Commentary
For Digital Publishers: Ditch the Survey and Embrace the Scientific Beauty of Web Analytics
- by Philip Sawyer , January 28, 2010
After 23 years of conducting advertising research, I have reluctantly come to this conclusion: Most digital ads aren't very good, and unless advertisers alter their approach to digital-advertising research, many are likely to assume the medium doesn't work for them and prematurely give up on it.
Why are so many online ads so very bad? It's not because the medium is inherently weak, as some suspect. Rather it's because very few online advertisers are getting useful feedback about their online efforts and so aren't exploiting the medium's extraordinary power to create dynamic and, most important, mutable advertising that consistently attracts and holds reader interest -- and promotes sales and branding.
Here is the problem. Most online advertisers are still using research methods that made sense in the last century and for the advertising vehicles of the time. Those research methods employ what we might call an "exogenous" approach: the advertiser and research are on the outside looking in, using surveys with small samples and asking respondents to report on their attitudes, buying habits, and advertising readership. Of course, there are several problems with this approach, two of which are that: a) some respondents lie, and b) the rest, being human after all, are often inaccurate, uncertain, or incomplete about their attitudes, buying habits, etc.
Yet, some digital publishers still use sample-based online surveys to measure evaluate advertising effectiveness for their clients, and none of the information gathered from a survey is likely to provide advertisers with the kinds of insights that can turn a weak advertising campaign into a juggernaut.
It doesn't have to be this way. With a little creativity - and a willingness to break new ground - an online publisher can move almost any advertiser from the outside to the inside: from simply observing and waiting for advertising results or for a research report, to delving deeply into the center of an advertising campaign, getting vital intelligence about what's working, what isn't, and why - and, most important, quickly enacting and testing changes to increase the power of the ad until it starts to fully realize its goals.
Surveys don't allow you to do that. Web analytics does. And if they're going to thrive, publishers have to become more adept at web analytics to demonstrate their true value, which is real and considerable, but awaits discovery through the use of accurate measurement tools.
Advertisers don't need surveys, not when they can harness the flow of information at far less cost with web analytics, which delves deeply into the center of an ad campaign and allows you to monitor and test creative elements (headlines, layouts, photography); engagement with the material (time spent on a page, downloading of featured items, usage of links to Facebook, etc.); and action elements (quality and content of inbound phone calls, coupon downloads, requests to contact representatives). No survey can even approach the quality and utility of this kind of advertising information.
With this kind of information, the advertiser gets real-time, behavioral feedback, not fallible memories of behavior. Perhaps more important, with the analytic information in hand, the advertiser can immediately improve the creative, an opportunity that no other medium provides to this extent. And the publishers who can provide advertisers with this service? They reclaim the client relationship from ad networks and earn client loyalty for providing such a valuable service.
This win-win approach isn't new, just recontextualized: Right now, Google provides and promotes the Google Analytics system to its advertisers for free. The reluctance of so many publishers to provide analytics tools for their clients has negated the Internet's great advantage - its measurability - and publishers have missed the opportunity to demonstrate their value, which our analytics data suggest is considerably higher than they - or their advertisers - fully appreciate.
Survey research, done right, is still a valuable tool for print and broadcast advertisers. But digital advertisers who ignore the power of web analytics in favor of survey research are more likely to approach excellence than to achieve it.
While I agree w/ Philip that web analytics should be an important component of any marketing effort, I don't think it makes sense to ditch the survey as a total replacement - the attitudinal analysis derived from a questionnaire is just as important as the behavioral.
He is right that human respondents can potentially provide misleading feedback (intentional or perhaps not); however, web analytics can also mislead. What about the scenario where a visitor comes to a website from a particular display ad...surfs around looking at merchandise....and then abandons the session to ultimately purchase from that company's physical store? The web analytic analysis would deem the display ad referral a failure and decisions might be made to scrap or alter the creative; attitudinal analysis based off survey feedback would paint a different picture, with high visitor satisfaction levels and offline purchase intent. For the "Research Online/Purchase Offline" segment, the display ad creative did the job and should get credit for attracting the right type of visitor - and the survey would quantify this dynamic...not web analytics.
When publishers have both techniques at their disposal, it's like they have a full brain: web analytics is like the left side, analyzing the "what" via behavioral tracking; survey feedback is like the right side, assessing the "why" and "what next" through attitudinal evaluation. Both are critically important to any publisher marketing effort.
There’s no excuse for not tracking phone calls and transaction details through your web analytics today, but some of the more challenging issues involve tracking web 2.0 interaction: Facebook pages and groups, Twitter profile interaction, mobile devices, content on blogs and forums run by people outside your organization, RSS feeds, local business profiles, etc.
Tracking and measuring all of these channels can be a real challenge, especially when you want to understand how these various web channels impact lead generation, product sales and brand name awareness. Web analytics alone is inadequate for collecting target audience data in today’s web 2.0 environment.
Here is a list of excellent resources, research, services and tools that equip marketers to know, strategically plan for and target their market:
http://www.capturecommerce.com/audience-engagement.php
Qualitative metrics (including survey results) are just as much a part of web analytics, as quantitative metrics is (including your example of clickstream data from google analytics). Clickstream will tell you the "what" whereas qualitative metrics like survey results will tell you the "why". If you're only using quantitative web analytics, you're missing a key piece of the web analytics equation, and one that can bring much-needed context to the numbers. Both halfs of web analytics are equally important. And in my opinion, I see it the other way around: far too many organizations are paying attention to quantitative metrics, and completely ignoring qualitative.
*disclosure, yes I work wth Phil. Like the thread of thought here and wanted to throw in two cents.
Your website is a 24/7 survey. Thinking in this manner, ask your analytics different questions and set different goals. This creates an entirely different question tree, and result.
Eric, points out the problem that is true in many analytics methods. Goals are set for only one type of visitor. In the example, visitor educates his or herself on website goes to physical store and analytics show this as a failure. Not true, if you set say an Engagement goal, say of duration - that you want to set a goal that says: "I want 40% of the traffic from this display ad to spend 5 minutes browsing in the online store." This result has a value to the product - a brand value, and the display ad (and the publisher) isn't getting credit for this value.
Finally, you keep your online actions whether online purchase or not). You now have much more experiential/behavioral analysis simply by changing the goals per the "visitor demographic" and adding engagement goals like time.
The recent additions in Google Analytics to set Engagements, to increase the number of actions per profile, and to add time as a potential segment class I think is a very important step that can help foster this idea.
De-quantifying analytics is simply done by asking the analytics data "Why?" instead of asking, "How Much?"
Thanks for the post, Nick - your point about setting web analytics goals is a good one....interrogating the behavioral data to drill deep into a visitor session will produce further insights.
However, there is still a risk in just looking at the behavioral data against a specific goal - in your example, "Time Spent" as a rule for "Engagement." If you set the threshold at 5 minutes, and the visitor only needed 3 minutes to conduct the purchase research (i.e. the display ad was successful AND the website was extremely efficient as a research tool in driving offline behavior...), then web analytics alone will still be misleading.
We conduct a lot of Engagement type analysis w/ clients - sometimes longer website visit duration equates to positive Engagement....and sometimes a longer visit is a negative (i.e. Navigation is poor, the visitor is trapped and won't exhibit positive future behaviors based on that visit.)
Great discussion!
Like many of the other commentators here, I agree that web analytics provides part of the analysis, and a very important part. But web analytics can be combined with data from other tools to avoid the "last click" problem, attributing too much of the credit for the sale to the last site visited before coming to the web site (usually the search engine). Attribution tools, like those from ClearSaling and VisualIQ, as well as Ad effectiveness metrics (like those provided by Adometry and others) can be combined with web analytics to provide a more complete picture of the purchase funnel.