Internet Marketing Business Track: Demystifying Online Attribution

This is a guest liveblog post by Alan Bleiweiss, Internet marketer and blogger at Search Marketing Wisdom. He focuses on providing SEO site audits and action plans, including follow-up consulting to aid developers, designers and content specialists implement SEO action plans.

This is going to be a great session because not only do I get to liveblog, but I went to the restroom BEFORE the session this time – live and learn right?

No – seriously I’m looking forward to this session because I’ve recently been heavily involved in consulting for an agency client on the issues site owners face in proper online attribution.  So I’m hoping that not only will I be able to provide a quality liveblog record of the session, but that I might even learn a thing or two along the way myself.

We’ll see.

Here we go!

Moderator: Vanessa Fox, Contributing Editor, Search Engine Land

Q&A Moderator: Laura Lippay, Director of Technical Marketing, Yahoo!

Speakers:

Jonathon Colman, Internet Marketing Program Manager, REI
Cameron Cowan, Product Manager, Omniture, An Adobe company
Dennis Goedegebuure, In-House SEO Manager, eBay, Inc.
Richard Zwicky, CEO, Enquisite (now called Eightfold Logic)

Vanessa’s opening the session, welcomes everyone, and now a word from our sponsor:

It’s Brian Bowman with Reply.com [insert the usual marketing spin here :) ]

Vanessa: How many of you do online attribution?

A smattering of people raise their hands.

Vanessa’s now introducing Laura Lippay and explaining about asking questions online for this session, as well as tweeting, and as crazy as it sounds, raising your hands right here in the session!

First up on the panel is Richard Zwicky.

He works with clients like Razorfish and Microsoft.

It all boils down to one salient point – every day, millions of potential customers are telling you exactly what they want but you are not listening.

If you haven’t read Vanessa’s book, you really should.

Attribution starts with findability – the path from customers to you – it’s about visibility wherever they are.  And it’s about relevance.  Attribution is all about understanding it all.  (He’s now quoting Guy Kawasaki, so I’ll opt to not help Guy out.)

You want to speak to your customer – a series of referrals over time builds confidence and engagement.

Attribution – understanding the whole series of events that lead up to a conversion.  What was the series of conversations?

Traditional offline and online channels contribute together.

Offline is not easy to measure.

Online is completely measurable.  It’s about the individual finding you. It’s about measuring people.  How they heard of you, when you were relevant.

You can’t manage what you can’t measure.

Its how they fit together – there’s a combination that leads to a sale.

He throws up a slide with a 10 line chunk of code from a raw log file.  I say (out loud) I’m not going to live blog that – Richard responds – are you sure? It’s going to be up a few times… – Yeah, uh, NO! [Blogger-speaker interaction! I like that! —Virginia]

By looking at the raw code of a visit you can learn where they come from in a given session – and you can gather this for each session over time, and multiple domains and multiple sub-domains.

Kayak has multiple sub-domains and they need to be able to understand that cycle.

Core tracking data:

• Pull the data, push it into a text file and pull it into Excel.
• Referrer info might have changed but the persistent and network cookies remain the same if they come from one of a group of related sites or sources.

Richard is now getting into very deep technical process here – Vanessa says the slide deck will be available after the session – which is good for you because I just can’t keep up with Richard’s rapid fire preso. (OMG fast & technical!)

Essentially it’s about being able to track one visitor who comes back over and over through various paths and to see the pattern of that multi-visit process.

Tracking can involve cross-domain and sub-domain visits. Example – they might come in from more than one micro-site over time and end up in a single sale.

Once you’re properly optimized you can properly track activity.

Where did they come from on the Web? Where did they come from social? You can see where the overlap is when you’re looking at the data. You can then tie dollar value to the transaction.

This is different than “last-click” attribution and lets you apply your budget where it’s most effective across the overlapping channels. Or else you’re investing in fragmented marketing strategies and missing most of the opportunities.

Our business is to improve our customer acquisition rate and we want to invest where the most value is, and attribution allows us to do that.

Vanessa:  How many of you use Ominture?  Several hands go up

Up next is Cameron Cowan.

He doesn’t promise us any answers.

What is attribution?  Who gets the credit?

Online acquisitions – Search marketing – SEO, Social, display advertising…

And once they get to your Web sites, its tracking what happens on your site.

But we’re also talking about offline advertising and what conversions happen offline.

He says a consumer may go to BestBuy.com to find out where the closest store was or if a product was in stock.  But he’s a tactile guy and wants to see what those headphones feel like and sound like.

Things we overlook about attribution:

• First click
• Linear allocation
• Decaying allocation
• Reverse decay allocation
• U-shaped allocation

The above are all ways to map the multi-touch point of a user from initial contact to final conversion, then assigning weight either earlier on in the process, closer to the middle, later on, or some other mix.

Other things to track:

Duration – cookie expiration – how long do you keep cookies live in the attribution cycle?  Is it 30 days or 2 years?  Is it the right time frame for your business?

What if the first click was a month ago? Do I give the same value to the first click?  Am I going to assign the same amount of credit to each click?

Are you even looking at this data?

• Marketing Mix
• Banners
• Paid
• Organic
• Mobile
• Social
• Affiliates
• E-mail
• Offline

If you’re just doing Search in a funnel, that’s fine, but understand that’s a choice and impacts your ability to evaluate. What marketing mix am I going to be tracking?

• Depth of Engagement
• Impressions
• Clicks
• Product Views
• Cart Additions
• Checkouts
• Orders
• Revenue
• Units

Did they sign up for a newsletter? Did they come in, check it out, and then leave because they had to go ask their wife before they made a purchase?

Does an impression deserve the same credit? Or any?

Are you making the right decisions for your business?

Creating an attribution model for your business can be really challenging, so start simple and evolve from there.

Doing something is better than waiting until you can do everything.

Does it apply to you?

How many orders happen on first visit vs. multiple visits? If 99 percent happen on the first visit, you don’t have a problem and don’t need to study or track multiple touch point conversions.

Online retail example of multiple visit iterations and considerations
Which attribution should get credit? First Touch? Last Touch? Linear Attribution? Latency? Should latency affect weighting of credit?

Understand the points of contact and whether they have value other than conversion – the consumer experience and how you can influence them along the way – these are factors to consider:

• Awareness / branding
• Consideration
• Conversion
• Retention

He’s not answering these – they’re questions you should be asking.

A lot of people think a lead is the end of it. Why can’t you understand how coupons online are leading to in store sales? There are solutions that allow this type of analysis.

He’s now putting up 4 pictures asking what is common to these 4 pictures? Answers from audience: they’re four corners of buildings; they’re all storefronts; they’re all in the U.K. He says: they’are all buildings on the same street intersection – the four corners….

Learn to ask the right questions.

Now it’s Jonathon Colman from REI (the camping gear retailer), @jcolman.

Cross program attribution: We’ve talked about all the difficulties, so many questions and few answers so we conducted a test.

Marketing program that operates in a silo:

• Doesn’t allow sharing customer insights tools and reporting
• Cannibalization of results, diminished impact of the whole
• Leads to duplicate incomplete work and unproductive competition

Giving credit where it’s due:

• Last click analysis lessens cross-program collaboration, effectiveness
• Delays insights into customer behavior

Advertising costs:

• No budget can support CPC’s that increase without limit
• Takes budget from other programs, new opportunities

When we don’t have a good picture of attributions we take away from programs that could be more effective.

By integrating search marketing programs you:

• Drive sustainable traffic and sales
• Gain insight into customer behavior
• Save budget for other efforts

Test for efficiency in paid placements. Test for priorities in non-paid optimization.

Test 1: Correlation between paid and natural search

For different keywords they invested a lot more incrementally over time. He expected to see a massive increase in results over time. He expected no change in natural search over time with the increase in paid search.

They found a direct correlation increase in natural search slightly after the investment in paid search – as the paid search rose, so did the natural, and as the paid was reduced, the organic reduced, yet the natural search results didn’t drop as much as expected.

This could be due to several reasons so next they needed to understand what the cause of that legacy natural search value is.

With this kind of data you can leverage internal political pull of the paid search program in order to get more natural search work done.

Test 2: Subtractive placement in paid search

REI owns 5 out of 10 placements in the SERP for the test phrase – #1 paid, top 2 organic, and 2 of the 5 product results. They wanted to find out if they had to be in all five. Could they save money by reducing maximum CPC? Could they increase traffic in non-paid programs?

While they saved money, they saw a significant drop in conversion when only using non-paid.

Additional testing: test a greater variety of terms, test additional levels of cost reduction – maybe you don’t eliminate the paid ad, experiment with different bids – the more you experiment the more accurate your results.

Test 3: Additive placement

We all know about the value of having more results show up on the first page. What if you could create more efficiency and save money by optimizing more on the organic side? Will it increase natural traffic and decrease costs of paid search?

Results are pending, but numerous studies say to should expect 53 percent increase in total clicks when appearing in top paid and natural results.

Next steps:
Test seasonal terms, test messaging of high conversion paid search for use in natural search content. Test ROAS (return on advertising spending) of adding paid search ads for high performing natural search.

Integrate search marketing programs:

If you do all these things, impress your boss, take a long weekend and go camping! (LOL what a plug!)

Vanessa mentions a lot of this type data is available in her book.

Next up is Dennis Goedegebuure, @TheNextCorner, in house SEO manager at eBay.

We’ve had a lot of discussions at eBay about attribution.

He asks “Who doesn’t know what eBay is?” Oddly nobody raises their hand!

eBay makes products available through 22 sites. (Wow – really – I honestly had no clue!)

We roll up revenue based on location/country but buyers are coming from multiple countries and different channels.

Dennis puts up a great infographic here – “iPad Exports from the US: Where do all these iPads go?”

At this point Dennis provides examples of different ways buyers come to eBay, bid on products, and then from there, how each touch point gets assigned attribution based on several combinations of factors. It’s fascinating.

Except this is a liveblog and there’s no way I can properly explain this without writing a university Thesis. So I’ll just synopsize and hope you grasp the basic concept by the time I’m done. If not, blame Susan for allowing me to liveblog. :)

US based buyer: bids on the iPad, wins the bid. Direct attribution of winning bid to last click.

What if person clicks on paid or organic search results, but then clicks on a display ad on the site, then buys it? Attribution goes to first click and last click.

UK based buyer: bids on iPad. Attribution is direct on the click but the revenue is in the U.S. (This is where their data tracking gets really complex!)

If a UAE buyer sees the auction, and bids, same thing happens – attribution to the last click, but revenue to U.S. – and UAE buyers generally bid hundreds of dollars higher.

If the UAE buyer bids after the UK bidder, eBay gives the UK bid attribution as well (the U.K. bidder starts the process and motivates the UAE bidder to bid higher so they get attribution credit).

He’s now going through variations on the bidding process – and eBay applies different attribution models accordingly. And my head is already spinning with the permutation possibilities!

Cross-country correlation is important to determine effectiveness of acquisition.

Second-to-last-click correlation is important to determine effectiveness of cross-channel buyer acquisition.

These two combined make for interesting investment opportunities.

It’s Q&A time.

Question: Generally, what kind of return are you going to get investing in a sophisticated allocation model?

Cameron: There are no generalities. If you can squeeze out a few thousand dollars, for some business owners, that’s enough. Larger businesses need to see a higher return.

Jonathon: It depends on your business, and who your customers are. It’s a question of how much you want to miss out on in terms of sales and traffic.

Question: Can you talk about attribution with mobile and specifically how Omniture is handling this?

Cameron: As people become more diversified with their online experience how do you track that? As far as I know, nobody has come up with an ideal answer. But you should track it down to the user experience and Omniture allows that.

Dennis: For us it’s a bit easier because we have a sign-in so we can track a user.

Richard: A lot of mobile goes on within apps like Yelp – and you have to treat those referrals different than a general search referral. They may have done research before that session though.

Vanessa: Urban Spoon says 99% of their visitors came through their iPhone app.

Question: Secure protocol (https) doesn’t allow this kind of tracking. With the new Google ability to search with https have you tested with these to find out how many people convert, and can you apply attribution?

Vanessa: It shouldn’t affect your analytics that much – very few people use it.

Richard: I think the effect will be negligible. Like Search Wiki – nothing came through that. Our logging script accommodates both http and https.

Cameron: All our testing so far it has a negligible effect.

Vanessa: Google had to develop Universal Search because nobody knew about the tabs at the top of the results, so few people will use https search.

Question: If attribution is harder because someone uses multiple browsers, cookie blocking, cookie deleting, is that harder to assign tracking?

Dennis: at eBay we use a guest login and a member login – cookies are required so we can track it.

Richard: User login resolves that. We collect tens of millions of referrers – we see 3 to 4 percent with JavaScript disabled so it’s not significant enough to impact our overall results.

Vanessa: You can track the IP address.

Question: Jonathon, for your first test, was it the same keyword sets or for brand terms?

Jonathon: We saw a lift in both.

Follow-up question: Any correlation for same keyword on both paid and natural?

Jonathon: Look at how many keywords, the size of your ad groups. But with a larger base you see a higher lift on brand terms.

Me (Yes, I actually asked a question – and look – I’m liveblogging it as it happens! Either I’m a liveblogging superstar, or I’ve just cloned myself): What about call tracking?

Dennis: We did a test with Skype. And we sold Skype (uh, I’m not sure about what that is supposed to mean, but Vanessa has a good laugh…)

Cameron: We’ve been tracking calls since 2006.

Richard: You have to have call tracking, track coupons that someone prints out and uses in the store, all of that offline conversion data – you have to be able to associate all of it. It’s crucial.

And that’s it! We’re out. Now I get to run downstairs to liveblog You&A with Matt Cutts. Because I’m a liveblogger now and my physical well being no longer matters! :)

[Hee. Alan is a fast learner! ;) —Virginia]