Excellent Analytics Tip#3: Turbocharge Your SEM/PPC Analysis

CactiSearch Engine Marketing / Pay Per Click, is the hottest thing around. Everyone is doing it and if you are not doing it you dare not admit it for the fear of looking crazy. There are more agencies and consultants out there than you could care to count and all of us are trying to work with all of them trying to do our best.

My hypothesis is that while a few companies are laughing all the way to the bank as a result of their SEM/PPC campaigns the rest are probably not maximizing return on investment.

We in the Web Insights world can play a killer role in helping our companies truly measure effectiveness of our SEM/PPC campaigns. We are unique because we can cut through the crap because of our awesome analysis powers, our access to all the data and our ability to think beyond base metrics like CPC or CPM. Yes people are doing CPA now and we are even measuring Conversion Rate by Key Phrase. But there is more.

Here are my suggestions to us Web Insights Analysts for providing deep insights to our decision makers, and potentially a competitive advantage to your company:

# 5: Measure your Bounce Rate (aggregate and by top key phrases)

My definition of Bounce Rate is someone who stays on your website for less than five seconds. Some folks define this as visitors who stay on the site for one page or less, this is fine as well.

The conversion rate for most sites, per shop.org’s most recent study, is around 2.2%. When we do PPC/SEM campaigns we are thrilled to get 3% conversion rate and we ask for more money to spend. But since we are paying for this traffic we should expect much more than a bit better than normal conversion rate.

Measuring the Bounce Rate for SEM/SEM traffic in aggregate is an eye opener. It usually turns out that the Bounce Rate is really high (60%+, often even higher). Decision Makers are shocked that 60% or 80% of the traffic stays for such a small amount of time because we are paying for each click at the search engine.

It gets even more interesting if you measure Bounce Rate by Key Phrase. This will highlight stinkers very quickly, phrases that drive traffic that stays for five seconds or less (or only one page).

The point of the recommendation is to really dig into your traffic. It is possible that the key phrase is the right one but people are landing on the wrong page. It is possible that the phrase is ok but you are not repeating the value proposition on the landing page, or the offer is missing etc. Once you know you can fix, and oh do set the bar higher for PPC/SEM campaign conversion rates.

# 4: Audit you Vendors/Agencies

The de-facto standard is agencies are running campaigns using some very sophisticated analysis (bidding) and measure “success” using third party javascript tags on your website. This is good progress. But there is more.

There are often problems with third party cookies. Each agency seems to be using a different kind of data capture and data processing system. They also define the same metrics differently (even metrics say like repeat visitors or visits to purchase) and they don’t do any analysis on site effectiveness because they are, rightly, optimizing your bids.

My recommendation is to request (or force as the case may be) your agency to add an extra url parameter at the end of the URL string that goes to the search engine. This should be a parameter that your analytics tool will recognize (usually something simple like kaushik.net/avinash?source=google-avinash_is_nice will do). Now power up the segmentation in your tool and slice off the SEM/PPC traffic and match up key metrics like count of visitor, time on site, bounce rate, conversion rates using your own analytics tool.

The numbers will not tie, but if they fall within 10% you are fine. What you will usually find is they will be off by 20 or 30%, that would be a great time to compare notes with your agency because if they are measuring success 30% higher than you then it is likely your campaigns are not performing as well as you think.

Another benefit is that you can now truly, with inherent knowledge of your business that your agency does not have, analyze what this traffic is doing and understand their behavior and work on site optimization.

I have to stress that the Agencies are doing the best they can, they are not trying to mess with you. It is simply that there are limits to what they can do and you need to step up and take your responsibility.

# 3: Measure Cannibalization Rate (vs. Organic)

Another false measure of success, and very prevalent, is that we start buying our “branded key phrases”, traffic shows up, we pay google and overture, there are sales and success is declared. But if you have done decent SEO (search engine optimization) it is possible that you might still have gotten traffic from search engine (especially your well recognized branded key terms) from Organic results. Do you know how much you are cannibalizing your Organic (free) traffic by doing PPC (paying per click)?

This is a very powerful metric and something you should be watching all the time and based on the results of that analysis optimizing your SEM/PPC spend. Many good Agencies are doing this, demand it from yours.

I will share a very simple way of measuring cannibalization. Take some of your most popular branded key phrases, assuming you don’t have heavy weekly seasonality factors, and turn off your campaigns for those key phrases for a week. Measure number of visitors, bounce rates, engagement with site (if you have that metric)  and conversion rates. Turn on the campaigns next week and measure the same metrics again, for the PPC/SEM slice. Compare the two sets of numbers and you have a nice idea of how much you are cannibalizing your Organic (free) traffic when you spend on PPC.

My recommendation above is not the most awesome, but it is a really simple way to measure cannibalization. I am sure you can think of ten better ways, what is important is that you measure cannibalization and then make informed choices. Why pay for traffic that you could get for free?

# 2: Aggressively push for Testing & Experimentation

Testing and experimentation is not normally considered the scope of web analytics, but it should be. One of the beautiful things about key phrases is that we can infer intent from them. It is likely that 60% of your site traffic is No Referrer and you don’t know squat about them and you can’t give them relevant content, but with your search traffic you know something of value.

Partner with key marketers in your company and implement a testing strategy (multi variate, AB, DQB, RRR does not matter) that will allow your to try different content, pages, structures, layout, navigation based on what you understand about the intent of those key phrases. The result will be optimal customer experiences for your PPC/SEM traffic and a huge spike in conversion rate, whatever you are selling.

(And if you set persistent first party cookies you can “remember” the SEM/PPC driven traffic and the next time they come you know who they are, which test experience worked for them, or did not, and now your site can start to have much richer interactions with them.)

# 1: Strive for “Multi Goal” Understanding of Visitors

No website exists for just one purpose, not even the simplest one. Yet almost all SEM/PPC campaigns are measured by one outcome goal (usually a conversion rate of some kind, purchase or a lead submission or visiting a page listing your core content). If the traffic comes for a few different purposes why measure them all by one goal? (To beat a dead horse, this is another reason to do testing.)

For your SEM/PPC traffic set multiple relevant goals. For example along with Conversion Rate you can also measure Customer Satisfaction and Task Completion Rates by surveying that traffic (if your web platform allows that). This will tell you oh 10% are converting but the others were looking to print pages or your tech support number or came to read reviews. You can then take relevant action to optimize your campaigns.

Or you have a frequent turnover site, set goals for Multiple Purchases in each Month, great complement to just conversion to first purchase. Or here is a nice one, if you follow the advice in #4 above, amass data for x period of time then compute “Live Time Value” for SEM/PPC campaigns and use that to optimizing your bidding as well as set a new bar for success. Bottom-line, if you are measuring “single goal” success for your campaigns, there is opportunity to get more sophisticated.

As usual ; ) I have one bonus recommendation for you…….

# 0: Help Measure the Value of the “Long Tail”

The concept of the long tail, as applied to Search Marketing, states that there are hundreds of key phrases that individually drive small amounts of traffic to your website, but if added together they can be a huge amount of, potentially profitable, traffic.

Thanks to our current analytics packages we are all used to only looking at Top 50 key phrases from each search engine. Maybe you look at the Top 150 because after that you have the “onesies” and “twosies” and we don’t care. But the delightful thing is that these key phrases are not a part of cut-throat bidding process in ad-words or overture or ad-center.

The implication is that if you can identify your “long tail” and make that a part of your SEM/PPC campaigns (assuming you are using some automated bidding system), you can capture a large share of this traffic and do so at very effective costs and turn great profits for your company. 

There are limitations to identifying the long tail just using your clickstream package so use other tools that are on the web and/or ask your Agency. I also recommend HitWise as a great source. But help your company look beyond the obvious / “top” key phrases and go after the long tail and help them measure the value of the long tail.

What do you think? Do you agree with these recommendations? Most of you are much much better at PPC/SEM than I am, do you have suggestions to add? Please share your feedback / critique via comments.

[Like this post? For more posts like this please click here.]


  1. 1

    Hi Avinash,

    I Absolutely agree. After reviewing the bounce rate reports in depth, we can prevent the click fraud
    upto some extent.

    The Line "If the traffic comes for a few different purposes why measure them all by one goal? (To beat a dead horse, this is another reason to do testing.)" This is where most of agencies are lacking and Reporting the success, while they dont think about this at all.

    And as your Bonus Recommendation suggests , After analyzing the keyphrase referal in detail we can use more effective keywords for SEM/PPC.

    Waiting for your new post to learn something new.

    Best of luck

  2. 2

    Intersting stuff; not all analytics packages will allow you to pick up the bounce rate of a keyword (don't know any that do that offhand, but I'm sure some will do that). Sounds like ClickTracks does that as I now that's what your using at Intuit.

    Adding the Parm to the url is a nice, and easy thing to do in most cases (except where I happen to work – where every single change needs to be examined and signed off on – we probably could not easily to it there). But for everywhere else, yes, I think this is a great idea. In fact, you want to be able to match up the traffic the agency says ther're buying for you vs. what the analytics records you got – as well as the other measurements which give you a lot more behavioral data.

    I like the Measure Cannibalization Rate (vs. Organic as it's easy to measure (again, it depends on how easy it is to convince the client to turn the PPC off for a week so you can gather the two sets of data). I posted something on this last March http://www.webmetricsguru.com/2006/03/cannibalization_by_organic_sea.html

    Now the long tail is a very interesting subject – worth a couple of posts (and intersting to any large site with a lot of content – much of it deep in the site). The Long Tail, as applied to Search is the 3, 4, 5 + word queries that there are only one or two searches for a month- but in aggregate, can make up half or more of your traffic. The Long Tail, as it applies to Advertising is something else – it's all the possible sites your ad can be seen on, including viral media.

    Studying the Long Tail and how it applies to various online industries is facinating.

    Sounds like you were busy listening to someone, we know, who knows a lot about this stuff, and some of those ideas got into the post.

    I'll write up a post or two on this at Webmetricsguru later on today.

  3. 3

    I have spent some time this morning rereading your Excellent Analytics Tip #3 post. And I also went back to some earlier threads, in particular following the discussion about cannibalization of organic results.

    I have come to the conclusion that you — and probably a majority of your readers — know a whole lot more than I do about SEM/PPC.

    I'm a relative newbie to Web Analytics, with experience in more traditional forms of marketing/communications and a broad knowledge of Web development and electronic communications. Basically, I have had the good fortune to hang around Robbin Steif long enough to know that a marketer/Web developer like me must understand Web marketing and analytics in order to do justice to my client's goals. Right now, I work with a lot of graphic designers and design firms who know they need to offer Web marketing expertise to their clients, but they don't have that expertise themselves.

    In a former life, I studied the history and philosophy of science at the University of Pittsburgh, and I like to think this allows me to bring a bit more to the table than just an ability to crunch numbers and give an effective presentation.

    Sometimes the opinions of a relative outsider can be very valuable because it's hard for "insiders" to see the forest for the trees.

    But, when it comes to Excellent Analytics Tip #3, I think you have got a better handle on the issues.

    I must say, I see some strong similarities between the Web analytics community and the scientific communities that I studied as a grad student at Pitt.

    History and philosophy of science tries to understand a lot of different aspects of science, especially those that are not examined very closely in the normal day-to-day proceedings of science and society.

    As I follow Web Analytics discussions like the one that ensued after your Top Ranked Web Analytics Blogs post, I find myself wondering a lot about the power of argument and presentation of evidence. I think you have alluded to this aspect of your job in your blog a number of times.

    As analysts, we may have what we consider to be an air-tight case — or at least a respectable case — for doing one thing or another, but, unless we can convince the client (our boss, etc.) and/or our peers of its validity, it does no one any good!

    There are a lot of "non-scientific" — or "non-analytic", as the case may be — elements that go into decision-making. At the very least, the definition of what is considered "scientific" changes over time.

    Thanks again for spurring discussion!

    Mark McLaren

  4. 4

    Mark: I would hypothesize that it is your ability to bring more than "crunch numbers" to the table that makes you super valuable. There are a lot of us who might be good at number crunching but in the world of Web Insights it is the ability to step away from the numbers and look at the whole picture with the business context that is important.

    Your philosophy background probably makes you unique because it gives you a fantastic ability to relate to the "web interactions", beyond just "hits" that the numbers show.

    IMHO that's the difference between Web Analytics and Web Insights.

    Thanks for taking the time to comment, I appreciate that very much.


  5. 5

    Your long tail comment is right on the money. We manage PPC/SEM campaigns here at Agito Internet Marketing and have found that single strategy to be the life saver of many accounts.

    I can think of three instances recently where the client felt they were not getting acceptable ROI on their PPC campaign – and we saved it by appliing the principles of the long tail.

    Basically, by taking a PPC campaign and expanding the keywords from dozens, to hundreds, the traffic increased dramatically and the average CPC dropped by the same.

    You can use your analytics program to find the actual phrases people used to get there, and keyword research tools to find other obscure phrases and then bid on these.

    Excellent blog – I would never image you are a newbie :-)

  6. 6
    Gradiva Couzin says

    Hi! Thank you for the excellent post! All very good suggestions. As an SEM consultant I have struggled with clients who have a hard time committing to any sort of metrics at all. The best way to get around this is to make the metrics "actionable" (also to make sure you only report things that the client's can easily understand and digest).

    Cannibalization of organic by PPC is definitely something I've seen, and it's quite sad! We work to do the opposite: provide a PPC campaign up front but then phase it out as the organic positioning takes over we may provide a PPC campaign up front but then phase it out as the organic positioning takes over! This is a bit of a twist on things but it is one way to provide a more measurable ROI for organic SEO.

    Thanks again for the helpful blog,


  7. 7

    Search Marketing from a Web Analysts View.

    Web analytics must be the theme for the day as Avinash posts some useful advice for managing a PPC campaign from a web analysts view.

    Some of the great advice includes measuring bounce rates and determining whether you are matching the best keyword buy with the most relevant landing page. He also recommends adding tracking parameters to your PPC URL strings, allowing you much more granular tracking.

    I'd definitely add that companies working with agencies should share as much analytical data as possible. Too many clients guard their web data as if it's the secret recipe to KFC. For us search marketers, it's like navigating a plane through a thunderstorm without instruments.

    Avinash also suggests that companies could be cannibalizing their organic traffic by bidding on keywords they rank well for organically. He suggests that you try turning off your PPC bids for a week and measure the impact. I'd certainly caution against this, especially if you are a relatively unknown brand. While "big brands" can often maintain their traffic levels from organic alone, most smaller brands need the positive reinforcement, projected to potential visitors, that comes from having both an organic and paid listing. If you decide to take Avinash's test, be sure you are acurately measuring all metrics, not just "did we lose a lot of click-thrus".

  8. 8
    Matt Donovan says

    This is an excellent article from beginning to end. As I am fairly new to the search marketing industry, I find it interesting that many companies, when conducting a PPC campaign, forget some of the more basic business rules. For instance, measuring "multi goal" understanding of visitors should be a critical piece of any campaign. However, I would be willing to bet that every agency has measured success of a campaign by only one goal. It is imperative that multiple goals are established for each client at the outset of the campaign. However, the agencies must be diligent in monitoring the relevancy of each goal to ensure overall effectiveness of the campaign. For example, if a client states their goals to be drivin traffic to the site and increasing conversion rate, the agency will have to deal with competing priorities. More traffic lends itself to less qualified leads, thus ending in lower conversion rates.

    As for Testing and Experimentation, don't be afraid to think outside of the box. There are many web analytics tools out there that will assist with the structure and monitoring of AB testing. Some web analytics providers provide assistance in creating multivariate testing as well. With my background as a statistician, this is one of the most intriguing uses of experimental design that I have seen.

    In closing, I wanted to comment specifically on cannibalization. You will rarely (if ever) find a client that will be willing to shut down a PPC campaign for a week just to test cannibalization. We have designed a new way to track cannibalization by utilizing historic data (sometimes gettign this is like pulling teeth!). Using robust forecasting techniques that track chronological fluctuations, we estimate future organic traffic. As we implement our PPC campaigns, the "new" organic traffic is compared to the forecast to provide a decent estimate of cannibalization.

  9. 9

    Matt: My hope was to push the conversation about cannibalization to the forefront becuase my hypothesis is that most of us don't even think about it (and it is costing us real money). I agree that going dark for a week is very hard but people don't possess excellent methods like the one you outlined then push to go dark or do whatever it takes to figure out your cannibalization rate. I bet people will be surprised if they know what their number is.

    Thanks for the comments, it is my favourite part of blogging.


  10. 10
    Dylan Lewis says

    In addition to measuring and improving the "bounce rate" to increase user satisfaction and finding the keyword stinkers, you are ultimately providing better content to search engines.

    Better landing pages will increase your quality index score in Adwords which should help reduce your costs and increase your exposure.

    Here is a recent article that explains why landing page quality is not only good for your site but good for your PPC pocketbook:

    Now there is a another economic incentive for making your landing pages match your keywords to provide the best experience to your visitors and to the search engine visitors.


  11. 11

    Thanks for this, I'd never really considered looking at multi goals and trying to identify what people's goals are from THIER behaviour rather than what we decide thier goals should be.

    One other thing I would mention is dynamic keywords. I use these for long tail, I put all of the long tail words in one group and add a dynamic keyword into the ad copy so the word they searched for shows in bold in your ad, with little comptetition from other ads you should get a high CTR

  12. 12

    Excellent Blog. Thorough and practical to implement. I am on my way for a job interview & find this to be one of the best sources of information on web analytics. THANK YOU!

  13. 13

    Translation for this article into Chinese is done just now, via: http://youping.cn/archives/391


    What is the abbreviation for DQB, RRR?


  1. Turbocharge Your SEM/PPC Analysis – Avinash Kaushik…

    I promised I would devote a post or two to Turbocharging your SEM/PPC Analysis which Avinash just covered in his blog.  Too exhausted to do a long post but I'd like to add a couple of things to the comment I……

  2. […] One of the advantages of online marketing is the ability to track visits and conversions on your web site and to connect them to your campaigns and your sales results. Avinash Kaushik has written on making it easier for business people to connect to the analysis of online marketing programs and has some great ideas on turbocharging your search marketing PPC campaigns. […]

Add your Perspective