Why Does Search Get All The Credit? – SES London
This panel was particularly interesting to me as I work for a large media agency (referenced in both these talks as a potential for lack of coordination and multiple tracking tools) and also because quite recently we have been working on dealing with some of these very attribution issues across the team – so a lot of these issues were all too familiar to me.
We had Nick Beck of Tug as well as Paul Mead of VCCP search talking about a good number of issues with attribution. Essentially the panel was formed on the back end of a meeting Nick had with a client in which they found themselves facing an issue: clearly 200+200+200+200 != 200. As a result of working across a number of agencies it is far too common that everyone tries to claim the conversion based upon different tracking methods and the result is a false reporting figure that far exceeds actual revenue figures for the client.
Nick Beck – (Tug)
He also though that surely he heard this 5 years ago, we must’ve moved on by now?
However, obviously tracking and attribution is still NOT being done right and he felt obligated to speak about the concepts and tools to try to help clear the air and demystify this a bit.
-standard cookie on click
-tracking pixel on confirmation page
-capture conversions and revenue – optimise for ROI
-set up goals through GA
-analyse & adjust campaign over time
Similar to PPC (cookies)
Tracking pixel on confirmation page
Capture conversions or other KPIs
Integrated media – this is what we all want and all agencies aim to do. But which click wins!?
1999 was the year of integrated marketing, now we’re still talking about it because it still hasn’t happened and there are ever more channels as we’ve discussed to worry about.
Sometimes first click wins, but usually last click wins – and that’s usually PPC!
Where tracking goes wrong?
100% attribution for one and only one click
No path to purchase, no measurement of other impacts
Results are thus being over reported and the actual channels being under/over estimated.
This makes it extraordinarily difficult to estimate results and plan for budgets?
Why does this happen:
BAD TRACKING – there are a handful of other causes but ultimately it comes down to bad tacking in most cases.
What you really need is good tracking- but using multpile technologies and multiple agencies see sales attributed to multiple chanels and this is why so many people struggle with this and why so many people are increasingly interested in “holistic” marketing solutions.
Reporting would still only based on last click, but at least that way you won’t over report
Use one tool and one pixel for tracking.
You need historical data to implement this successfully. With this data you can then sit down and analyse results and the click path to conversion meaningfully. You can then give a percentage weighting based on this. Without this historical data you are really just putting a finger in the air and not really providing meaningful predictions.
Milk & More –
Background – PPC and SEO since 2009
All display run through a different big media agencies. One using DART and one using Atlas.
Floodlight tag generated in DFA (Double Click)
All display shifted to DFA in two days
Measuring it right is the first step and ultimately the key to fixing these problems with attribution.
What of the fragmenting landscape?
Nick showed us a great chart showing the different number of sources of possible attribution then (19880s) vs. now and the massive number of additional areas trying to claim credit for sales. The newest to join the list are a growing number of social media sites and has even led to some (Facebook) pulling back some of their reporting data due to the many resulting complications.
Social for a lot of companies is still a box ticking exercise. The KPIs are horrible but there are a lot of brands out there who don’t even seem to care. We will need to start figuring out how to actually account for this but at the moment it is just not doing it properly. In the meantime, Nick suggested the following as starting points:
- Use GA to understand traffic impact.
- Use sentiment tools (Sysomos & Radian6)
- Third party software (Kenshoo, Searchignite)
Paul Mead (VCCP Search)
According to Paul and his experience from a PPC background everything used to be about last-click wins, but now attribution is a huge issue. Paul likened the issues with attribution to the meaning of life: it touches all of us but we can get by on a day to day basis without thinking about it too much.
Everyone is coming from a different background and has a vested interest.
Perhaps the most interesting point that Paul made about all of the struggles to make attribution work is that real truth is unattainable. Everything in attribution modelling is a kind of a fudge but it can be profitable and insightful.
Paul talked about a number of areas of concern that should always concern a client if they see them in a report (or you if another agency is producing such data):
- Pretty Pictures
- Unnecessary Complexity (media mix and econometrics). Really dangerous because they can never be questioned or interrogated.
Things Paul suggest to consider before leaving last click attribution model:
1. Different tracking system (digital media, affiliates, PPC (bid management), “non attributed” and organics)
Tracking system can still have problems (one have all paid for and one has direct – GA)
2. If attribution is the answer, what was the question? What specifically do we want to answer – Does PPC cannibalise SEO, does display drive search, are my voucher code partners parasitic, are my affiliates delivering incremental sales against my PPC, what % of brand searches are navigational
3. One SERP, 4 ‘channels’ PPC, SEO, Aggregators, Affiliates. Users clicking in and out of all these different silos. It’s actually the same group of people often, not all uniques.
4. Search and display – display drives search. In some cases there is absolutely no value.
Practical things to try:
- Single tracking solution on all channels
- Automated deduping of sales through conditional tag calls
- Address any significant revenue leakage (e.g. call tracking)
- Extend cookie window to look at average touchpoints to sale
- Analyse mix of new versus existing users (e.g. brand search) – tag the login. If a brand visitor immediately logs in they are likely just navigational
- Set up control tests for display and Email
Insights first, change later!
Attribution Weighting Models
If everything comes through one channel you don’t need weighting! However, if you do there are a number of complications and you need to be very careful when looking at active touch points versus passive.
Example Scoring system:
Direct click on a banner – 100
Nonbrand SEO click – 100
nonbrand PPC – 100
Facebook clicks – 20
Brand SEO – 10
Brand PPC – 10
Impression on a banner – 2
Weighting this sort of thing can help you account for active versus passive impacts and really attribute value and budget more appropriately. However, obviously this doesn’t do anything to address the tracking issues and would require a functioning single tracking system to attribute sales properly to the different channels.
Paul Closed the session with a handful of sweeping generalisations that he suggested we test if any of these apply to us. As an SEO I take personal issue with a few of these (disagree to an extent) but I still think they’re interesting insights and worth testing.
Sweeping generalisations: (test if they apply to you)
Display works on a longer cookie period
SEO cannibalises PPC more than PPC cannibalises SEO
PPC users click around, SEO users don’t
Affiliates and voucher codes can be very parasites
Brand PPC can be navigational and where no competition exists not incremental to SEO
Often not much of a purchase cycle in search
What we don’t see (a clear pattern) is a big part of what makes this difficult. How many people start at work, go and do it again at home. There are loads of gaps and behaviour is actually a lot more random than you would have thought in Paul’s experience.
Understand the limitations of attribution
Think about the specific questions you want to answer
Lots to do before touching last click
That’s it for me today! We’ve got more coverage coming to you tomorrow and some really great sessions. For those at the conference enjoy the party this evening and enjoy the ongoing coverage from State of Search.