Credit where Credit is due: Demystifying Attribution
This is a write up of the panel session “Credit where Credit is due: Demystifying Attribution” at SMX London 2011.
Chris started this session by discussing the value of the keyword in terms of conversion rate and sales. This is a discussion that often uses the last click wins mentality when calculating ROI values. Attribution is often far more complex than this and this session aimed to clarify many approaches.
The following are the top tips from each presentation:
Cameron Cowan – Product Manager Adobe
1. Attribution is more than just the allocation model. Last click attribution is the most commonly used but misses multi-channel touch points.
2. Attribution timing is vitally important, understanding the length of cookies used is very important. 30 days is default but it is often way too short and can skew an attribution model. Tie your cookie duration into your sales cycle.
3. Regency and Latency is important to understand and implement in an attribution model. Does the length of time in between touchpoints affect the value of that touchpoint?
4. Marketing channels need to be tracked in order to create an accurate attribution model. Email marketing is often overlooked.
5. Depth of Engagement is very important, bounce rate is often the most overlooked metric as your website hasn’t answered the question that the searcher asked.
6. There is no right model for attribution, it’s the perspective that you want that should determine what method you use and this should be tailored to your business.
David Sprinkle, VP of Analytics, ROI Labs
Benefits of Good attribution
1. Consistency across channels, SEO gets the raw deal most of the time
2. Limits dominance of brand keywords
3. Values the whole buying cycle
4. Improves budgeting and ROI
3 things to remember
1) Single source of truth – ensure that you’re measuring all sources on the same tool. PPC tools often don’t have multi source attribution, analytics does.
2) You can’t analyse what you can’t track
3) Make an informed decision
Pros and cons of each model:
1) Favours direct channels
2) Emphasises brand terms
1) Favours awareness channels
2) Good for keyword research
3) How long is too long?
1) Don’t recommend this, value diminishes more as more sources are added
1) Requires judgement calls and guesswork
2) Few tools do this on the fly
1) Weighted by the level of time passed
2) Sophisticated and accurate but it’s very hard to code
3) Usually requires manual analysis
How long is too long? Track your visits until purchase and this will help guide cookie duration. This should be measured from source and overall as it will differ.
Taking the stress out
1) Choose your tool wisely
2) Give yourself options
3) Translate analysis into optimisation targets
You can manage what you can’t measure!
1) Track both first and last click
2) Include all sources
3) Use Campaign stacking, appends all the channels and allows great data for deep analysis
4) Analyse time to convert
3rd party tools data will vary but you can get them closer by:
1) Mirroring cookie expirations
2) Capturing transaction ids to avoid duplicates
Jeff Greenfield , COO, C3 Metrics
Using data to make smarter decisions, the main issue is that we are usually drowning in data but still thirsting for knowledge.
The porn industry set the standard of the 30 day click window. Most other business units have nothing to do with studies; they just adopted the norms even though they aren’t relevant.
Last click is flawed, is it right to reward only the last relay runner? Re-targeting often has the best conversion rate but it’s the last relay runner which depends on the first 3 to get them in the game.
Each client and campaign is different, data is collected and analysed before assigning values. Output from the analysis should drive meaningful change.
Craig MacDonald, CMO, Covario
There is a simpler way to get to the answers that you need. Looking at things at a higher level allows you to respond quicker.
There is a great deal of doubling counting. If you added up all the self-reporting from your different agencies you’d be making 5X the revenue that you actually are.
Digital only drives 15% of Media Spend according to Forrester but the majority of spend is still mostly offline.
A different model is Multivariate Regression Approach, it maps improvements in ROI based on the changes in budget spend across all channels. You can implement this in your PPC campaigns by grouping keywords into the various sections of the funnel altering budgets and measuring outcomes.
This model also allows you to factor in lifetime value of customers and how the traffic generation efforts are generating social media awareness and remarketing efforts later on.