Efficient Frontier Insights

Making the most of Yahoo!'s Partner Network Bidding Features

Recently, Yahoo! introduced some features that make it easier to manage the performance of their large Partner Network. If you've ever pulled an Ad Distribution Report to look at your performance on their Partner Network, you'll know what a welcome change this is. Jonathan Beeston recently covered this topic in an excellent post. Today's post looks at how the Partner Network performs across a larger sample of our client base and discusses some tactics for using these new features.

Partner Performance Overview

The chart below looks at the Partner Network performance for a sample of several retail clients totaling about $4MM in search spend.



The chart shows the percentage of clicks and revenue coming from the Google and Yahoo!’s respective partner networks. The dots show the RPC for each Partner channel relative to the overall SEM RPC. Note that these averages are not weighted by client spend and they represent an equally weighted sample of different businesses’ experiences.

A couple observations that are as expected:

• A greater proportion of Yahoo!'s click traffic comes from the Partner Network – about 40% vs.  30% for Google.
• Yahoo! Partner traffic does not convert as well as Google's. Google Partners have RPCs that are almost double those of Yahoo!.

Though the profile above is typical, we have seen large variance across accounts. In some cases we've actually seen better performance from both the Yahoo !and Google Partner Networks. It really depends on who Yahoo! has enlisted in your vertical. Also, Yahoo! continues to aggressively police the quality of their Partner Network traffic, so you can expect to see continued changes in the makeup, and thus the overall performance.
Looking at the same industry sample used above, the charts below show the wide variation in performance across businesses.
 




In the context of Yahoo!'s new partner bidding features, the performance disparity between Yahoo! Search and Yahoo! Partners is great news for advertisers. Without separate targeting, advertisers have been forced into a one-size-fits-all bid. Bids have been based on the blended performance of these two very different channels. Now, with separate bids you can spend more on the high converting search engine traffic and less on the low-converting Partner traffic.
 
Managing Partner Traffic

There are actually several tools provided by Yahoo! for managing you this traffic, so the implementation possibilities deserve some discussion. First some historical context, separate targeting for the Partner Network is the latest in a series of advertiser-friendly new features from Yahoo! designed to improve the performance of their search network. For a long time, they have provided the ability to use domain exclusion to block individual partners. Then, during the last year, Yahoo! has started actively trying to improve the quality of their Partner Network by excluding the most egregiously poor performers. In addition to this, they instituted Dynamic Pricing to automatically discount the CPCs you pay based on partner quality. (Quality being click-based so not necessarily aligned with conversion rate).

Now Yahoo! has provided two new tools for managing partner traffic.

First, users can apply a blanket discount on bids for the partner network. So if you chose a 20% discount to Partner Network bids, a keyword bid of $.50 would be $.40 on the Partner Network. Note that the minimum bid constraints still apply. If your discount puts you below the minimum bid, your bid will be inactive on the Partner Network.

Second, advertisers can target the Partner Network separately. You can mirror campaigns to have one targeting Search and one targeting the Partner Network.

Here are our guidelines for implementing these approaches:

• Proceed with caution! As outlined above there is significant upside but there is also risk. These are new features whose actual behavior and performance have not been validated. Start with a small test or, if you’re risk averse, wait a couple months to see what happens to others.
• If you’re not already doing so, look at your Ad Delivery reports to see if there are partners that are consistently delivering very poor traffic. Consider blocking partners with particularly bad traffic.
• For high volume campaigns consider mirroring campaigns and having them target Search and Partner Network separately. This is where the real opportunity lies. If you bid smartly, you should be able to unlock significant efficiencies here.
• For low volume campaigns, use bid discounting. Mirroring campaigns adds to administrative overhead and also slices your data more thinly. If either of these are a concern then at least experiment with discounts (or premiums) for partner bids. Again this is uncharted territory so proceed with caution.
• Whatever you do keep a close eye on your performance. Yahoo! is aggressively managing their partner network quality. The volume and quality of traffic coming from the partner network can change quickly. Also, as your competitors take advantage of these features their bids will change. Specifically expect Yahoo! Search traffic to get more expensive. Be vigilant and be ready to adjust.

At Efficient Frontier, we are excited about the changes Yahoo! has made. Combined with domain exclusion and the relative size of the Yahoo! Partner Network this represents a major opportunity for increased efficiency and growth. These are exactly the kind of tools that smart, data-centric, and scalable SEMs can use to their advantage on behalf of their clients.

Shay O'Reilly

The Energy Was Up at SMX West 2010

The market reacted well today to encouraging signs for the U.S. economy. Retailers said that store sales in February were up by the largest amount since November 2007. And US factory orders rose by the largest amount in 4 months. Well, let me add some more positive news to the mix: using SMX West as a sample, the energy level at search conferences is up 2X year over year (for you statisticians, I made that up).
 
Danny Sullivan and team put together a strong event in SMX West 2010. Kicking it off with Steve Ballmer was a coup for SMX and his energy and enthusiasm for search carried throughout the event. And the level of interest in the sessions and enthusiasm for search (and display) marketing was evident among the attendees I spoke to over the last few days.
 
I participated in two very different panels. On Wednesday, I spoke about the opportunities in integrating search and display marketing. A clear take away from the search and display session was that marketers are hyper-focused on making progress against the challenges of attributing success correctly across marketing channels. We’re eager to partner with our clients and prospects in tackling the complex problem of attribution!
 
Today, I spoke about the Microsoft-Yahoo deal. I received many questions (asked in different ways) about the future relevance of Yahoo! without a core search offering. (Note to Yahoo!, it doesn’t seem like your messaging around “controlling the user experience in search” is breaking through with advertisers.) Clearly, Yahoo! has some amazing properties that most of us use every week if not every day (ex. Finance, Mail, News, Sports). They will remain very relevant for the foreseeable future, just perhaps not as a search engine.  

It was fun to engage with peers on the coming trends in our industry. The cloud of “cautious retrenching” that very appropriately dominated events in 2009 may have lifted. With the search landscape evolving, display shifting to auction pricing, and mobile coming of age, we have huge opportunities on the table over the coming months and years in digital marketing. 

Justin Merickel
Vice President, Marketing and New Product Development

Monthly CPCs for February: Mixed news

The CPCs from February  2010 are presented in the table above. From a CPC perspective, there are some interesting trends. While automotive CPCs experience a seasonal decline, the decline in finance CPCs dropped significantly. This is not surprising if one notes the performance of the S&P 500 in February. The overall index fell 9-10%. We have noted in the past that the financial vertical spend and stock market performance is very tightly correlated.
 

On the flip side, we saw a double digit rise in retail and travel CPCs  This is noteworthy because travel and retail CPCs typically fall in February (see 09 data) due to seasonality.  In the case of travel, could this be the first sign of an uptick after a year and a half in the doldrums? We shall see what March brings us.

The Fat Tail

We all know the importance of the long tail. Below is a quick graphic emphasizing the importance of it and how to best analyze your data.

1. Look at all of your keywords their conversion over the last year.
2. Bucket them by the number of conversions they had in the last year. If you've got a head term with a large number of conversions, for example 5/day, that's 5*365=1,825 for the year.
3. Multiply the number of keywords getting that many conversions.
4. Graph it out.

The result is a chart like the one below showing the total volume of conversions coming from the head to the tail.

This chart is based on actual client data and displays a typical head-tail distribution. Collectively, the most insignificant keywords, the ones that have only one conversion all year, account for as many conversions as the largest head term.

Head terms and Brand terms are always going to be a major part of your spend and revenue but it's these low volume terms that are crucial to account growth and profitability. Therefore, they should be an important part of your account strategy. There are two components to a healthy fat tail: coverage and efficiency. You need all possible converting searches included in your account. A retailer, for example, should endeavor to have their entire product catalog included in their account. Second, you need to tailor your bids efficiently. This is an art and a science unto itself but at the very least you should be evaluating and bidding on each term regularly - ideally daily. Setting a bunch of flat $.10 or $.20 bids is just leaving money on the table. The fat tail highlights the importance of managing your long tail keywords effectively. If managed well, not only is the long tail economical but is also a big volume driver. 

Shay O'Reilly
Business Analyst

Strike The Right Match Type: Part 2

In Part 1 of this blog, we spoke about common mistakes made by PPC marketers when using match types in their search campaigns.   It is easy to highlight areas where campaigns are falling down, but we wouldn’t be doing our job if we didn’t highlight some practical ways to redress the match type balance.

The example below shows some of the keywords which Google deems to be related to the term iphone.  Would an advertiser selling iPhones want to bid on steve jobs or should they, in fact, exclude this term altogether?

 
The Google AdWords Keyword Tool and the Wonder Wheel give a great insight into how Google’s matching algorithms work.  Insights aside, for practical implementation purposes, there is no substitute for a search query report (SQR).  

Here are practical steps you can take to ensure that you’re making best use of match types in your campaigns:

1. Run SQRs on a regular basis (weekly for high volume generics)

• This is the most valuable insight you have into your audience’s interaction with your search campaigns
• SQRs will help you to build highly targeted campaigns which give you every chance of maximizing your potential to convert
• If there is sufficient volume available for a search query which has been matched to your keyword, then consider adding it into your account on all three match types
• Exclude search terms which are irrelevant to your account/a particular campaign

2. Don’t be tempted to bloat your account by triple matching all terms

• As a rule of thumb, consider how many words make up the keyword.  If there are more than three, there is no need to add exact/phrase match versions to begin with.  
• Use your common sense to inform what does/doesn’t need to be triple matched; cheap flights is a high volume generic which needs to be on all match types, but find a three star hotel in Paris probably only needs to go on broad match in the beginning

3. Don’t assume that Google will always match the search query to the exact match keyword, just because you have that version of a keyword in your account

• SQRs will give you an insight into where the ad rank of your exact match terms is being trumped by that of their phrase or broad match counterparts
• You may have to add negatives just to ensure that the traffic is channeled in the way you want it to be

4. Ensure that your ad groups are tightly themed

• Also ensure the keywords are clustered according to the message you want to serve to the user.  This will mean that, when your exact match terms are triggered, the ad returned speaks to the user’s intent.

As ever, we’re interested in hearing your thoughts on this and all other performance marketing topics!

Charli Rogers, UK Client Services Director

Strike The Right Match Type: Part 1

Search marketing is all about targeting your audience with the right message at the right time, so that the campaign’s potential to convert is maximised.  Effective use of match types is critical to ensuring that you have maximum control over your search campaigns. 

This is easier said than done; three common mistakes made are:

1.    Overuse of broad match keywords
2.    The old triple-match trick
3.    Poor or no use of negative matches

Let’s look at these one at a time.

1. Overuse of broad match keywords   

Overuse of broad match is very common in PPC.  It’s very easy to allow Google’s matching algorithms to take control of your targeting.  Unless you are constantly reviewing your keyword lists, you’re not going to be controlling who sees what message and when.  This can equate to inflated CPCs, poorer CTR and lower conversion rates, all of which stand to make your paid search activity less efficient. 

Too often, we see disproportionate spend vs. revenue vs. ROI for different match types.  The example below shows a huge proportion of spend going through broad match keywords, generating a significantly smaller proportion of revenue:

Put simply, this advertiser is allowing Google to dictate which message/product/offer gets shown to the user (let’s not forget that this is also a potential customer!).  Would that advertiser allow someone to walk into their high street store and hang a sign saying “Women’s Clothing” over the menswear section?

2. The old triple-match trick  

It is a common misconception that triple-matching (having every single keyword on broad, phrase and exact match) leads to an optimal campaign.  This is not the case.  

Managing PPC at scale can often lead agencies and advertisers to take short cuts when building out keyword lists.  Concatenating a comprehensive list of seed terms and modifiers and then triple-matching the output might generate an impressive volume of keywords, but that volume is likely to be the very reason for a campaign becoming unmanageable.  Size, in this case, is definitely not everything.

For high and mid-volume terms it makes sense to invest time up front in utilizing the full range of match types.  For tail terms with low traffic volumes, triple-matching prior to seeing the behavior of those terms is likely to do more harm than good.  It will lead to account ‘bloating’ and make analysis and campaign optimization far more problematic than is necessary.

3.    Poor or no use of negative matches

Negative match types are underused by PPC campaign managers.   We spend so long trying to capture the right traffic by using qualified ad copy and leveraging positive matching, so it seems crazy that this vital step in the process is overlooked.  Time and again, search query reports show that advertisers are paying for traffic which is not relevant to their business, or a particular product.  

Stop by later this week for part 2 of this blog post, where we’ll outline some practical steps which can be taken to ensure you’re taking best advantage of the controls that match types can give. Alternatively I will be presenting this very topic at SES London tomorrow, so please stop by. 

Charli Rogers, UK Client Services Director

CPC - January 2010

On a monthly basis, we will now be publishing our monthly CPCs by category along with some trend analysis. For January we have the following numbers:

The above table shows the CPCs that we saw among our advertisers in 4 verticals. The trends in CPC broadly reflect the seasonality in these marketplaces. Finance dropped slightly as automotive CPCs dropped 10%. A big CPC drop was seen in retail due to the after effects of Christmas. Travel bucked the seasonality trend. Traditionally, travel picks up strongly in January and we see an increase in CPC. However, this year we are seeing a CPC drop indicative of the continuing weakness of this sector.

Search & Display: Understanding Cross Channel Influence on Conversions

At Efficient Frontier, we recently updated our platform to include our first cross channel reporting capabilities inclusive of both search and display. The report gives marketers a quick view of all conversions on their website and insights into what advertising channels are driving those conversions. Additionally, our cross channel report shows how many of the attributed conversions were preceded by events from other paid channels.



Many advertisers continue to have display and search teams working separately and often use different tracking and optimization systems.  As a result, conversions with overlapping events (ex. a search click preceded by a display click) are typically double counted. That is, each channel (search and display) will independently take credit for such conversions. Lacking the visibility into how each advertising channel is influencing the other, advertisers over-estimate advertising efficacy and may over-pay for placements and terms. Let’s examine a site retargeting scenario to illustrate the challenge.

When you are retargeting visitors to the advertiser’s site you are inevitably reaching visitors that came to the site through paid search clicks. This means users who have come through paid search clicks and then viewed a retargeted display ad and then converted will be double counted if the advertiser is not using the same tracking systems across these channels or not using the right attribution model.

We typically see 30-40% of all conversions arriving from a display impression, or click, are preceded with a paid search click. This overlap is significant enough that marketers need to pay close attention to attribution. However, double counting conversions is just one problem. The other issue is that advertisers are not getting the proper visibility into how one advertising channel is influencing the other.

In our campaigns, we have seen that up to 10-15% of conversions that had a prior display impression had the last event as a paid search click. And when we looked further into the path to conversion data, it is interesting to see that many of these users see multiple banner ads and, within minutes of seeing the last banner ad, visited their favorite search engine and searched and clicked on a search ad prior to conversion. Clearly display influenced the search, although that ad will rarely be given credit. Looking at the issue in reverse,  we see that 20-30% of conversions that had a prior display impression or click had a paid search click prior to the display event. In essence, this 20-30% overlap consists of users that came to the advertiser’s site through a paid search click and then saw a re-targeted banner ad prior to converting.

What this data shows is that there is significant cross channel influence between search and display. The key for advertisers is to have both a centralized tracking system to ensure they can obtain cross channel insights and an optimization platform that can activate this data in display and search buying. Only with centralized tracking can different conversion attribution models be tested, making cross channel optimization possible.

The cross channel report is a new addition to our optimization platform. The vision for Efficient Frontier is to optimize media spend not just based on the direct conversions but also account for the indirect conversions. We deliver strong results by providing the right data, partnering on the right attribution model and optimizing across search and display campaigns.

Suman Basetty
Director, Product Management

The Importance of Click to Conversion Time

Click to Conversion is an interesting topic and often overlooked metric. My latest Analyze This column on Search Engine Land discusses how it can help you understand your clients and also help you better manage your campaigns. Take a read of the first article in a two part series on this topic. In this column, I have provided insights on:

1. The connection between the average order size and conversion time.
2. Conversion time and cookie windows
3. How conversion time can be used for display retargeting campaigns.
4. Using conversion time to understand your shoppers.

Dr. Siddharth Shah

More Reasons Why Biddable Display is Similar to Search

As Efficient Frontier continues to dig deeper into the world of biddable display advertising, we're finding a lot of similarities to search. I have just posted an article on this subject on AdExchanger. Feel free to read and tell us your thoughts.

Justin Merickel