So You’re Thinking About Attribution
Published: January 11, 2012
Author: Sean Marshall
So you’re thinking about rolling out an attribution model? If not, then you should be. Proper attribution is the quant jock’s way of dealing with the tangled web of online marketing. Traditional marketers have known it for years – it often takes more than one touch to get a customer to convert. What they never had, however, was the ability to measure when and how a customer interacted with their ads and content. Today’s online marketer has that ability. If you aren’t taking advantage, your peers (and competition) are passing you by.
First, you need to figure out if you’re looking for cross-channel or SEM only. If you’re serious about this, you should go cross-channel. For too long, SEMs have had blinders on – ignoring the influence of SEO and email on their campaigns. Outside of seeing a big spike when the weekly email blast goes out (and trying to take credit), 100% SEM credit was the norm for most of the last 10 years. The reality is that AdWords and 3rd-party management tools take too much credit. It’s not what SEMs want to hear, but the fact remains that these tools take credit for 100% of conversions when they might have only influenced a purchase. There are benefits to isolating search as a channel and understanding the behaviors between non-brand and brand clicks, but if you really want to be accurate, go cross-channel. This is particularly important if you have multiple paid sources. While integrating SEO data is interesting, it isn’t as actionable (at least in a short term view) as comparing two paid sources and adjusting the amount you bid/pay on the spot.
Google Analytics and the multi-channel view is a good starting point. As with most new GA tools, it’s a nice addition and, given the price, a no-brainer. That said, you aren’t going to get the level of detail, or usability, you would get from more sophisticated solutions. GA really only allows you to process this data at a channel level. The problem with that is that you can’t distinguish between sub-categories within a channel to the extent you need to be effective (i.e. search can’t easily be broken down to search vs. content, brand vs. non-brand, text vs. display for content, etc.). Each channel has its own complexities you need to take advantage of.
On the PPC side, most marketers understand the dynamics between brand, non-brand, the engine, and the contextual networks. Content traffic, in general, carries far less intent than search traffic. For years, people would look at content as the red-headed stepchild of search. Performance was poor, conversion rates were low – it was a bust. Problem was that no one was thinking about its role as an introducer and influencer during the purchasing cycle. In a system that credits the last click, the brand campaign is king. In the same system, content is a peasant.
Imagine a sneaker-phile is browsing his favorite sneakerhead website looking for news and upcoming releases. Now imagine you sell the type of sneakers he’s into…but he’s never heard of you. With the right copy, you can catch his eye and draw him to your site. The problem is that this person probably wasn’t looking to buy that day, but your site probably made an impression. A few days pass. Same user searches for your brand, clicks your ad (yes, buy your brand!), and converts. Content gets shut out, brand steals all the credit. No wonder content looks like a dog…
Another thing to note here is that attribution models are most valuable when dealing with multi-step or considered purchases. If less than 90% of your conversions happen the same day as the first click, you need attribution. If not, you might be able to do without it, but eventually, it’s best to roll it out.
You’ll often hear clicks in a conversion path referred to as introducers, influencers, and closers. These are just first, middle, and last touch points, respectively. In a system like GA, or an SEM-centric solution offered by tools like SearchForce, you’ll need to work with proxies to figure out how to shift credit. These systems aren’t able to distribute credit for each individual transaction, so you need to get a rough sense of how different important scenarios play out.
For example, if you find that 30% of your brand conversions were introduced by a non-brand click, you need to create proxies and shift the credit back to your non-brand campaigns. In effect, take the volume generated by these influenced brand conversions and use it to recalculate your non-brand ROAS and CPA targets. Have a 200% ROAS goal for brand conversions that happen to equal double the non-brand conversions? Perhaps you should shoot for 100% ROAS and anticipate double credit after you account for the assisted brand conversions. This only gets more complicated if you factor in latency. As these models should be deployed for programs where conversions don’t happen the same day, you have to wait to see how things pan out. Proxies are a good stop-gap but not a perfect solution.
If you really want to take it up a notch, I’d recommend a tool like Convertro. As with most marketing tech companies, these guys are specialists. A tool like this will enable you shift credit for any transaction depending on how different channels interact. The beauty is that you can do this at the most granular level – knowing exactly what KW and match type gets credit. Outside of granularity, customization of attribution is the star. Want to give organic clicks zero credit if they interact with paid? Done. Notice affiliates influencing a conversion 10 seconds before purchase? Arm yourself with data and push back on them.
Now, I can’t get into all the details of a tool like this, but I highly recommend looking into it. I think digging into it could give you some insights into how to best balance your model. Until then, a few final recommendations:
1) Shift credit away from organic search if it interacts with paid – you can’t just magically push position on organic searches, so move the credit where you can make a change.
2) Same goes for paid brand search – you’re likely dominating first position, so what good does taking credit away from non-brand do? If anything, you’ll be able to push non-brand head terms more aggressively and dominate auctions – leaving your competitors asking just how you can afford to stay in position 1.
3) Favor the introducer. Influencer clicks are important, but you should generally assume that the first click introduced the user to your brand in the first place. Might not always be the case, but under that assumption, you’ll be able to drive awareness AND ROI from top-of-the-funnel channels like display and Facebook advertising.
4) Go with longer cookie windows – most often, people cap SEM cookies at 30-60 days because they don’t want the channel to take credit for actions taken long after clicks. Given that you’re shifting credit properly, let that cookie go as long as possible. Your model will ensure the right channel gets credit at the right time.
– Sean Marshall, Director of Search Engine Marketing
– Questions? Comments? Email us at blog at ppcassociates dot com.