Thanks again to Ryan Gibson from the Rimm-Kaufman Group for sharing some top-notch tips on paid search. If you missed the call, catch up with the full length replay of the Key to PPC for Online Retailers in our eCommerce webinar archive. If you’d like a summary in blog post format, here’s a recap of the main points with answers from Ryan to questions submitted after the webinar ended:
Keyword Development is 50% of Your Success
Keyword research is essential because it ultimately determines which searches your ads will appear for. Keywords must be appropriate, specific and capture the variations of how people search for the products you sell.
“Long tail” terms are desirable. Although they don’t get a lot of clicks, there is not much competition and it’s the specificity that makes them high-converting. You can also get them low-cost, and the lower the cost, the lower the risk. You’re not spending anything unless they are clicked on. Long tail terms in aggregate may be attracting a small percentage of clicks but be driving the highest number of sales because they convert highly.
General (and shorter) queries may indicate one is in research stage and thus, convert lower than more specific searches, when customers are aware of exactly what they are looking for.
Keyword List Tips
Test 3-10 Keywords per SKU
Modifiers don’t count, such as “buy widgets” or “buy widgets online.” Also, branded terms don’t count. Rather, you want to test out unique terms. So start with a URL (product page, right down to the sku – so if you sell blue and red widgets, they have their own keywords, as do different sizes). 2 benefits to this approach are:
1. Ensures you have the product you’re creating the term for
2. Makes it easier to tie up correct landing pages
Beware of keyword research tools like Google’s Keyword Tool. They are good in concept to find additional terms to fill in holes, but you can’t just import keyword suggestions without human oversight. Keyword research tools don’t understand your product, or generate terms that are irrelevant to your business. (I think keyword research tools are great for finding negative keywords, but that’s only if you’re using broad match). This is important, if you sell musical instruments, flutes, don’t run paid search for wedding and champagne flutes.
Jason mentioned using Ebay for keyword research, if you missed my guest post at SEOmoz check out the link.
After testing your keywords (letting them run for two weeks, or 30 days or more depending on your traffic), typically 50% will remain active and 50% will be paused, deleted or moved to negative match.
How Do You Determine a Poor Keyword?
1. People click on general terms before specific terms, so general terms don’t get credit they deserve. (Conversion would be credited to more specific term, though customer clicked general term first)
Busted: Multiple PPC clicks are rare, generally user clicks same ad twice.
Ryan recommends crediting the last click that came through for the conversion.
2. People search for non brand phrases, then brand phrases.
Busted: small percentage of brand sales may follow this pattern. Allocate credit to non-brand in this scenario.
Keywords – Managing Match Types
The value of exact match traffic is twice that of broad match because it’s perfectly targeted. When bidding on “laptop battery” with exact match, you won’t show up for “how to extend the life of your laptop battery.” But what if you want to appear for a long tail search like “buy replacement laptop battery online”? You’d want to be broad matching “laptop battery.” You can use multiple match types, just make sure you bid higher for your exact match so Google will select it over the broad match (Google wants to maximize revenue) so it will appear for exact matches, and your broad match will still appear for other queries.
Because exact matches convert higher, you can afford to bid higher for them. (Also, if your quality score and click-through rate is solid, you could be getting good positions for a better cost-per-click anyway).
Tip: Turn “automatic matching” off. Automatic matching may serve your ad for a keyword that you don’t have in your Ad Group. It’s currently in Beta, so you won’t be included unless you have an invite from Google to participate in the Beta.
Good ad copy is better than ad copy. Great ad copy may only be slightly better than good ad copy.
Make sure you “echo the search string” – you want to show the keyword in the ad copy (will appear bolded). This includes headline, text and display URL. This confirms you have what they are looking for, rather than showing “Lucky Brand Jeans” for a search on “denim mini skirt.”
Provide a compelling “why shop here?” value proposition. This could be a promotion, numbers, percentages or “free.” But use “free” appropriately. If free shipping is only available in a certain geographic region or after a certain purchase threshold, customers could abandon your offer because you weren’t up front about that, and the offer is no longer appealing.
(Note: Don’t be afraid to include prices. If your prices are high, it discourages bargain shoppers from clicking and costing you money).
It’s paramount that your landing page reflects the promises in your ad, including correct product and price.
TEST! Test, and test.
Don’t just evaluate click through rate, evaluate overall conversion and efficiency. You want clicks, but only if they are qualified. Don’t waste your money on semi-targeted traffic.
When testing ad copy variations, make sure you have turned off
To do this, go into your campaign, click Edit Campaign Settings / Advanced Options, and make sure your Ad Serving is set to show ads more evenly so you can really compare the click through rates of each.
Even if you find a “winner,” never stop testing. Keep your winner and test new copy against it, using the same ad serving option. Test offers vs. no offers, too. Offers will usually have higher CTR%, but might have lower conversion rates – so you have to think beyond the click-through success.
Selecting Landing Pages
Select appropriate landing pages to minimize clicks from the ad to the cart. There is a big difference between good and bad copy. However, testing good and great copy will give you incremental improvements but won’t move the needle that much.
Select a landing page that takes user to the exact product they are searching for. But don’t make the decision for them – if they search for “Merrell shoes,” don’t land on one particlular model of shoe, rather the category page.
Test landing pages to discover what works for your pages and your products. Dynamic Keyword Insertion on landing pages can help maintain the “scent” and convert higher – keeping tightly focused relevance.
You may also consider testing custom pages for your highest traffic or highest converting keywords.
Understand YOUR PPC Economics
Brand vs. Non-brand Searches
With branded searches, the customer is aware of the item before search. Non-brand searches are “incremental.” Paid search *shouldn’t* be credited for branded searches, because the awareness could have come from other advertising, email marketing or word of mouth.
Non-branded search is really what you should care about in terms of how is your search program performing for you.
Economics: Order Latency
How long from clicks to order will be indicated by the price or the category it’s in. A considered purchase like a desktop PC will take longer to make the decision than a shirt (usually). Using your sales data, determine how many days it takes for 80% of your orders to come in after the click. (At this point I was dropped from the conference call and missed a slide or two)
Metrics that Matter:
Margin should drive ad spend to ensure you can really afford the clicks. A target rule of thumb is this formula:
1-cogs-vc/2 -> 1 – [ (Cost of Goods Sold) – (Variable Costs) ] / 2
If COGS is 50% and VC is 10%, your target Ad Spend to Sales ratio would be 20%:
1-.5-.1 = .4 / 2 = .2
Margin based bidding:
A single cost/sales ratio ignores the margin differences between categories. Margin based bidding results in better bottom line ROI.
Good solution: set different ad spend targets based on category margin structures.
Best solution: bid based on observed margin with frauds and cancels factored in.
Example: A click on water bottle might drive the visitor to your site, then the visitor buys a kayak instead. This is much higher in value, but you still should credit the water bottle keyword for the conversion. It attracted the visitor.
Frauds 30-60 days after the fact should be tied back to keyword, a keyword might be driving a lot of returns or frauds and you should identify that.
Time of Day/ Day of Week
Observed SPC (sales per click) doesn’t vary by position on the page, but it often varies by time of day and day of week. You may find that researchers, mindless browsers click on search ads during lunch hour and purchase at 4 pm. Base your decision on what times of day clearly have highest sales, and daypart your campaign accordingly.
• Paid search programs should be evaluated on incremental (non brand) results
• Smart ad copy is about echoing the search and offering a “Why Shop Here” message
• Broad match can be dangerous, use it, but watch carefully
• Drive the program to keyword level economic goals
• A margin-based metric is the most direct approach for driving ppc profits.
Questions and Answers
Ryan kindly answered audience questions after we ran out of time, here’s the answers:
Would you spell out the margin based metric again?
The ‘rule of thumb’ guide for calculating your A to S is as follows:
A / S = (1 – COGS – Variable Costs) / 2
For instance, if your cost of goods is 50% and your variable order costs (pick, pack, ship, phone support, etc) are 10%, the formula would look something like:
(1 – .50 – .10) / 2 = 20% A / S
It’s important to remember that this is simply a rule of thumb. Ultimately, the number has to make sense for your overall program. If you’re finding that PPC drives very loyal customers and they’re buying again after their initial purchase; it may make sense to spend a little more, knowing you’ll make it up on the subsequent purchase.
For any math fans, Alan provides a detailed explanation of the calculations and limitations.
Would phrase match would work if the searcher looks for “Blue and Gray Widget” and your keyword is “Blue Widget”?
Phrase match means that your keyword term must appear somewhere in the search string, with the words in the same order and nothing in between.
For example, assume you have both “blue widget” and “gray widget” in your program on phrase match. Your ad for “gray widget” would be served for a search on “blue and gray widget” because those specific words appear in the search string. Your ad for “blue widget” would not appear, because the search string includes additional words between “blue” and “widget”.
If “blue widget” and “gray widget” were in your program on exact match, neither would be served for a search on “blue and gray widget”.
If both keywords were in your program on broad match, it’s likely that the engine would serve whichever keyword had a higher max bid.
Where in Google Analytics do you see the 30 day time? I can never find that.
Apologies, I was thinking of the Adwords cookie window when I said 30 days, not Google Analytics. Adwords has conversion tracking that you can enable and that window is set to 30 days. The default cookie window for Google Analytics is 6 months.
Is there a way you know of to change the reporting/tracking window on Google Analytics?
By default Google Analytics looks at 30 minutes for same session data and a 6 month overall cookie window. You can find info for customizing these timeframes here: http://code.google.com/apis/analytics/docs/gaJSApi.html#_gat.GA_Track…
f you do break out separate campaigns (eg US vs. Canada), what’s the best way to sync the ads and keywords?
It depends on what you mean by “sync”. I’m not sure if any of the engines will “sync” the keywords in two different campaigns.
If you’re looking to run the same keywords and ads in both geographies, Google allows you to copy keyword, ad and campaign info through their Adwords Editor tool. I believe the other engines offer tools through their user interface to copy campaigns.
If you’re talking about grouping the data together to get a better look at terms, categories and groups across geographies, I believe that the reporting tools that the engines or analytics provide can group this data.
Can you talk about geo-targeted terms a bit. Do geo-targeted terms tend to cost more?
The costs for geo-targeting terms vary and really depend on your industry and the competitive landscape, just like buying clicks for a national campaign. As a whole, we don’t typically see that geo-targeted terms are any more or less expensive than those that are not.
If you’re considering a geo-targeted campaign, be sure to keep in mind the way that people are searching for your products. Should you geo-target product terms, use geographic modifiers along with the product terms or both?
For instance, if you’re selling a special kind of widget in Pittsburgh, do people in Pittsburgh search for that product by looking for “widgets” or by looking for the geo-modified “Pittsburgh widgets”? If it’s the former, a geo-targeted campaign for Pittsburgh that doesn’t us geographically modified terms may make sense.
On the other side, if you find that people use the geographic modifier more often than not, you may have more success running a national campaign for “Pittsburgh widgets”. A national campaign for “Pittsburgh widgets” would also capture traffic for people in Cleveland who are looking to find “Pittsburg widgets”.
How many clicks and orders do you need for a term to evaluate it’s cost effectiveness? What does it take to have statistically significant info?
This is something that’s difficult to answer. Ultimately, statistical significance is dependent on a lot of variables. Conversion rates matter. AOV (average order value) matters.
For instance, if a term is converting at a much higher rate than the rest of the program, you can probably determine sooner than later that it’s productive.
If the AOV for your site is $50 and you’ve already spent $500 on a term, $500 seems like a fairly significant number to me regardless of the number of clicks.
What do you mean by use your URLs to create keywords, can you clarify?
When you’re creating a term list for your products, we’d recommend looking at your website and creating keywords for each URL (web page) on your site as opposed to looking at a product list, catalog, etc.
This ensures that the ads will be specific and appropriate for your website’s product offering and the ads will ultimately be driven to the most appropriate landing page.
If you have a phrase match ad for “Aerosmith tickets” and broad match for “Aerosmith London”. which ad would come up for the query: Aerosmith tickets for London?
Assuming you have “Aerosmith tickets” on phrase match and “Aerosmith London” on broad match in your program, the engines will likely serve the one with the higher max cpc for a search for “Aerosmith tickets for London”.
In this example, I’d expect the ad for “Aerosmith London” would likely have a higher conversion rate because we know the user is looking for a specific location and the opportunity exists for more tailored ad copy and targeted landing page.
What are your thoughts on the content network?
The conversion rates are going to be lower than the search network because you’re not finding folks when they’re looking for a product. They’re reading an article on another site and your ad is next to what they’re reading.
Content can be successful when carefully managed. Following Google’s best practices for content helps, too. A few tips:
1. Be sure to separate your Adgroups that you’re using for content from the Adgoups for your search keywords.
2. Also be sure to limit the number of keywords in your content Adgroup to 50. Google only looks at the first 50 keywords in an Adgroup for content targeting.
3. Make use of their site exclusion options. If there’s a site driving a lot of clicks and no conversion- try excluding that site from the content program. (This is different from their site targeting program, where you’re able to select each site where your ad runs.)