This Halloween, I wanted to ask some scary questions. Not about ghouls or goblins, but about the health of your paid search campaigns:
Is your search advertising program healthy? Are you getting as much out of it as you could? How do you know? If you outsource, is your SEM doing the job well?
Don't assume that just because you're paying them big money that they're paying appropriate attention to the details â€“ we've seen plenty of examples of gross negligence on the part of major SEMs.
Just because your search results in aggregate look good, doesn't mean the program is healthy and running at maximum efficiency. Once a year, or more often, a detailed evaluation of your program will pay big dividends in the long run.
Here's how to do it:
First: Gather Data
- Keyword-level Costs and Sales. Your in-house search manager or SEM should be able to easily provide you with clicks, costs, orders and sales dollars by keyword phrase and engine for the last month or two. Don't be content with the "top 50" or "top 100" terms, you need to see everything. If it isn't easy for them to pull this together quickly, then one of two things is true: 1) they aren't watching the important details of the program; or 2) they're hiding the details because they're ugly. Wave the red flag -- your program is broken.
- Google, Yahoo, & MSN Term Lists. Have your in-house manager or SEM dump the current active terms with the creative (ad text), the destination urls, the current bids and the current position (average position over the last week, or day is fine). Again, this should be trivially easy to access.
Second: Analyze the Data for each engine
- Brand vs. Non-Brand. Sort the Keyword level Cost and Sales data alphabetically by keyword. See what fraction of your sales are coming from your brand name and permutations (store name, storename.com, www.storename)? When you look at the costs and sales of these brand terms, and more importantly, the cost and sales of everything else do the numbers make economic sense? Don't let sales on your brand name subsidize poor performance on the non-brand phrases.
- Bid Management. Sort the same spreadsheet by cost descending, so that the most expensive terms come to the top of the list. Search is a "spiky" business, and it's important to look at a month or two worth of data, to eliminate some of the spikiness.
- For the twenty â€“ fifty most expensive terms: pull back the current bid and position from the appropriate Term List sheet. Do the bids and positions make sense for these critical terms given the economic results you're seeing? Are there instances where you're bidding a lot on poor performers? Are there instances where you're bidding too little on great performers and you're not high on the page? Bidding on mid- to low- traffic terms is a complex statistical game; bidding on the high traffic terms isn't. Don't accept a bunch of mumbo jumbo about the bids, if the bids don't make sense on the most important terms you can be sure they don't make sense on the rest
- Gauging the rest of the bids. Delete the "brand phrases" from the performance data, as these phrases distort the real picture. Re-sort the list by click-counts descending. Look at the sum of costs and sales for terms with more than 1000 clicks. Repeat this process for terms with between 900 and 1000 clicks. Repeat all the way to the bottom, ending with terms with fewer than 100 clicks. Bid management for the lower traffic terms is tough, but if each "bin" of terms doesn't have reasonable efficiency, that may indicate problems with the system. It's easier to break a fine watch than it is to build one, but if you're paying a lot for the watch, it aught to withstand this scrutiny fairly well.
- Ad Copy. Is the ad copy compelling, accurate and specific enough to the product? It usually isn't cost effective to hand write copy for every single ad, but the templates used should be at least category specific. And, it may well be worthwhile to put extra attention into your top 50 to 100 most viewed ads.
- Destination URLs. Grab a random selection of phrases and copy and paste the destination urls into your browser. Is that the page you would pick to land the user on? If not, find out why the landing page is what it is. Getting the landing pages wrong costs you sales, as many shoppers aren't patient enough to navigate more than is necessary.
Term Lists. How thoroughly do the term lists cover your offerings? A decent rule of thumb holds that you should have ~5 to 10 keywords for every product you sell. Numbers aren't the ultimate measure, as it's easy to inflate a poor list by simply appending or pre-pending "shop for", "buy", "online", "on sale", etc. to each term. Creating a truly comprehensive list with permutations that actually help capture more traffic is hard work, but well worth the effort. Low traffic, detailed terms convert at a much higher rate than more general terms and are often less expensive.
If you get to this point and everything looks good, congratulations, your search program is in good shape. If not, depending on the severity of the problems uncovered, ask hard questions or look for a new SEM. If you can't do the examination because you can't get keyword level performance data, fire your SEM or encourage your in-house manager to get the tools they need to do the job well.
Search is complicated, and getting every detail perfect is impossible, but if you uncover grievous problems take action. There's too much at stake to tolerate incompetence.
To read more, check out my talk at the DMA earlier this month with the Sharper Image.