 Pay Per Click Prognosticator, Analyst and New Media Enthusiast |
|
||||||||||||||||||||||||
A .pdf file, Research Paper from May 2006 on the buying behavior of Google Adwords Visitors on 3 major internet retailer e-commerce web sites. 
PPC Match Type Comparison and Analysis: What are we really bidding on? A .pdf file, Research Paper from June 2003 Comparing Overture Standard Match vs Google's Broad and Exact Match and the advantages/disadvantages of each. Accompanying Data to Match Type Study A .pdf file, excel doc with PPC Match Types and their corresponding Actual Queried Terms over a one month period in the finance vertical. (Note: Using Google SiteMaps to see the actual queried terms that trigger a paper about matching actual queried terms is both interesting and ironic.) Contributing Author in the book: "Search Marketing Strategies:A Marketer's Guide to Objective Driven Success from Search Engines" on Google Adwords ROI optimization strategies. Click Here to read excerpt in Google Book Search Google announces switch from CPM to CPC Email from Sept 2002. Found this cleaning out my old email. Thought it was interesting. |
|||||||||||||||||||||||||
|
Other ramblings: Complexities of PPC analysis: posted 5.09.07 While keyword research, position and creative testing are fundamental aspects of search, they are just the beginning challenges when one begins to truly understand what is happening "under the hood" of PPC engine tools. What is often, truly happening "under the hood," are the aspects of search ad serving which are not communicated through PPC interfaces, and often not known, or communicated by search engine reps. This is both good and bad. Its bad because it makes performance improvement more challenging. Its good because it means there can be significant potential for improvement, beyond focusing on the search fundamentals. Below, a few of these strategies are discussed. Match Types: Match types make keyword level performance reporting and analysis archaically complicated. Google, and the other major ppc engines, set the default match type to broad match. For this reason, many keywords run for long periods of time on broad match. In an effort to improve efficiency, it often becomes necessary to then submit those same keywords to exact match or phrase match. Many times, however, it would not be wise to simply switch off the broad match version in order to try and maintain the current level performance, while we test. However, by virture of submitting the exact match phrase, the broad matched phrase as defined by Google will often appear as though it has declined in performance or impression/click volume. If one continues to analyze keyword level reports, this can be deceiving. Often times, the exact matched actual queried term actually makes up the majority of the volume coming through the broad matched listing. Thus, when the exact matched version of the keyword is submitted, the broad mached listing often severely drops in volume. Even more importantly, the conversion rate may drop as well, since the exact matched queried terms, which tend to be the most relevant, no longer come through the broad matched listing. Add in "phrase match" to the mix, and the complexity is further exacerbated. This makes it very difficult to accurately compare the exact matched version against the true broad matched version of the keyword, as they need to run at the same time for a true test. Reporting date ranges: As a search engine marketer, or anyone who runs runs a ppc account can testify, we often log-in and check on things. We'll run a random report for the previous week, to see how we did, or see how much we spent. Maybe we submitted a couple of keywords last week and wanted to see how they did. Possibly a client may ask us for a Year to Date report or a monthly report, so we'll run a keyword level report for this time period. The difficulty is that we frequently make changes to different parts of an account or campaign, however, we run date ranges that report the data for all KPIs which are all tied to the same date range. The danger is when we make changes during this time period that pertain to some KPIs and not others. If I have one adgroup running from Jan 1 to March 31, and on February the 15, I submitted several keywords to this adgroup that drive significant conversions, a report on this adgroup from Jan 1 to March 31 will be totally misleading, as the conversion rate will be calculated by a number of conversions that occured from Feb 15 to March 31 is divided by the impression number from Jan 1 - March 31. This is partially a flaw of reporting simply on an adgroup level. Another example shows the necessity of carefully choosing date ranges to run reports. Toward the end of April, I changed a number of creatives to include DKI. This drastically effected the performance of a number of keywords which did not previously drive significant traffic. A colleague ran a report from Jan 1 to April 30 and one specific keyword shows that it spent $500 over this 90 period and only drove 1 conversion. All other keywords in this campaign drove conversions with reasonable efficency except this keyword. Unfortunatley, for this keyword, it *appeared* to have been performing poorly over a 3 month period, yet was simply the result of a creative change in the prior week. "Tail Terms:" This brings me to the next complex question of what to do with keywords with one conversion. For years, it made sense to run weekly reports for clients, and then analyze them. As per usual, branded keywords, and other strong keywords, would have mutliple conversions in the past week, and then there is a group of keywords that have one or two conversions. Some would have strong efficiency and some would have weak efficiency, so we would change bids accordingly. The next week, we run the same report, some keywords that had one conversion in the previous week, would have none this week, and some keywords that had 0 conversions the prior week would have 1 or 2 this past week. If we try to form bid management strategies on these data sets, we're essentially playing an endless game of keyword whack-a-mole. In this case, CPC's will rarely have an effect on CPA/ROI. The difficulty is that if we simply run a report for a longer time period, the keywords that were "tail terms" in a weekly report, become high volume performers in a monthly or yearly report and a diferent group of keywords would appear as "tail terms." Every keyword has its own level of popularity and volume, and therefore, requries its own amount of time in order to accumulate enough statistically significant data. Regardless of the need to feel as though one needs to "do" something to show that we are busy, it is largely futile to analyze or make changes to anything that has not accumulated enough data to make a confident, educated decision. This begs the question, "how much data is statistically signficant?" which is another discussion. A major challenge for search engine marketers, is that many business's keywords do not accumulate statistically significant data fast enough, that we can make informed decisions and produce positive change within the "30 day" trail period, 6 month contract or simply, "soon enough to prove ourselves before the client gets ancy." "Daily Budget Caps:" Many businesses have campaigns with no budget caps, due to unlimited budgets or relatively few keywords. When daily budget caps are in effect, another layer of complexity is added to the already complex situations above. We all know by now, if budget caps are in effect and low, ads will not be shown all of the time. Google will do its best to space them out or, or we have the option of accelerated delivery. The diffulty here is that a compaign with one daily cap can have mulitple adgroups and/or keywords. The adgroups or keywords that have the most volume, tend to consume a larger portion of the budget. Or if the cap is low enough and the CPC's high enough, which keywords happen to get clicked first, consume all of the budget. When we run our weekly reports, keywords that have no impressions or clicks, *seem* to be poor performers, but this is merely a function of the daily budget cap. Adwords has recently begun to show "alerts" that show what percentage of impressions we may be missing out on, on a campaign level. The difficulty is that keyword conversion performance can be so disparate, we really only want some keywords to get full exposure and not others. I wonder how long until we are able to put budget caps on specific keywords. I used to think that Google would never do this, but in my experience, by separating keywords that simply "seemed" like they would do well with full exposure in their own campaigns, I've often been able to increase performance and spends by 4 to 8x in a matter of weeks. Everyone was better off. I've seen this work on campaigns for 6 of the top 50 etail sites, 1 top 10. Clients made more, Google made more and we performed. I predicted 4 years ago that Google would eventually indicate how much impression "opportunity" an account would have, I was just hoping it would be at the keyword level. If Yahoo or MSN wanted to do something ahead of the curve for a change, they would go ahead and do this. Each PPC engine even has its own twist on this, depending upon their ability to deliver against a budget cap. While I have a degree of respect for "Wall Street" based keyword portfolio algorithms, they simply do not address a number of these issues. They tackle, or should tackle, questions like statsitcal validity and reporting, however, they simply do not, and can not deal with the underlying issues that are not communicated through APIs. Many of the factors listed above, have made or broken search campaigns. Not only do these not surface through APIs, they are not apparent anywhere in the interfaces and often times, not realized by search engine account reps. The complexities of advanced ppc analysis are clear. For the search engine marketer, this presents another challenge, which is one of communicating these complexities. Once one is able to read "through" the reporting tools..to see what is "actually" going on, the more difficult it has become to explain to clients why some decisions are made. For many who are new to ppc, the mental building blocks needed to understand the complexities of this level of analysis, are themselves complicated and take a while to soak in. Keyword Relevancy Factor |
|||||||||||||||||||||||||
|
posted on 4.20.2005(but never really finished) by Andrew Holladay There is a direct relationship between the keyword relevancy factor and the potential conversion rate/revenue a keyword can drive. Keywords can be rated according to their relevancy to a web site. We'll use a scale of 1 - 10. 10 being the most likely revenue generating terms, and 1 the least likely. The fundamental question the semarkerter has to ask about a keyword is: "When a user/potential customer types this keyword into a search box, how likely is it that the preconceived idea/concept/image in their head at the time of typing this in, matches the current content and offering of my website? and to what degree?" 10 - Well known Branded terms, "Acme Inc" and they get www.acmeinc.com. The more well known a brand name is, the higher the percentage of overall revenue it is likely to drive.(this is assuming campaigns and bids are set at maximum keyword exposure* across all keywords). 9 - Unique branded terms with fairly good brand awareness and clearly could not be confused with any other business. 8 - Unfortunate branded terms, brand names which also happen to be another common word or is very similar to another brand name. 7 - The brand names of specific prodcuts that are owned by one or only a few businesses that can only be found on one or only a few sites. 6 - Branded keywords of medium to smaller business, which may not be as well known. These won't perform as well as other branded terms due simply to lower search volume. 5 - Specific Keywords(keyphrases with many keywords and/or model numbers) which represent the majority of the content of a site. 4 - Specific keywords of which the site has a large selection, but is not the primary, or major, focus of the site. 3 - General keywords(typically one word keywords, such as "camera") which might represent a good portion of a site and its focus. 2 - General keywords of which a site has a small selection or percentage of content related to it. 1 - irrelevant keywords(probably won't be approved, or will be disabled anyway.) - not only do branded keywords tend to drive the most revenue, they tend to have high ROI due to lower CPC's for these keywords. This is due to SE policies restricting the guidelines on competitors of these keywords, and due to controls many companies have over the amount their affiliates are allowed to bid. Obviously, many factors determine whether a keyword converts. To view it chronologically according to the path of the search user, it would go something like this. These are also rated on a scale of 1 - 5, according to their level of influence in the conversion process. 1 - not as influential, 5 - most influential. 1. Keyword Research - which keywords do you show up on. (5) 2. Keyword Exposure - how often and where do you show up (5) 3. Creative/CTR - how do you attact/pre-qualify etc..(1) 4. Landing Page - do they land on the right page, does the page create momentum(4) 5. Shopping Cart process/site usability, how well does the site convert etc....(3) The above description of the keyword relevancy factor should be considered by whoever in the organization is responsible for step #1. In order to drive as much revenue as possible, as soon as possible, one should, first and foremost, micromanage the keywords that fall into the keywords with the highest keyword relevancy factor. This means putting them in their own campaigns, choosing the right matching options and CPCs and telling your affiliates to back off and not bid higher than you. It is likely that these will drive the majority of revenue. If you need to drive more revenue or conversions as soon as possible, micromanage current keywords. New keyword generation can take days for internal and external reasons. If you have monthly goals, 3 - 5 days is 10 - 17% of a month. *Maxmimum keyword exposure refers to both ad serving frequency and position together. Now that both Overture and Google both have daily budget caps, which force the spacing of ads, its important to read the reports with a grain of salt. In reports, When a keyword has an average position of 1 or 1.1, as the case may be, this is fairly deceiving as this is only the average position of your ad, WHEN IT WAS SHOWN. If you showed up position 1, in one out of five searches, who cares. Your competitors showed up Pos 1, four out of those five searches. Therefore a report, which shows an average position of 1.1 can be deceiving as the total keyword exposure could have been minimal." |
|||||||||||||||||||||||||
|
Andrew Holladay Zoominfo
ppc analyst |
|||||||||||||||||||||||||
