Large Scale Bid Management – Pubcon 2007 session digest

After a nightmare traveling (I haven’t had much luck lately) with 6 hours to go from SFO to Las Vegas, I arrived only in time for the Tuesday afternoon session, “Large scale bid management”. I was anxious to listen in having done that very thing at HP and now find myself starting such campaigns again. My thought? This should be the perfect discussion for major brands, comprehensive catalog businesses, and sites with many many pages of content.

”The PPC programs from the Search Engines have grown more complicated and robust over the years. The amounts of variables have increased as it relates to ad text, bidding, budgeting, targeting, positioning, and distribution. While some small and well-structured campaigns can run with little or no maintenance, the challenges for managing large scale PPC campaigns across multiple channels compounds daily. This session will examine real life examples of large scale bid management, techniques, and tools.”

First up was Jon Kelly, President of Surehits purported to be the ad network for insurance and loan products. His premise? “Calculate click value, reward user choice, and watch data;” calculating click value requires consideration of the probability of conversion and the value of that conversion. Jon has an interesting approach to organizing a campaign, one which I feel is highly advanced but not technically possible given the paid search platforms available to us through Google and Yahoo. I’ve shared with you the importance of structuring your search campaign to analyze performance in aggregate instead of mining keyword level data (which is overwhelming with thousands of terms), Jon goes further with a strategy of tagging your keywords with multiple categories of organization.

• – Organize by – Model your market: geography, product, request (white paper, application, rates, etc.)
• * Do not do buckets – everything Oklahoma or all products = you lose data
• * (HOWEVER, and this is my editorial comment, buckets are better than nothing!)
• – Instead Tag keywords
• * City, state, type, service, price range, request type, etc.
• * You can then model various questions with predictions for performance based on conversions and value of conversions in each tag -> City is bad, state is good, this product is good, request is bad = good keyword phrase or bad?

Keyword tagging is a brilliant direction for the industry, in much the same was I’ve been frustrated that we can only organize by campaign and ad group when there are multiple layers/facets of keyword organization required to effectively model and measure program performance. The challenge is that today, keywords can’t be tagged in analytics or search management platforms. For now, use campaign and ad group segmentation to bucket keywords as best as possible and push vendors and search platforms to enable tagging so you can structure your keywords in as many dimensions as possible.

Jon goes further on optimization with the following:

• – Reward User Choice: customize landing pages to appear more relevant to users coming in through search
• * My caveat – don’t go too far. I for one can’t stand John Batelle’s “Welcome Google Searcher” personalization (see right) which puts this large box prominently on the page to help direct users. That said, I’m easily sold by data/results so while I don’t like it, John’s approach may be a best practice considering most users are not as savvy as those of us in this industry. I’d love to hear from John or perhaps he can point me in the direction of a case study on behalf of his customization.
• – Watch campaign data closely:
• * With localization, are you running city phrases or brand phrases? Is “Houston Mortgage” an actual brand? Mobile Alabama = Mobile Home Loans
• * Recognize unusual amounts of traffic on what you consider tail keywords; that may be a result of Content

Final thoughts: Jon seems to be doing a lot of work to determine potential value of keywords when I’d advocate tracking and weighting while running to manage bids on all possible keywords. To his point about critically analyzing keywords like “Houston mortgage,” I’d still be there as you should be bidding relative to conversion and value; that is, it doesn’t matter if you get unusual amounts of traffic on a keyword because it is someone else’s brand, if you bid relative to performance, it is the result that counts.

Next was David Rodnitzky, VP of Advertising for Mercantila which runs a series of specialty retail store websites; David also writes blogation.blogspot.com. David first takes a broader approach to simplifying the questions many large scale paid search advertisers have:

• – If you have no clue: outsource everything
• – If you have a clue but no resources: outsource the tail and use 3rd party software for the head
• – If you are an expert with a team: get 3rd party software to run your program

He shares other nuggets of wisdom:

• – Define the tail but don’t presume to know – the tail is not pre-determined!
• – For most, the “head” (or most significant keywords) is under 1000, really under 100
• – Simplify your management of the tail – simple bidding rules (i.e. 20% margin), standardize campaign negatives (negative keywords), run three generic ads that work across all campaigns, search and destroy bad keywords weekly
• * My caveat: certainly this is the approach you should take if you aren’t tracking and bidding relative to conversion. But failing to do performance based bidding is the greatest mistake you can make as an SEM.
• – The Tail can be a problem: what drives little volume now can change overnight
• * I disagree, see previous. But his underlying point is insightful: news, scandal, buzz in a related brand or product can drive monstrous search activity for a tail keyword. Keeps you bids in check relative to performance and benefit from the hype – don’t shut down because you can’t bid properly for the increased traffic.
• – “Tail” doesn’t even really exist due to Google broad match (now we’re on to something!)
• – Focus on the head: right keywords, bids, messages, and landing page
• * Expand that list! – plurals, synonyms, misspellings, etc.
• * Implement advanced bid rules – optimize for position (IS HE NUTS!?!?), factor in assists and latency (Yes! See Yahoo Search), Day-part, Geo-target, IP/Site target
• * Improve ad text – optimize for eCPM – remember greed, vanity, exclusivity, fear (test them)
• * Test landing pages – Google website optimizer, Offermatica; Bounce rate, exit rate, etc.

I’m mixed on David’s approach. He has some very advanced search marketing techniques such as factoring in assists and latency, recognizing the truth of the tail, and some conventional wisdom out outsource vs. in-house BUT his focus on the head, optimizing bids for position (!?!?) and lack of performance based bid management (apparent through “optimizing for position”) can lead to a waste of money or loss of opportunity.

Finally was Kevin Lee, a closer peer and someone with whom I’ve chatted and considered working. Kevin is co-founder and executive chairman of search vendor Did-It. Kevin started with an audience poll. All in all, his time was most insightful but it provided tactical direction, I didn’t hear the “how-to” that is most valuable from these presentations:

• – How many feel large scale bid management means lots of keywords? (scattering of hands)
• – You can’t afford top position all the time for every keyword – nor should you attempt to!
• – Focus on your campaign and what matters because search engines don’t mind your waste = search engines make money
• – Smart way to buy clicks – time of day, content vs. search, day of week, age
• – START WITH CUSTOMER DATA!
• – Higher interest level in your brand? By geo?
• – More offline conversion from the website? Track latency and indirect sales!
• – Are you measuring cart size or by product?
• – Tracking life time value?
• – Better targeting = better click rate = lower CPC for you = better advertising for consumer = more money for the publisher
• – Geographic segmentation may be the most powerful segmentation!
• – Pay attention to conversion rate 1.5 vs. 2.1 is 25% higher meaning you can afford 25% HIGHER
• – Daypart not based on clicks or volume but conversion rate! Of course you see the most volume during the day and around lunch time but is that when people buy?? You want that audience don’t you?
• – Synchronize creative and landing page – test different landing pages
• – Pick the right partners, technology, and strategy development

The large scale bid management session at Pubcon was diverse. That is perhaps the best review I can provide. All speakers were good and bad in their own way; the time was filled with good suggestions and bad, some gave a taste of their exceptional experience and skill but not enough information to really act. This was the kind of panel where I wish an audit had been performed beforehand to ensure consistent, effective information is shared. I don’t mean to sound harsh but here’s the one question from QA worth sharing; even this answer is insightful but woefully ineffective:
“Can you recommended bid management tools?” – (general response) no options given. There are impressive PPTs which make claims of effectiveness, quality, and cutting edge technology. Do a “bakeoff” of multiple tools. Give them the same campaign with which to work and a time frame in which to deliver. Which performs best but be sure to consider ease of implementation and de-implementation (if they require unique tracking codes be wary).