it was stronger for paired ads.
On the eye tracking side, search is still the most effective tool in attracting attention. There’s a roll over impact on a multi-channel exposure. If you see search and display or search and content or search and display and content, it’s more effective than just seeing any one of those.
Key takeaway: The power of three. There’s a synergistic branding impact across content, display and search ads.
None of this data is public information.
Larry Cornett steps up to the podium.
His talk will build on John’s presentation in a lot of ways.
Users do a lot before and after they’re on the search page. He’s going to talk about that, about the research they’re doing, how users experience search, a little about crafting search and how they get from ‘to do’ to ‘done’.
The reality is that the search page is just a tiny slice of online activity. Before the search, the user somehow comes to need to do a search. After the search, they want to go somewhere. They’re going somewhere because they want to fulfill a task. The task is not getting to best buy. It’s getting an iPhone. You need to know what happens after and how it all links back. How do you support them through the whole lifecycle of what they’re trying to accomplish.
There is no single methodology that gives you the whole picture. Some ways that Yahoo does testing are:
- Search editorial
- Bucket testing
- Metrics & Analysis
- Search Science
- Focus Groups and surveys
- Eye-tracking research
- Ethnographic studies
How users experience search
- Starting context (what have they seen and experience before they query)
- Quick Scanning (one to three seconds)
- Information Scent
- Matching intent
- Quick Decisions
- Looking for answers (not a homework exercise. “Don’t make me work”)
- Feeling safe
They try to help crafting searches with ‘search assist’ (suggested searches). For most people search is hard. They’re not experts.
Focus on the ultimate goal. They’re looking to do something, they’re wanting an answer. Yahoo SearchMonkey is an attempt at giving them that answer. It gives the user more information about what’s behind the link and what’s important to know.
What does this mean for marketers?
- Before the SERP
- Starting context
- The “Real task”
- On the SERP
- Intent and information scent
- After the SERP
- Fulfilling expectations
- Being their “answer” and living up to the promise of the search result.
We thought that the reason people were having trouble with search was that it was an artificial session. But field studies showed us that the users were really having trouble formulating queries so we really tried to implement something that would help them.
Bill Barnes is the last to speak.
Their research is grounded in their search marketing and grew out of that.
Why is the first listing seen so important
Why do we scan in groups of 3 or 4
Why branding is important
[Standard heat map image, you've seen it a million times.]
They did experiments with the top SPONSORED listing and played around with really great ad copy versus just ‘okay’ ad copy.
When they did a survey they didn’t ask about the listing, they asked about the search engine and if they’d use it again. The only difference was the ad copy but there was a huge lift in trust in the engine with the great copy.
Working memory: It’s what comes to mind with recall. For some reason, we’re hardwired to think in threes or fours.
[Oh no, my battery is dying]
There’s a 16 percent increase in brand association when brand is the Top Sponsored and Top Organic Results. On an unbranded query. The really interesting thing is that the recognition of OTHER brands drops away at the same time.
There’s an 8 percent lift in brand purchase. If you’re not there, you lose 16 percent brand lift.
Even for branded queries, you get a brand lift if you appear. Should you buy your branded terms? Yes.
Eyetracking finding: Brand fixation only occurs in the TITLE and the URL not in the description.
If you’re a familiar brand to the searcher, they will often skip the sponsored listings at the top. If you’re buying the top sponsored, write your copy for a NEW user.
If you have brand A and brand B in sponsored with Brand A in top organic, brand A gets a HUGE lift.
- INTENT is the most important thing
- Organic and sponsored combined give the biggest brand lift.
- Be aware of who else is on the serp
- Write your ad copy to new clients.
- Don’t assume your brand will be in the consideration set. If you’re not on the page, you’re forgotten.
The first question is does offline affect offline. The answer is yes, though the panelists don’t say that. Go read the Re Search Online, Purchase Offline session from yesterday.
Why do search views get longer?
Pavan thinks it’s because searchers are looking for something in particular whereas display and content ads are push forms of advertising.
Do they really only spend 1-3 seconds and how often do they click?
Larry: It’s on average. In some cases, for navigational queries, that’s less than a second. It might be longer at home but yeah, it’s amazingly fast.
Bill: Females look longer and shop around, males just go straight to results. There’s a free paper available.
Pavan: Search intentions lead to searcher behavior. Fact based search stays organic. Commercial searches tend to be more broad. It also varies by culture. Chinese spend twice as long as Americans.
John mentions that his contention is that the site search queries are the same queries that are being typed into the search engines but they’re just not getting to your site.
Is the suggested search condensing the search queries?
Larry: Yes. People are moving to longer queries and that search assist does jump them to the queries that will get them to the answer faster.
Are there differences in lift by categories?
Pavan: Yes there is a difference in lift across different verticals but in all cases it does result in lift.
[Long set up about pretending to be a confused searcher and poor SERPS] What can be done to help confused searchers?
Larry: Search assist is just one way. It works mostly for shorter queries?
Does the golden triangle change with non-roman character sets?
Pavan: In Chinese, the scanning pattern is very different. It’s a rectangle. You have to look at everything to put together meaning.
Would you suggest not trying to dominate the organic?
Bill: No, never. Always optimize.
What plays into search assist? How does it affect PPC?
Larry: Nothing is paid in those.
John: The hidden message there is: No you can’t spam the suggestions.
John says that the other thing site search is good for is manifesting usability problems and for doing competitive intelligence.
If you rank 1 on a non branded term, should you also be number one in paid search as well?
Bill: That’s exactly what our research showed. That said, always test and retest and see if the ROI is worth it. Clicks went 50/50 on paid and organic, so make sure that you’re testing and monitoring.