Chris Sherman introduces the panel saying that metric people looked at was simply click through. That evolved into conversions, ROI, CPA and so on. Today we have lots of metrics to use to see, plus we have the ability to track a customer over time. That is what this session is about.
Ben Perry from IProspect was up first. If you use an agent you get a big difference in results when compared to humans. He showed how the agent (software) was much more aggressive in bid prices versus the human. The agent brought in a 37% increase in ROAS and a 55% increase in conversion. Time matters in bidding. He overlaid clicks and conversions on a 2d graph, for a specific client, it was very interesting to look at. There are also weekly trends, especially with b2b markets. Then he plotted a monthly view, which showed a combination of weekly trends, but it doesn't always look the same way for each client. Then again he backed it out by year and you see the same trends, and seasonal trends. Time should impact your bidding strategy. For a low consideration product, 88% of all conversions happen within the hour (batteries, snickers bar vs. car purchase, home, plasma screen). For high consideration products, 46% of all conversions occur between 1 and 60 minutes. 46% of all the other conversions too more than an hour. He plotted a scale by days and it took up to 6 months or so. Customers who waited longer had an average revenue per sale of 329% higher then those buying within the hour. For this client, 44% of revenue is driven by repeat customers (life time value of a customer). So what if we did not consider the 44% of revenue as a life time value within the bidding. Non repeat revenue would have been 59% lower if we had not counted lifetime value.
Alan Rimm-Kaufman from Rimm-Kaufman Group was next up to talk about the time warp. 3 questions and 4 tips. How soon can we tell if something is working? Why do we care? How does this wait affect bidding? The sample: 1 million paid search click (random sample from 6/1/04 through 6/1/05 Google & Yahoo. 41,377 subsequent conversions (6/1/04 - 8/1/05. 85% / 15% split B2C vs. B2B advertisers. Long duration cookie. How long must we wait? many order come quickly, 50% within 28 minutes. But getting nearly all takes longer, 75% within 25 hours, 90% within 12 days and 95% within 4 weeks. Are they looking for you? Search phrase contains client brand: 50% in 28 minutes, 75% in 3 hours and 90% in 7 days. Search phrase down not contain client brand; 50% in 32 minutes, 75% in 1.5 days, and 90% in 13.6 days. Big ticket purchases? $0-99 1.5 days for 90% to order, 100-199 about 2 weeks, 200-399 about 2.2 weeks and 300+ over 2.5 weeks. Holiday affect, as the holiday season gets closer, they order quicker (they need it now). Vertical? Hobby supplies, 50% in 38 min, 75% in 20 hours, 90% in 6 days... Consumer electronics almost 16 days for 90%. And professional leads take about 3 days for 90% of those orders to close. Why do we care? On average it takes about 2 weeks for advertising to sales ratio to stabilize. Tip # 1: Be patient. Don't overreact to short-term results. Long cookies; method, long cookies 90+ days, order crediting in reporting; 14 days, 30 days as per client. Preserves full data, can recast history to show impact of max interval decision. Tip #2: Handle click to order interval in reporting, not in cookie. Bid management & day parting. Bid management must address click to order delay. Day parting must be on time of click, not order. Example; over 25% of Monday midday orders driven by pre-Monday clicks. Tip#3 compute economics on a P90 conversions. Tip #4: If day parting, day part on click date, not order date.
Finally we have Rob Gaudio from MEA Digital to discuss a case study for one client, Oakley sunglasses company. This is more specific about allowing enough time to measuring conversions for your paid campaign. Oakley, Inc. World's leading manufacturer of premium sunglasses. An expanding line of performance footwear, apparel & accessories, sold in more than 100 countries. Investments in Direct Busines; new e-commerce system in Q1 2004, paid search campaign in Q2 2004, redesigned site in Q2 2005. They only sell MSRPO. Brand conscious, historical focus was on brand image and not ROI focuses. SEM Program: They developed an extensive keyword portfolio, they implemented core metrics to track and optimized daily, weekly, monthly and compared to quarterly and annual data. Technology integration; Google/Yahoo PPC, Atlas OnePoint and CoreMetrics. SEM Test Plan: Objectives; optimize SEM program in time for peak buying season, discover optimal time to test creative. They executed a creative copy test, suing time as an additional variable. Used Google AdWords. Isolated other variables like keyword position, cost and daily budget. Selected a stable time period to avoid major seasonality. Success metrics; click to purchase, cost per sale, cost per order, conversions, etc. Consider outside influences; other media, world news, weather, internal company issues, holidays, and competition. Implementation. Creative tests: used top performing creative from 2004 as a benchmark. They ran 4 creatives at same time (did not use auto optimizer), for 4 weeks. Monitored results for 7 days periods, and cumulative. Compared data on 7, 14, 21, and 28 day periods. Results: Summary conversion data was consistent across all weeks to the total (would not have known that if we did not allow the time to past). Determined the best performing copy at the Ad Group level. More aggressive call to action creative performed best. Products are seasonal and require attention to the product level. Some creative did not have enough statistical data to warrant a change in Winter. Next on the horizon is to retest in Winter. Use learnings for future launches; international SEM initiatives in 2006 and new products and keyword categories. Monitor sales data from test period sales for repeat purchases. (1) keep good historical data, (2) seasonality is important (3) give it time, (4) keep it simple and (5) watch for outside influences.
Forum discussion at Search Engine Watch Forums.
