Archive for March, 2008
Posted on March 21, 2008 by Traci
For anyone who is reasonably new to the world of Pay-Per-Click (PPC), let me share a few words of wisdom — if you don’t have an extensive list of negative keywords, you’re probably paying too much for your traffic.
Suspect you might be one of those advertisers who is feeding the Google piggybank? Well, let’s put a stop to that! In this post, you’ll learn what negative keywords are and how you can generate a starter list of negative keywords in no time at all.
So what are negative keywords? Let’s start with an analogy. When you do pay-per-click advertising, it is like throwing a party. You’ve created a guest list of people who are invited to your party (these are your keywords), but you’ve also hired a bouncer to keep out any undesireables (these are your negative keywords).
Here’s an example. Let’s say I offer French lessons. My keyword list likely has a number of keyword variations that users might type into a search engine, like “French lessons”, “private French lessons”, “French language lessons”, and “French language instruction”.
But what if someone types in “private French Horn lessons”? Or “French language instruction online”? Or “French language lessons on DVD”? Without negative keywords, you will be showing your ads to all these types of bad traffic — inadvertently lowering your quality score and overpaying for your traffic as a result.
Now, you’re probably wondering where you get your negative keywords from? Do you have to pluck them out of thin air by sheer ingenuity? Am I going to ask you to brainstorm negative keywords while sitting in the bathtub, or keep a notebook on your bedside so you can think them up as you’re drifting off to sleep? No, no, no — there’s a much easier way!
Here’s what I want you to do. Just as you probably use a keyword tool to help you develop keyword lists, I’d like you to use a keyword tool to develop negative keyword lists. And if you don’t have fancy tools at your disposal, don’t worry — for our purposes, the free keyword tool in your AdWords account will do just fine!
First, you’re going to type one of your keywords into the keyword tool:
Then, scan the list for “bad traffic” terms:
For each “bad” search query on the list, find the offending word in the phrase and add it to your negative keyword list. (Below is a screenshot of adding a negative keyword in AdWords using AdWords Editor, but this works for whatever search engine or interface you happen to be using.)
See how it works? It’s not hard at all really and is well worth the effort.
Good luck and remember — don’t feed the Google piggybank!
View Comments (16 Responses) | Categories: Paid Search
Posted on March 19, 2008 by Robbin Steif
Wow, did you all see John Marshall’s dreamy new look in the five videos he and Avinash Kaushik created? John, you look awesome with contact lenses. BTW, if people didn’t see the blogpost or the videos, they were all about which referrer gets credit for a conversion when the visitor has looked at multiple sources before converting, and/or cleared cookies — a really great topic.
You may not have seen them for the same reason that I put this off for a full week: I am not a video fan. Sure, I do them when I have to, or when it is hard to understand something. But I am always dying to read – I feel like I can read 40 minutes of video in about 5-6 minutes. The question is always, will I come away with as great of an understanding?
I am probably the next-to-last person in the universe who would rather read than watch, so for that last person, here is a synopsis. I have also embedded the Google Analytics interpretations where that makes sense. Suggested reading time: 5.5 minutes.
Act I, whereby Avinash and John explain who gets credit for the sale with only one referrer. This was an easy situation: A visitor is looking a
the NYTimes online, clicks on a banner ad hosted by Doubleclick, goes to the site and purchases. If the Doubleclick ad is correctly tagged and all the pages on the site are correctly tagged, Doubleclick will get credit for the sale. This is true for all analytics packages, including GA.
Act II, whereby Avinash and John explain who gets credit for the sale with multiple referrers. This was a harder situation. Everything is correctly tagged. The visitor starts with the same NYTimes ad hosted by Doubleclick, he looks at a few pages, but doesn’t purchase. He comes back a second time using a paid search ad, and he looks at a few more pages. However, he still doesn’t purchase; instead, he bookmarks the product of interest. The third time, he uses his bookmark to come back and he purchases. Now, who gets credit for the sale? John and Avinash said, “Your analytics package will usually give credit to the last referrer, direct (none).” This might be true for many analytics packages, but I know it is not true for Google Analytics. In GA, a referrer always overwrites the last referrer, with one exception: direct (none) will never overwrite, or if you like, override, a prior referrer. It is possible in GA to avoid overrides/overwrites, but it is not possible to force GA to give credit to the bookmark. So to summarize: if someone comes on a banner ad, comes back on PPC, and finally converts using a bookmark, GA will not give credit to the bookmark — it will give credit to the last “real” referrer, which in this instance is a search engine’s paid ad. (John Marshall discussed this with me, and he pointed out that it is not clear that the way GA does it is the “right” way. This is arguable, but we weren’t go there now.)
Act III, whereby John and Avinash point out why all those different measuring tools you have show different results. Ok, we understand how your web analytics tool works, now, especially if it is Google Analytics. But what about all those other measurement packages you have? How does your DoubleClick report work, how does your paid search report work? After all, that person started with a DoubleClick ad, came back with a paid search ad, and converted on a bookmark. Well, strangely enough, everyone gives credit to themselves! Doubleclick takes all the credit for the sale, as does the paid search conversion tracking when you look at it using a tool like the conversion tracking that comes with Google AdWords or Yahoo Sponsored search.
So, (Act IV) John said, this is a great reason to use your web analytics to look at Time On Site – even though your analytics doesn’t give credit to all referrers, you can use Time on Site to understand how important those referrers might be.
Hmmm. How would we do that with Google Analytics? The best way would be to create a special profile for each important campaign. Then you’d be able to use the time on site and depth of visit charts (under Visitors > Visitor Loyalty) to get a “bucketed” understanding of how many people referred by DoubleClick stayed for 3 minutes, 4 minutes, etc. Of course, you’ll still want to drill down to the adGroup and keyword level, and there you will only be able to look at averages, but it is better than nothing. And if you have one really important keyword, go ahead! create a profile for it.
Act V: Whereby John and Avinash discuss what happens when cookies get deleted, or privacy tools block cookies, etc. Doubleclick gets deleted, usually. Even your PPC tool’s cookie usually gets deleted. So by the time the sale happens, there really *aren’t* any cookies to take credit for the sale! And this is quite a wonderful revelation in my eyes, because now it explains why so many accounts have a potentially disproportionate amount of traffic in “direct”, even when all campaigns are coded correctly. In this situation – no cookies left to take credit – even Google Analytics will give credit to the sale to direct (none.)
Robbin
View Comments (5 Responses) | Categories: Google Analytics, Web Analytics
Posted on March 12, 2008 by John Henson
When looking at your web analytics reports, it is important to know what you’re looking at. What may initially look like errors in the data, could just be the result of not understanding how the report or the individual metrics are defined. Consultants, who frequently work with multiple analytics packages, must be especially careful. Here’s a short example of what I mean:
A client came to me with 2 reports. One was the product SKU report for a particular SKU. The other report was that same report, but segmeted by source.
In the report that was segmented by source, all of the summary data was considerably higher than on the non-segmented report. The client didn’t know which one was correct — why was the data changing when he segmented the report? (And when you see two different sets of data you think should be the same, you wonder if either is correct.)
Open up your own Google Analytics ecommerce report and follow along.
First go to the product SKU report by choosing Ecommerce -> Product Performance -> Product SKUs from the left sidebar menu in the reporting interface.
You’ll see a list of all SKUs that sold during your selected time period. Click on one to see a summary screen for just that SKU. It will look something like this:
Now use the pulldown menu to segment by source. Most of you will see the summary numbers jump up, like my example here:
In the example images, the Quantity went from 545 to 715, just from segmenting. And Product Revenue jumped from $14,385.33 to $19,585.61.
When Google Analytics displays the segmented report, it is pulling Quantity, Product Revenue, etc from the Transaction Level. That is, the Quantity is now the total number of items purchased in all transactions that included your selected SKU. Likewise Product Revenue is the total revenue for all transactions that included that SKU — not just the revenue generated by that product.
If you go straight to the segmented report, you might not even notice that the data is different, and you could be making decisions based on the wrong information.
It is not that the data presented is wrong, it’s not. But it may not be the data you are expecting, which can be just as bad.
Although my example is in Google Analytics, it’s important to consider regardless of your analytics package. Make sure you know how a report is defined and if you find something that doesn’t seem quite right, be careful of your assumptions and don’t always believe what the report tells you about itself.
-John
View Comments (3 Responses) | Categories: WA Tools, Web Analytics
Posted on March 10, 2008 by Traci
When I begin working on a new pay-per-click account, I really never know what I’m going to see. Sometimes the keywords are far too generic to generate a good conversion rate (a dog walker advertising on the term “dogs”) and sometimes the keywords are far too specific to generate any real traffic (such as “dog walking services in Pittsburgh’s North Hills”). But you know what never surprises me with new clients? Poor segmentation!
So, for anyone who’s just delving into PPC for the first time (or who’s been bravely running their own PPC campaigns), let’s get down to business. What is segmentation? And why is it so important?
I’d like you to think of your PPC account as an investment portfolio. You might look at your portfolio as a whole from time to time to see how you’re doing overall, but in general, there is very little actionable data at the account level.
The actionable data comes from seeing how various segments of your portfolio are performing. Let’s say my fictional investment portfolio is made up of energy stocks, manufacturing stocks and tech stocks. If my energy stocks are getting phenomenal returns, my manufacturing stocks are getting steady positive returns and my tech stocks are losing money each month — well, *that* is some actionable data. I’m probably going to put more money into my energy stocks, hang onto my manufacturing stocks and sell my tech stocks.
Your pay-per-click account can be seen in much the same way. When you divide up your account into 6-10 campaigns (generally, your main product lines or service areas), you are going to begin getting some clear, actionable data. You’re going to know what your top-performing campaigns are and you’ll be able to turn up the volume on these campaigns (for example, increase your bids to get a higher volume of high-converting traffic). Similarly, you will know where your efforts are wasted and can stop the “slow bleed” of poorly performing campaigns.
And don’t forget to extend your segmentation efforts one level deeper — each campaign should be segmented into relevant ad groups (let’s say about 2-4 ad groups for each campaign). Although adjustments at the ad group level are somewhat more “fine tuning” than “big picture”, the principles are the same — it’s still about doing more of what’s working and less of what’s not.
So, if you’re one of those many PPC advertisers who have one campaign and one ad group, start segmenting! What you find out may surprise you.
View Comments (5 Responses) | Categories: Conversion Science
Posted on March 5, 2008 by Robbin Steif
Thanks to everyone who took our survey. We had 72 responses (and we have about 1500 subscribers, plus tens of thousands of unique visitors who read on a non-subscription basis), so there are definitely issues with statistical significance. I will do a separate post on statistical significance of surveys (i.e. where you only have one numerator and one denominator, as opposed to all those calculators that let you compare tests.) In the meantime, many thanks to Judah Phillips for lending a hand in that department.
We only asked five questions:
1. Please describe your “relationship” with the LunaMetrics blog
2. Why do you read this blog?
3. In the area of Google Analytics, you consider yourself (novice, intermediate, expert)
4. In the area of multivariate and A/B testing, you consider yourself (novice, etc.)
5. Freeform place to tell us what you want
At a simple level, we learned that most responders to the survey are also blog subscribers or read very often (63%), they are mostly interested in GA (46%), although the second area of interest was split between learning about conversion in general and learning about web analytics in general. This one was hard, because I didn’t learn enough to stop beating myself up about doing so much GA stuff and so little conversion stuff. The least important reason people read is to learn about industry trends. (OK, we will remember not to do that. An easy request.)
We learned that among responders, there is a nice bell curve of GA expertise, with the majority of responders (45%) considering themselves at an intermediate level
and experts at 18%, newbies at 32%. On the other hand, the vast majority of responders considered themselves newbies at MVT and A/B. Again, this is helpful information, because it reminds me and other bloggers here at Luna to write at a level that new and intermediate analysts can understand.
I tried to do a lot of neat correlations that fell apart (too little data.) However, I pushed the numbers by hand, and was continually awed at how MVT came in last place to most readers. Of course, there is a strong problem of what causes what; after all, we write about GA so much, that it would be surprising if the audience didn’t self-select and want to hear about it. On the other hand, it always feels to me like MVT goes hand in glove with web analytics….
Although everything was completely anonymous, a number of people left some really great comments, and a few even left their name. And responses were so nice! (Maybe no one knew how incredibly anonymous they were – no IP addresses or anything.) Here are only some of the comments that we got. I copied and pasted (and only added hyperlinks to other places in our blog.)
Freeform answers to, “What is your relationship to this blog?”
I´ve discovered today, reading a spanish analytic blog.
I clicked on the link today while I was visiting your website…just wanted to know more about LunaMetrics…in particular, your CEO, Robin.
I just subscribed after the recommendation of Avinash’s book. I use Google Reader, which i read daily.
accidentally found it yesterday through google and it was the only website to make understand RegEx properly 
Yes, I am embellishing on my answer!! My answer is really A: I subscribe to this blog or read it very often, but I also use some posts as reference guides whenever I need them. Amazingly, I still use Robbin’s Regular Expression guides from 2 years ago, because it’s always good to go back and cover the basics again. [Dear writer -- me too!! There are all these RegExen, like negative lookahead, that I don't use that often, and then when I need them, I go back to the blog and learn about them again - Robbin]
Freeform answers to, “Why do you read this blog?”
I don’t care for the way this question is laid out. I don’t think that I can “rank” the reasons why I read a blog. But I do know that this is one of the blogs that you MUST have bookmarked if you’re in web analytics and consider yourself “staying on top of the industry”. Therefore I have checkmarked that answer, but everything else would come in as 1a, if you will.
I wanted to see how much LunaMetrics knows about paid search.
Interesting insights, How to’s on how to get at the data in GA to answer relevant questions.
because Robbin will ask me if I read it and I must be truthful!!! But your team is so freaking smart I want to read it.
Freeform answers in the place where I said, “Go ahead, tell us everything you wish we had asked, what you wish we would write about, anything at all.”
More than anything, I get to see when the web analytics industry moves to analysis of customer behavior over time (i.e. prior loyal customers have a 20.3% chance of visiting in January 2008, visiting an average of 3.4 times, and has a 32.4% chance of purchasing, spending an average of $125.) When I see this type of information, broken down by Recency/Frequency/Monetary/Channel/Personas, I’ll know that the Web Analytics community has arrived. I believe the folks at Luna Metrics will be among the first to get this.
I like stories about what other people have learned, tried or are thinking about. Since I am a novice, I want to learn as much as I can. I don’t really like the technical posts, but I am not the technical person so I just sort of skim them and move on. I’d also like to learn about how your clients are overcoming organizational obstacles. [Ooh, this is a good one, although Avinash just wrote a nice piece on this.]
I really liked it when you brought up the issues with documentation on the Google Analytics Help Section. I’m all about organization and it would be nice if you pushed that issue some more. Since the issue was brought up, I have seen several changes in the Help Files organization. I also really like it when you conduct experiments (like your Visitor Loyalty experiment from about a month ago). Thank you, Robbin!
You might have asked about your audience. For example, website managers might be interested in a certain level of detail on how to do something, but CEOs might be in more top level stuff and may not have time to slog through the blog.
I *love* that you are putting tools on your blog, like the FireFox extension for copying goals, and the macro for deduping parameters. Thanks!!!
I wish more WA bloggers would write about “Doh!” moments they’ve had – where they’ve spent x amount of time (too much) trying to solve a problem, or configure a profile, or segment a group, or whatever, and after spending all this time, they have an “aha” moment, where they realize the answer. Sharing those moments with other practitioners/blog readers…priceless
View Comments (No Responses) | Categories: Surveys
Subscribe to this blog's feed
