Jan 242007

12 tips for winning at paid search in 2007

Here are 12 suggestions for winning at paid search in 2007. These tips are from my January column in Catalog Success magazine.

  1. Focus on Google. The reality is, Google controls more than two-thirds of the search market and is growing rapidly. Yahoo! continues to lose market share each quarter. MSN is a far distant third. Ask is even further back. Allocate your attention proportional to your ad spend. Don’t completely ignore Yahoo! or MSN, but invest the most love and attention in your Google campaigns. You’ll be rewarded with the largest return for your time.
  2. Grow your term list. As a rule of thumb, test three to 10 times as many distinct keywords as there are pages on your site. For example, if you carry, say, 5,000 SKUs, test between 15,000 and 50,000 terms. Assign tracking codes to gauge sales and costs at the keyword level of granularity (or, better, at the by keyword-by-landingpage-by-copyversion-by-matchtype level).
  3. Make your landing pages relevant. Don’t send paid search visitors to the homepage. Instead, deep link them to the most relevant selling page on your site. If you have more than one candidate for a relevant landing page, assign tracking codes and test all options.
  4. Split out branded from nonbranded terms. Many catalogers find that a large portion of their pay-per-click traffic and sales come from searches on their brand. Advertising on your brand makes sense; the cost is low and the sales are high. Do it, but realize that traffic on your brand name isn’t incremental. These are searchers who are looking for you due to your reputation and other marketing efforts. Don’t allow your agency or your in-house team to use your brand portfolio to hide a poorly performing nonbrand portfolio. Separate the portfolios in all your reports. Charge your search team to grow nonbrand portfolio sales aggressively and profitability.
  5. Use a strong bid management technology. If you’re spending more than $5,000 a month on paid search clicks, you’ll benefit from a bid management tool. Seek a tool that:
    • optimizes your selected performance metric across your keyword portfolio;
    • operates at the keyword level, as opposed to AdGroup level; and
    • uses strong statistical models to infer performance for low-volume terms
    • sets bids based on economics, not by simple position.

    Look for a flexible tool that allows you to specify different metrics or different targets for different types of terms. Avoid optimization algorithms that are “black boxes.” A tool should be reasonably transparent. It should provide a sensible rationale for each bid. All bid management platforms are not created equally -- a stronger platform can lift sales or efficiency by 10% to 15% just based on algorithmic strength.

  6. Track paid search into your call center. Some catalogers have begun to display a visible Web source code on the bottom of each Web page. This code indicates which tracked link drives the click that initiated the current session. I think this is a brilliant innovation.With such a code, when a caller tells your customer service rep (CSR), “I found you on the Web,” the agent can ask the customer to scroll to the very bottom of the page and read off the letter code in the brackets. The CSR then punches this Web source code into your order management system as if it were a catalog source code. This powerful idea allows you to track the spillover from paid search into your call center.
  7. Bid by margin. If your products have considerably different margin structures, fold these variances into your bidding logic. Do this at the category level or preferably at the keyword level. Compute margin based on the actual products ordered, not on the terms advertised, as certain ads will generate significant orders for products beyond the one advertised.
  8. Use contextual advertising with caution. Contextual advertising (AdSense, Yahoo! Publisher Network, etc.) performs poorly for many catalogers. First, these potential customers are viewers, not searchers, and therefore are less qualified prospects. Second, an overwhelming majority of click fraud occurs on the content networks. Generally, less sophisticated search marketers should avoid advertising on the content networks. If you do so, separate the tracking of these ads from the reset of your search marketing campaign so you can monitor them carefully. There are quality clicks at reasonable prices to be found in the content networks, but it takes expertise to find these needles inside the haystack.
  9. Scrutinize your affiliate program. Many catalogers drive up their paid search costs by mismanaging their affiliates. Analyze where your affiliates get their traffic. Don’t allow affiliates to advertise on your brand name. And by all means, don’t provide affiliates with better offers than those on your site.
  10. Manage your agency. If you use a search agency to manage your paid search campaigns, consider the following:
    • a large fraction of your sales will be driven by your brand name;
    • these terms typically have low cost; and
    • your agency played no role in brand reputation.

    Then consider paying your paid search marketing agency as a percentage of ad spend, as opposed to a percentage of sales. Establish clear return on investment metrics. Ensure your contract has an easy out (30 days or less) if performance or service targets aren’t met. Negotiate a maximum monthly dollar cap on your agency fee to protect you from soaring bills in your heaviest traffic months.

  11. Audit your search programs every six months. Compare the costs shown in your search reports with the actual invoices from the engines. Compare the sales reported in your search reports with order detail records. Determine whether any orders are counted twice due to different cookie tracking. Analyze keyword-level efficiency by computing the ratio of cost to sales by keyword. Sort this list by descending ad cost to ensure your most expensive terms are pulling their weight. Analyze product-category-level efficiency, folding in your category margin data to ensure your campaigns are generating profit, not just sales.
  12. Obsess about conversion. In the long run, the marketers with the best site conversion can afford to pay the most for paid search traffic, allowing them to grab the largest share of available inventory. The ultimate winners in search are the sites with the highest conversion, as measured by margin dollars per visit.



2 Responses to "12 tips for winning at paid search in 2007"
Alan, I enjoyed your overview for 12 tips for winning at paid search. In particular I noted the recommendation of bid management technology for anything over $5,000/mo.! A bit lower than the number I'd set, but that's not the point of writing. It leads me to wonder as I so often do when hearing about this general theme: WHICH VENDOR!? Not unrelated - I navigated to that excellent article by the analytics guy who has the recommended questions for analytics vendors. Similar issue... I'm pretty sure those are excellent questions but why hold back the all-important opinion: in the author's opinion, who's giving the best answers? Any feedback appreciated. This is in my ongoing quest to understand what others in the business are doing to serve their own clients (so I can serve mine), and also as a writer, revising a book on Google AdWords, needing to find out what other experts think about third-party tools. All the best, Andrew
Hi Andrew -- Thanks for the comment, and apologies for the slow response -- been travelling a bit. Rather than praise or damn specific tools by name, I'll list what I think are key attributes of a good third-party tool for bid management. In no particular order, here's a fast list: * Appropriate scale. A bid management tool that handles complex campaigns of 100k+ terms will likely be overkill for a campaign with 500 terms, and a bid management tool that is friendly and easy for 500 terms won't scale to 100k+ terms. * Econ-based bidding. Be it A/S, ROI, ROAS, CPO -- most marketers have a target econ effiency. A tool should maintain this target. A human should not have to tweak bids to ensure targets are met, this is the prime role for the machine. Bid-to-position is a nice option for branding efforts, but generally destroys profits in most other situations. * Term level clustering. Bidding phrase-by-phrase only works for terms in the head of the search distribution. The bulk of terms (not bulk of cost, bulk of terms) in a comprehensive campaign will be have no sales. Bidding these cost-only terms intelligently requires clustering them appropriately. That's the true "secret sauce" of bid management. * Term level bidding. Bidding should occur at the phrase level, not the adgroup level. * Strong algorithms. Bid algorithms matter. Better optimization algos and strong clustering leads to better results. We've run very clean head-to-head tests -- same client, same terms, same copy, same landing pages, the only difference between bid management platforms -- and seen significant differences in sales and efficiency (double digit improvements in both). * Transparency. If the tool bids 50c for 'widget' and 95c for 'wodget', it should be able to explain why in a way users can understand. Don't trust black boxes -- look for strong algorithms that provide clear rationales for their bids. * Post-sale considerations. For some retailers, post-sale returns, cancels, frauds, upgrades, signup-to-conversions, etc play a significant role in the value of an order or conversion event. If so, look for a tool which supports post-sale data inputs. * Strong reporting. The tool should present good summaries, allow drill-down into specifics, and slice-and-dice analysis. Costs (adspend) should be presented clearly alongside benefits (sales, orders, leads, downloads, whatever. * Time-of-day and day-of-week dayparting. Make sure all these time- and day-based optimizations are based on time-of-click, not time-of-sale. * Clear separation of brand vs. non-brand terms. For most advertisers, their brand plays a large role (high sales, low cost) in their campaigns. Smart search marketers should be looking to grow the non-brand portions of their campaigns. Look for tools which support the brand-vs.-non-brand distinction in bidding, reporting, and analysis. * Easy for humans. The tool needs to be friendly to the folks using it. The tool itself should assign tracking codes, allow A/B or A/B/C/D or multivariate testing, etc. Human slog should be reduced -- human slog efforts waste time and introduce errors. * Good exception handling. Marketing (and especially online retail) are human activities, and are thus full of messy exceptions. "Bid to economics, except for these phrases." "Don't adverise lead solder in California." "These products ship ground, so the Christmas ship deadline is one week earlier, so drop the bids on said products accordingly." Etc etc. A good bid management tool allows these special cases. * Different objectives for different portions of the campaigns. Marketers come to table with different objectives for different portions of the search campaigns: some terms are run for branding and impressions, others for efficient sales, others to make the owner happy, others to intimidate the competition, others to flush out overstocks, etc. Make sure the tool doesn't force you into a one-size-fits-all optimization objective. * Easy exit and data-export. If you opt to stop using the tool, make sure you can get out all your data. * Cross-engine ease. While respecting the Google TOS, look for a tool which enables you to handle all the major engines conveniently, while still accessing the unique features of each. * Items and phrases. Look for a tool which can report on items (skus, names, prices) sold, as well as orders. Look for a tool which can report on both the phrase-searched as well as on the phrase-advertised. * Appropriate price. The cost of bid management is friction -- the tool needs to generate lift many times the cost of the tool itself. For small spends, the cost of the tool matters. For big spends, the cost of the tool should vanish (in terms of percentage of spend), and all that matters is results. I have the sense you were seeking a review of tools by name, and so perhaps this response is a cop-out. If so, apologies. Not confortable putting up charts and rankings by name. I don't know of any tool publically available today that offers all these features. Indeed, that's why my firm built our own platform, and why we only run it for clients, we don't rent it a la carte. Cheers -- Alan

Leave A Comment