This Week in Relevance

Its do-able with strong tools and strategy, but you have to be incredibly careful when creating campaigns of this magnitude from a set of product data, as there are a number of mistakes that can be made that can irreparably harm the quality score for your AdWords account.

Some of the major challenges that you’ll face:

1. Ensuring that your data for building the campaign is of adequate quality to generate good keywords and ads. Product titles are often messy and too specific, leading to keywords that don’t get very much traffic and ads that have to use non-specific default text because of character restrictions. Product categories make very poor keywords because they tend to be very general so running a product specific ad against these terms will affect Quality Score in a negative way
2. Massive duplication of keywords — product data is at the individual SKU level and most retailers will have many SKUs that match a specific query, even a particularly specific one like ‘size 6 Jimmy Choo boots’ — if I have 10 different pairs, each in a number of different colors, widths, etc, would all pairs get that keyword? If not, how do I decide?
3. Account scalability issues — in the tools mentioned in above responses, 100k products corresponds to 100k ad groups which starts to push account size limits in the major engines and could slow down any changes to these accounts via API or AdWords Editor. Without a good strategy for managing the size of these product driven campaigns it is very easy to run afoul of Google and make it more challenging to do other tasks in your accounts.
4. Managing frequent changes to the product data — while it is resource intensive to build this campaign once without proper tools, it is prohibitive to actively manage changes to this data without them. Inaccuracies in pricing and product availability in paid search ads can impact relationships with your customers if they find those ads to be misleading — additionally you can increase your bounce rate which plays heavily into your quality score. To do this right, you need to ensure that you have the ability to update keywords, ads, and landing pages as your product data changes.

All that said, having your search campaigns effectively merchandise your entire product catalog is a winning strategy in search. You just need to be sure that you understand the potential pitfalls, and are leveraging scalable tools to handle the complexities mentioned above (as well as some others…).

Our company, DataPop, deals with these challenges every day via our technology and I’d happy to discuss in more detail if interested.

Dave
http://www.datapop.com

How do you set-up and scale Adwords Campaigns for 10000 to 100000 products?

In the time since our last post we have chosen to focus on 2 specific factors said to influence adCenter QBR. As mentioned previously, with adCenter representing a more meaningful percentage of search share, it is becoming increasingly important to focus on adCenter optimization rather than assume a well optimized AdWords campaign will perform efficiently. Our aim is to design a test that will provide data supported learnings about how much CTR and landing page quality influence “quality” on adCenter relative to AdWords. We have detailed the questions and proposed test methodology here:

1. How does (position normalized) CTR affect the CPC that you pay in adCenter?

It is well documented that CTR is a primary driver of AdWords Quality Score. As a result, advertiser optimization efforts are often targeted at improving CTR to, at least in part, improve QS. It is unclear at this point how much CTR impacts adCenter CPC relative to AdWords:

Testing Methodology:

Isolate a single keyword with significant volume (and a liquid marketplace) in an ad group with multiple creatives and run so each ad reaches statistically relevant volume. If you target a specific position for this keyword through your bid system, then we can use the CTR and CPC deltas to understand how fluctuations in CTR impact CPC for a given position.

An example: let us suppose a test produces the following statistics for our 2 test creatives – how do we interpret the results?

In this case it appears a 100% increase in CTR corresponds to 20% discount in CPC. It is probably not safe to assume similar changes in CTR will impact CPC equally in each KW marketplace. Replicating the experiment several marketplaces, at different bid levels should provide a more educated perspective on CTRs influence on CPC.

2. How much influence does quality of landing page have on CPC (and position)?

As mentioned previously, there has been a fair indication that adCenter is keenly interested in the quality of the landing page experience when determining quality. While it is impossible to know what specific landing page factors adCenter takes into account – and how granular their assessment of a page is – a test structured around 2 distinctly different, yet relevant, pages will hopefully provide some insight.

Test Methodology:

Isolate an ad group with two ads that consist of the same Title, Description, and Display URL but with differing landing page URLs. Equal ad rotation and an ad group level bid should result in similar CTR for each creative. Understanding variations in CPC and avg position should be attributable to QBR due to landing page variation. …

Again, let’s suppose a test produces the following data:

In this case, with the CTR being so similar, it appears the difference in lading page URL is resulting in 10% increase in CPC. At a lower CPC, Creative A should dramatically impact the advertiser’s bottom line over time. Running this experiment multiple times should give a representative set of data on the impact of landing page on CPC in an advertiser’s key marketplaces.

There is obviously a wide spectrum of different landing page experiences for possible use in such a test. One of the more frequent landing page questions we face is to show a category (or sub-category) level page vs. a product level page against Product KWs with some ambiguity. For example, 3M has multiple lines of duct tape – do they drop users on a specific product page – taking a best guess at what a user would prefer – or do they drop users on a more generic duct tapes page? The best answer is whatever landing page provides the best performance against an advertiser’s target metrics, but this test should help us to understand which URL adCenter feels is the most relevant which may provide some guidance on lowering CPCs which should help drive ROI.

The goal of these tests is to move from an AdWords focused knowledge of campaign “quality” towards data supported evidence about the nuance of managing quality in adCenter. Answering these 2 questions well should go a long way towards building an adCenter specific knowledge base and optimization best practices. Marketers can then start to make clear decisions on how to spend their limited resources optimizing their adCenter campaigns. We will be conducting similar experiments over the next few weeks and hope to have a more definitive point of view in a future post on how both ad relevance (CTR) and landing page relevance can drive cost savings in adCenter….but for now, we hope this framework helps in your efforts to demystify adCenter QBR.

We’re back. We have heard from our readers that they would like to see more original content and we are obliging. Moving forward, TWIR will focus primarily on what we know best – Search – and aim to push the Search Marketing conversation forward by examining big questions in the marketplace. As always, thanks for reading.

Understanding adCenter QBR (Quality Score)
With the recent Microsoft – Yahoo! Search alliance, adCenter is now monetizing approximately 20-25% of search share – up from less than 10% pre-merger. While Google AdWords still has a dominant search share, adCenter now presents meaningful source of volume. Understanding and optimizing adCenter performance now presents an opportunity to substantially impact your overall search ROI.

The importance of Quality Score in determining AdWords rank and CPC is well known but for all the blog posts and message board discussion, the actual calculation remains a black box. That said, Google does outline the core components of the QS formula, and, more importantly, exposes QS to advertisers thereby allowing them to measure the effectiveness of their optimization efforts in the context of the AdWords marketplace.

While Microsoft acknowledges that relevance is a factor in determining your ad position, it does not expose a Quality Score or other measure of relevance to advertisers. Anecdotally, we know (or think we know) Quality-Based Ranking (QBR), as it is known in adCenter, is designed to assess the quality and relevance of the ad and landing page to the query. We have heard in conversations that adCenter pays particular mind to the post-click landing page experience in addition to assessing the following:

• Bidded keywords
• Ad copy: Ad titles, text, and display URLs
• User experience

Generally, these seem to be very similar to the principals that determine AdWords QS but we have seen time and again that an equivalent campaign will perform dramatically differently on AdWords vs adCenter.

In Practice
We have observed some interesting trends when comparing our clients’ adCenter vs AdWords performance on campaigns with very similar (if not exact) KW-Ad-Landing Page combinations and bidding strategies. Specifically we have seen similar first page positions and lower CPC s on adCenter even though CTR is (in some cases dramatically) lower than on AdWords. Certainly this could be a function of the dynamics of the marketplaces (AdWords has a much wider advertiser base), search network distribution or both.

It can be incredibly challenging to design a controlled test between adCenter & AdWords to understand how specific optimization techniques impact performance differently. However, structuring an experiment to understand how the respective platforms react to equivalent campaigns and post-launch optimization efforts should provide some interesting directional information on how campaign attributes are judged in the context of the marketplaces.

Check back in the coming week/s as we explore methodologies and structures for experiments to better understand the nuances of managing adCenter campaigns for relevance.

Thanks!
The TWIR Team.

We have a winner in our contest and award our ticket to Under the Radar to Jonathan Rossi for his concise but insightful answer below:

What makes an ad relevant is when it connects and engages with the intended audience. This begs the question: so what actually is meant by *connects* and *engages*?

When communicating a message, be that in product marketing or even politics, one connects with the intended audience when there is a genuine understanding of what your customers (or voters) care about. This will require, of course, learning about your customers so that the ad, amid all the noise and information saturation, conveys a message that demonstrates that you (as the communicator) understand your audience. When there is a connection, there is engagement. This, IMHO, will result in an ad that is relevant to the intended, targeted demographic base.

As we’ve been speaking about on this blog for months, what resonates is what sells, and without an understanding of the intended audience (sometimes and audience of one, like in search) a marketer can’t begin to craft an engaging and relevant message. Congrats, Jonathan and we’ll see you at the conference!

The TWIR Team

DataPop is participating, along with 24 other leading startups at DealMaker Media’s Under the Radar Conference. This year’s conference will look at how retailers, brands and media companies are looking to improve their customer relationship and engagement levels. Improving their decision process, making it easier for them to get what they want, at the best price, in real-time and importantly, closest to them.

We are giving away one free ticket to this conference for the best answer to a question we posted on Quora that is near and dear to our hearts. We will close the contest on November 4th — happy answering!

The TWIR Team

This week Twitter announced a concerted effort to generate revenue through advertising. We think that it is about time – Twitter has the opportunity to create an incredibly powerful ad ecosystem around its platform due to the fact that billions of times a day, its users are expressing intent.

I think everyone would agree that the purest way for users to express intent, and especially commercial intent, is through search. Ad dollars follow intent and as such, nearly 50 cents on every dollar of online ad spend goes into search. However, we also know that consumers express intent in many other ways beyond search, and if marketers can figure out how to harvest and target it, then they will be willing to spend major dollars.
Twitter is a great use case for this. Below are some of the myriad ways in which Twitter users express an ‘intent’ that can be quantified and acted upon by marketers:

1) Users follow brands that they like in hopes that they will get access to timely promotional offerings
2) Users follow all types of personalities (entertainment, business, sports, or otherwise) in order to get a lens into their daily lives, interests, and potentially purchases
3) Users are curious about what the broader community is talking about now and thus engage heavily with trending topics
4) Users search for personalities, brands, and topics that are of interest to them
5) Users retweet, or share content, indicating how much that content resonates with them
6) Users indicate their location and ask for recommendations on what to do, where to eat, where to shop, or who to see

While many of these are not absolutely explicit statements of desire for a particular item, they do serve to indicate pretty specific information about who a user is, where they are, and what they are interested in. At the massive scale of Twitter, 160mm users and counting, this becomes pretty compelling for marketers. As evidenced by the recent announcements, Twitter is starting to tap into some of these ‘intent’ mines with its initial ad offerings.

By allowing marketers access to the trends section of the site, Promoted Trends taps into the Twitter user’s desire to see what’s happening and current in their world. If used effectively, marketers can leverage Promoted Trends, to highlight what is happening in their world’s in a way that is not entirely inauthentic (e.g. product launches, album releases, movie releases, etc.). This has clearly turned into something of a big business for Twitter already, but this is really low hanging fruit, much like Yahoo’s fanatical focus on selling its homepage placement over the years.

A somewhat more exciting ad vehicle recently launched is Promoted Tweets, which allows marketers to place a 140 character ad in front of users who are using Twitter’s search engine. It is clear that an AdWords-type marketplace complete with its own quality score algorithm is soon to develop as marketers vie for the attention of Twitter’s searchers (there are currently over 1b monthly searches). However, the reasons people search on Twitter and Google tend to be distinct and advertisers are going to have to adopt new strategies such that their ads are relevant in this marketplace. While there are some Twitter-centric variations to how advertisers play in this marketplace, this still seems like something of an AdWords knock off and not something that is going to make Twitter a truly unique promotional engine for marketers.

Twitter is clearly just getting started here with their early ad offerings, but we’d love to see them ultimately do advertising in their own way. Every single tweet, re-tweet, follow, and search is a signal, signals that could ultimately be leveraged by savvy marketers to message to those consumers. Layer in location, demographics, and extend beyond Twitter.com to additional clients, Twitter-enabled web sites, and applications, and you’ve got an ad platform with unique scale and flexibility. This will obviously take time, focus, data, and a marketing community that can push the envelope beyond what possible in targeting and messaging today. We’ll obviously be watching Twitter very closely in the coming months.

As anyone who’s been exposed to this blog can attest – we love data and tools that empower users to interpret it. This week TWIR takes a look at some recent news & trends in digital advertising reporting.

Breaking Down Silos
On the heels of our post a couple weeks ago on measuring social media performance comes this MediaPost article on iPerceptions releasing a suite of tools that analyzes individual customer experiences based on the combination of voice of customer (VoC) and Google Analytics data. The 4Q product allows users to integrate Twitter & Facebook links and provides global industry comparisons from more than 6,000 Web sites across 35 industry verticals. As the post author, Laurie Sullivan, points out – recent acquisitions and partnerships in the data space “provides an example of the need for companies to work together and allow data to flow across platforms, rather than get stuck in silos”. It will be interesting to see how budgets shift as tools improve the marketers ability to interpret succes across platforms.

Favorite Search Marketing Reports
This Search Engine Land post by Mona Esseily details 5 of her favorite Search Marketing reports. As Mona notes, picking favorite reports can be subjective, but the insights – whether from these specific reports or others – are the primary toolset at the markets disposal for iterating & improving campaigns & messaging. Here are a few called out in the post:

* Which geographic locations are performing best
* Additional product advertising ideas
* New campaign ideas
* Does advertising to “bulk” and “wholesale” consumers drives a higher cost per order
* Percentage of ad spend broken out per hour of the day

The beautiful part of these insights is that they are media agnostic and can be applied against several tactics. As stated in this blog before, the continuing evolution of digital advertising reporting will directly influence how advertisers perceive the market. As metrics become more and more consumable and actionable digital will become a necessity for branding, direct response campaigns and everything in between.

Square Peg, Round Hole
Brian Massey of Conversion Scientist has a series Search Engine Land posts on parameters for measuring social media and how to coax social media insights from Google analytics specifically. A common thread through both articles is the need to be calculated in developing content to drive the desired social action and the approach to measuring results. This all may seem like marketing common sense but until the analytics marketplace evolves to the point where signals from the respective social media platforms can be consolidated, the impetus falls on the marketer to structure a campaign that can be measured within the context of their existing tool set. We’re hoping the evolution happens quickly.

Revolution
In truth, digital media measurement is probably in need of a revolution rather than an evolution. The early days of digital marketing were focused on metrics that are now part of our day-to-day vocabularies such as CPA and ROI. We are now in a place where things like retweets, likes, check-ins, etc. have real value and being able to quantify those and incorporate results into future efforts is becoming essential for success. There will likely be a lot of dollars left on the table until we reach that point.

The TWIR community has spoken. With 35% of the final vote, congratulations go out to Theodore Hong, creator of “Red versus Blue: Relative support for Conservatives and Labour at the 2010 general election” map below. Given that Theodore is a phd student at Imperial College in London, England we’re guessing this is not his first Data Vizualization rodeo. Here’s hoping the iPad provides an occasional distraction from the rigors of study. The poll results were very close as Jakub Górnik and his HeatMap portraying UEFA Country Rankings 1960-2010 came in a close 2nd with 34%.

We are very grateful for the great turnout and contest participation. Its warms our hearts to know we share our data passion with such a large community. Thanks again to everyone that participated in the contest or voted – we hope you found some useful and/or interesting content at this blog and will continue to visit. In fact, we enjoyed the community’s efforts so much we would love to hear your ideas for a follow up contest. Please drop a comment in the blog or send us an email at info@datapop.com with any suggestions for an entertaining & engaging contest leveraging data analytics, relevant ad creation, geo, etc.

Lastly, a big thanks go out to Pete Warden and his OpenHeatMap tool. We love (non-excel) tools that empower data and Pete’s certainly does the job. Please wear his T shirt with pride.

Thanks again.
TWIR Team

The Winner:

A reminder that you have until 12 noon PT on Friday to vote for a winner in the Data Visualization Contest. If you haven’t yet, check out the great work of the TWIR community at the bottom of this post.

This week we continue to keep a close eye on Search. It’s hard to believe but this week marks the start of October and Holiday season is in full swing for retailers. TWIR takes a look at some initial research on the impact of Google Instant and some forward, keywordless, thinking. Thanks for reading and don’t forget to Vote!

How Google Instant has Impacted Paid Search
George Michie of RKG outlines their recent research on Google Instant’s impact on Paid Search to-date. As usual, the RKG team has built a great case study and measures how Instant has influenced several aspects of SEM. The traffic, keyword data and ad position data is certainly compelling as is the anecdote on how Google Instant was rendering search results for “Toothpaste for Dinner” when just “Toothpaste” was in the search box. Performing this search now, Google seems to have re-ordered priority and serves ads accordingly but it is worth noting that Google may infer user intent differently for somewhat ambiguous terms post in the Google Instant world. As the RKG post suggests, advertisers may be well served to contrast keyword performance pre and post Instant launch and look for any major performance shifts – its possible your message may no longer be relevant to Google’s perception of the Keyword.

Paid Search Without Keywords
This click equation blog post by Alex Cohen examines the prospect of Search Marketing without keywords. Allowing publishers to determine the audience segment your offers are relevant for will require a paradigm shift but the move to a CPA pricing model will mitigate advertiser risk and should ease the transition. Remove KW strategies and tactics from the SEM equation and relevant messaging becomes the biggest tool at an advertisers disposal for differentiating themselves and appealing to their user base. Now that’s putting the Marketing in Search Engine Marketing.

Vote!

Thanks to everyone that participated in the Datapop Data Vizualization Contest . We have collected all the entries and are asking you all to choose your favorite map of five finalists outlined here. Links to the actual Heat Maps have been included below if you would like to review them further. Voting will remain open through Friday, 10/1 at noon PST at which point we will announce the winner. Good luck to the finalists and Happy Voting!

Map Links:

The geographic distribution of wikipedia georeferenced pages

Red versus Blue: Relative support for Conservatives and Labour at the 2010 general election

UEFA Country Rankings in 1960-2010

All the available books that people are sharing at Bookcrossing.com

Where’s the watts? Electricity usage across the UK 2005-08