At both SMX West and SMX Advanced I made an effort to attend PPC panels, but I never ended up sharing my notes. This post features two conferences’ worth of my PPC education. I’ll try and avoid repeating / summarily address what is well-known and just address the higher quality stuff.
PPC Copy Session at SMX Advanced:
Mona Elesseily of Page Zero Search Marketing.
First understand your market by looking at the current ads. Then be different. For example, if 3/5 ads offer free shipping, test a different shipping offer. Or just avoid the topic.
2) MSN Labs’ search funnel tool lets you see the sequences of searches people do. Use that to feature the next/previous KW in common sequences. Mona also refers to the seasonality tool.
3) Spyfu is a nice tool; she uses it to gauge competitors’ ad spend. I’m not sure, but I expect that Spyfu ties into the AdWords Traffic Estimator Sandbox tool to get an idea of those ad budget numbers.
4) What motivates customers? Yada rehash … price, reassure them (official site, 24/7 support), use time sensitivity (deal ends soon).
5) Consider an industry’s feel – B2B v B2C, for example.
6) Consider the buying cycle and adapt your conversion accordingly. Move them to the next step in the cycle. E.g. research: Try this quiz. “Buy” KWs – move to sale. NICE TIP, original, impressed me right there!
7) Multi-Variate Tests: test finer details after AB testing the following Big Chunks:
- Different title call to action vs none
- Order of words/syntax
- www vs none
- Subdirectory vs none
- Order now vs order today
- Top line vs bottom line (e.g. alternate a phrase’s position; similar to word order tip)
FYI: If the MVT data is wonky, run the test longer.
From the audience, a question slipped in here (or at least, it did in my notes):
Q: Do you see variance in networks as to what copy works?
A: Yahoo Search Marketing (YSM) retail converts differently
Brian Kaminski of iProspect got up next.
He leads with some ok but bland advice on unique sales propositions, such as price, certifications, service, selection, reputation, new info deals… Then makes it interesting by pointing out that this will vary by keyword! Then leaves us hanging with no specific case studies or principles in how it might vary. Sigh.
Being from a hotshot firm lets you cite fancy studies and Brian tells us that 67% of search behaviour was driven from offline ‘stimuli’. So be consistent in on/offline voice.
Q: What’s the skillset/background you have in an ideal ad-writer ?
A: Math background (stats) and a creative mind.
Benu Aggarwal was alright and shared a case study. It was well-known stuff so I’ll move on.
David Szetela of Clix Marketing on Content Ads vs Search Ads
- Yell, don’t whisper. (Making me wonder: Is it possible to do this with upmarket brands? How can you shout besides color and animated ads?)
- Distract their attention
- Relevant ads are better – but this doesn’t happen automatically with AdWords’ algos
- Ads don’t need to relate to keyword ad text…
- Keywords are about the pages you want your ads on – largely about demographics and generic, all-encompassing (like top-category in a directory broad) keywords
- Test text v banners
- Use crazy headlines and giveaways
- Ease folks into the buying funnel since they’re not actively looking – deliver white papers and do lead gen.
For an ad to be successful, it must say:
- This ad’s for you (the target) – via connection with the page/subject matter
- Prequalify the click – e.g. luxury solution only wants clicks from big spenders
- Tell em what you want them to do on the lander.
- Telegraph a benefit
Content network quality score is mostly based on CTR and bid. Start high, go lower over time – the initial juice from your great ads should stick around.
The lower right corner of a banner ad is a good spot for a call to action
Video distracts attention.
Include a display URL. Doesn’t need to be the real destination.
More Q&A:
Q: What is the accuracy of a Spyfu?
A: Need not be 100% accurate – correlate it with other tools.
Q: Share keyword clues as to location in the buying cycle.
A: Problem statements indicate people looking for help. Then you’ve got make, models, features. Then brand terms, and then more make, models and feature keywords.
***
10 Terrible PPC Mistakes Panel – SMX West
The following seems to be from David Szetela’s (Clix PPC / SEM) presentation:
Dynamic Keyword Insertion: Don’t run DKI with campaigns bidding on TMs, misspellings or any broad matches. The latter get you made fun of a la eBay’s infamous “Find deals on Edward” copy…
Some content network sites don’t have an advertise-on-this site link – so how can you target them with placement targeting? Scrape the search results for those sites running adsense and paste the results into AdWords.
This was new to me at the time, but it seems lots of blogs (e.g. Diorex) have shared this.
When using broad match, your results will be better if the words have fewer synonyms.
I can personally attest to a related notion: bidding on acronyms can be really, really terrible. My dad runs a business selling dental CE courses / seminars and we bid on “AGD” – the Academy of General Dentistry. Turns out AGD is also a paintball gun maker, a game development company, a sorority, a crane hire company (hire is the UK word for rental) , etc. Words with multiple meanings (thus acronyms in particular) make terrible keywords!
Directly from my notes: “When describing pages for placement targeting, is “inanchor” useful?” – Answer is missing, so perhaps someone can elucidate the meaning? I imagine the idea was to find sites using the inanchor search…
As explained in this interview with Andrew Goodman and Amy Konefal, broad match has evolved to match any single word in the phrase you’re targeting. If you broad match ‘search engine optimization’ then your ads could show up for search resumes online, engine parts, and pc speed optimization.
Adi Connor (Director of SEM at Course Advisor): Account structure tips:
Base structure on business objectives and aim for lots of flexibility and efficiency. Pivot tables (a tool in Excel) can be useful for creating & templating [reports? campaigns?]. Good [account structure? pivot tables? both?] facilitate reporting. I think Adi meant good account structure.
Use modifiers in different groups to facilitate DKI.
Unfortunately Adi makes oversized presentations and blasts through them so that it’s hard to get good notes on her material. It’s quality, but the style of presentation is hard to follow.
Some miscellaneous Q&A
Q: What keyword should get credit? First? Last?
A: Study this with control groups. Some people use PPC as bookmarks, so heads up for that snag.
Q: How should PPC be tracked to phone calls?
A: At the keyword level.
Q: How do you deal with conflicting client-set metrics? EG Margin v volume?
A: Address before starting – define measures of success.
Also, share data regularly and relate it to agreed upon metrics.
Q from Googler: How do we help advertisers track based on margin?
A: Bid management tools.
A2: I’m tool agnostic – it depends on client needs. Look at media analytics tools (he means for managing offline-integrated campaigns).
Aside: Rely on attitudinal/survey data for the business cycle. B2B cycles are long – even 3 months. They know this using cookies. Need flexibility to differentiate between PPC conversions and repeat purchasers acquired from PPC.
Interesting metric for support sites: Call-avoidance.
You need sophisticated tech and individual tracking.
Big Management Today – For complementary notes, see Barry Schwartz’s coverage of Bid Management Today, which covers the slide material better.
Kevin Lee, Didit
Why it matters: See his slide with reasons.
Cherry pick the users that are most profitable. The winner’s curse is that the click-auction winner doesn’t know whether the click will be profitable. So: Bid management is there to compensate by helping to:
- Predict the value of an impression or click
- React accordingly.
Chris Zacharias – Omniture – Components of Bid Management
1) Data aggregation through APIs, analytics, internal tools
2) Initial Reporting User Interface – they still suck cuz management doesn’t see them, if I understood my notes right. Or else he meant this isn’t as important as the management console.
3) Forecasting algorithms to guess what the effect of doing what the system is suggesting will be/
- Important questions to ask relative to this are how often data modelling occurs and how often data modelling gets implemented. If I knew the English meaning of that jargon, these notes would be more helpful. Sorry.
4) Change match, geography, time, syndication options – all that should exist in the tool.
Data signals:
Most of what is tracked misses micro-conversions. Track adds-to-cart, services pageviews etc. Unfortunately these are not often in bid management tools – just web analytics tools. This is particularly helpful for low volume keywords. Additionally, if you cluster these keywords together, they’re more easily manageable.
See his slide on the types of management – rule, portfolio and algorithm based.
George Michie RKG (my notes said Chris, but I found no Chris Michie PPC expert online…)
- Tech requirements – see his slide
- Bid by value of traffic to you. Don’t bid by position.
- Get a system that is original, not rented. (Huh? Build your own? Really? Or only deal with firms that have their own?)
- There’s a tension between granularity of data and recency.
- If sending traffic to category pages, ask for item-level tracking to know exactly what you sold. On a related note, you want variable cookie windows depending on merchant needs.
- Backend feed capabilities for data input are also great to avoid lots of typing.
- Finally, get your resources, benchmarks and other ducks in a row before investing.
Bit of a shame George didn’t time his speech better before coming, because he obviously had many intelligent things to share. Here’s some more goodness. Can you tell I’m a fan?
Q: Should you build your own inhouse?
A: Kevin Lee (Didit, which has their own custom built tool): Costs several mil to do this. Less than 1/2 of online marketers who could justify it – eBay, Amazon, Overstock.
(Methinks that Kevin is exaggerating perhaps – a guy like XMCP could whup most devs asses when it comes to PPC, and from the little he’s shown me, he cleans up with his inhouse tools, and it wouldn’t take him millions worth of man hours. I do agree that most online marketers probably can’t justify it though. But again find it humorous that Kevin thinks 1/2 of online marketers could spend ‘millions’ to do it.
Caveat emptor: Kevin Lee is a PPC pro and I’m an organic search guy first and foremost. I’m probably making a fool of myself with the above statements, to people in the know, but still think I should put those feelings out there.
Aside: While clustering tail keywords is good, also segment head keywords by day of the week, time of day, geography etc.
p.s. Kevin was generous and gave away copies of his book The Eyes Have It, of which I got a copy. It’s an interesting argument, mostly directed towards CMO types who are new to search, as to why they should get into PPC. He makes a persuasive case. It’s not a how-to book for the most part, though, so know why you’re buying it if you do.
More miscellaneous notes:
- Engineers look at longtail keywords, see a low click volume and figure it’s a bad account. So perhaps put the least relevant ones in separate accounts to avoid tainting your better keywords.
- Quality Score is not on the domain name level yet.
- Clustering algorithms need to make sense for your particular business before you use them. The algos need to work to different goals within the same account.