One of the sessions at SES that I found most interesting was the Advanced Paid Search Techniques panel. In some ways, paid search opens up a different set of opportunities than organic - not all of which depend on how big your budget is. Through different headlines, ad copy and even smaller variables like capitalization or plural usage, it's possible to test and track ads for conversion and profitability in a way that just isn't there for organic (In theory, it could be done, but would generally involve cloaking which is something most engines frown on in organic results, even if done for legitimate reasons).
Once you've established a paid search campaign, what's left to do after establishing a certain bid price, clickthrough rate and conversion rate? Plenty.
Rob Pierre's presentation is on structuring campaigns to effectively grow and manage a keyword portfolio. There are 6 main categories Rob uses as the foundation of an effective PPC campaign:
- Campaign Structure
- Keyword Generation
- Ad Copy writing
- Campaign Site Development
- Destination URL Allocation
- Launch Bid Strategy
This presentation focuses on the first two.
It's a mistake to grow a campaign organically. Campaign structure and keyword generation can help with growth, and facilitate management and turnover to other people and divisions. It's ideal to start the campaign with a specific naming convention. Establishing a naming system ensures that all keywords are loaded on Google, Yahoo and MSN, loaded on all match types and across all channels (e.g. search, contenet network, etc.).
An example name structure would have a general account name that matches with a category name such as Cameras & Camcorders. Campaign names could be channel + keyword (Search Digital Cameras and Content Digital Cameras for example). And finally, ad group names could be based on sub-categories (5 megapixel, perhaps) and match (broad, exect, phrase, etc) and keyword types (generic, brand, event).
The end result is a campaign structure that can be managed across multiple engines and multiple keyword types. The goal in having keywords on all match types is to make the campaign laser targeted by ensuring that each keyword is associated with the most targeted ad copy possible and clicks go to the most relevant website page.
As an example, on broad match a search for Digital Camera Accessories and Sony Digital Camera can both trigger the Digital Camera campaign. Users would likely get generic ad copy, and land on a category level generic page. In reality, most people would prefer to show different ad copy and send the users to different landing pages.
Instead of trying to create phrase matches for each possible phrase, it's possible to use the established structure with "match type negatives" to force the right type of match to be shown on a given query. If a search for Top 10 Digital Camera comes in, here's how the process looks
- Broad match - "phrase match": in this case, that sends the query to the "Top 10 Digital Camera" campaign.
- "Phrase match - [exact match]: in this case "Top 10 Digital Camera" - [Top 10 Digital Camera] sends the query to the [Top 10 Digital Camera] campaign.
The query Top 10 Digital Camera in the UK would stop at the "phrase match" campaign. The query Top 10 Digital Cameras to Buy would stop at the broad match campaign. This will give the opportunity to mine raw search terms, identify negative keywords and create new ad groups. You can eventually point new [exact match] keywords to more relevant landing pages.
Patricia Hursh is up next with tips for long sales cycles. This could be a high priced service or high ticket item . The three main tips are
- Align ad copy with the buying cycle
- Offer landing page action options
- Understand benefits beyond conversions
Research and selection are the two points that generate the most search volume. Early and frequent reach is not the favored method of reaching customers, but is more applicable to those buying higher priced and long sales cycle items or services.
The 5 stages of a purchase are:
Awareness > Interest > Education > Comparison > Purchase
Translated for an actual purchase, the search path could look like:
Laptop Computer > Laptop Information > Laptop User Reviews > IBM Laptop Models > IBM ThinkPadT61
Ads should correspond to each query.
Overview of features, buying tips, avg price ranges
Compare top brands, user & professional reviews
Compare specific models, purchase prioritization chart, model discounts and special offers
Branded vs. non-branded searches: branded searches rise in the weeks before purchase. Generic searches are the ones that take place most often before the last click.It's worth measuring the search history of a person who converts on a branded term. Branded searches may get too much credit.
Landing pages should offer options that can capture those earlier in the buying cycle. Consider options that are less decisive than contacting or downloading to maximize interactions. Patricia presents a case study showing that total volume of online interactions increased by 30% by adding secondary action options.
Benefits of PPC beyond leads and sales should also be considered. These include brand awareness/recall, competitive market positioning and engagement and interaction
A brand had a 16% lift in brand awareness when they had the top Organic and top Paid listings. For brands who want increased visibility, keywords may not be high for lead generation, but can lift other areas.
Pete Wailes is up next with a presentation that's pretty brilliant, but not for the math averse. It covers using PAM-VAR (Pareto Analysis of Multi-VariateArray Results) testing to improve PPC campaigns, though the principles are applicable to anything that can be tested. Pete and Teifion created a multivariate testing tool to help with this.
The Pareto principle states that 20% of any system will generate 80% of the system's results. For campaigns this means you can test small portions of possible variants and have reliable, useful data.
5 variables with 5 treatments gives 3,125 permutations. Testing a sample of 25 scales. Just make sure to rotate variations evenly, otherwise data will be useless.
Using Taguchi arrays helps with this. The key elements of his quality philosophy are
- Taguchi loss function, used to measure areas of poor performance
- Designing tests so that they are insensitive to parameters outside the tester's control
- The use of an outer array for factors that are uncontrollable in real life, but are systematically varied in the experiment.
2 means testing variants at the same time. For example, a financial services company would probably have different results in January 2008 vs. December 2008 due to economic conditions.
When you've collected evenly rotated data, calculate the Variable Treatment Effectiveness Index (VTEI) values. VTEI is calculated using the formula CTR * Conversion rate * 10,000
Using the tool referenced above, you can quickly narrow ads down to the best performers. Pete was able to use the multivariate testing to reduce a client's spend from £7 to £3.
The final presentation is from Adam Goldberg, who makes the case for changing the metrics that e-tailers use to make decisions.
The first slide takes on Cost per Acquisition (CPA), Sale Price and Prrofit.
For the term Plasma TV, let's say that the respective breakdown by engine is
- Google: $225 | 1000 | -95
- Yahoo: $375 | $3000 | $165
- MSN: $450 | $3000 | $270
For lead generation, let's say the breakdwon is
Google: $40 | 2680 | -100
Yahoo: $75 | 1500 | 2020
MSN: $85 | 2805 | $2200
Using the wrong metrics can lead to wrong decisions for e-tailers. If revenue isn't taken into account, the acquisition cost may be off. Without tracking the profit margin, revenue number can be off.
For cost per lead, take into account the leads required to make a sale. Go further and take into account the sales price and profit. Additionally, lowering CTR can increase profit. For example, if there are 3 ads with 4, 3 and 2% CTR respectively, you may see an increase in profit as you get to the lower end. Why? If you're selling guitars, the first ad may be general and mention low prices. The second may mention specific brands or items and not focus so heavily on prices. The last ad may mention a high price point that lowers the clickthrough, but increases the order size.
The right metrics can vary by the type of campaign you want to run. For a Wal-Mart campaign, you want to run low margin/high volume campaigns. For a Harrods campaign you want a high margin/low volume campaign.
Never underestimate the air conditioning. If a merchant leaves the doors open in summer, the electric bill goes up, but conversely, so do sales. You can reduce the electricity bill, but it would also reduce sales.
Track all advertising together for a
complete view of revenue beyond the last click. Lower click rate with
qualifiers to increase profitability.
Premature analysis can occur with PPC analysis. Ignores possible purchase paths (referenced earlier by Patricia Hursh as well), can cause the keyword to be killed prematurely and can lead to setting wrong bid prices.
Brand keywords are not as valuable as you think. Brand terms close the deal, but other purchase paths start it. Additionally, people sometimes use them for navigation
There was time for Q&A towards the end.
Some of the takeaways:
Adam Goldberg: The content network can be a good alternative for the "air conditioning" if a budget doesn't allow for more expensive general search terms.
44% of products purchased
come from campaigns not advertising that product (this reinforces many of the points in Patricia Hursh's presentation, she asks if it comes from a study. Adam responds that it does)
Volume or CPA will be the constant in any campaign. Choose one or the other.
Test variables at once so that any external conditions affect the ads at the same time (i.e. economic conditions, holidays, lower general traffic)