{{ searchResult.published_at | date:'d MMMM yyyy' }}

Loading ...
Loading ...

Enter a search term such as “mobile analytics” or browse our content using the filters above.

No_results

That’s not only a poor Scrabble score but we also couldn’t find any results matching “”.
Check your spelling or try broadening your search.

Logo_distressed

Sorry about this, there is a problem with our search at the moment.
Please try again later.

Imagine you’re a football manager, your objective is to maximise the number of goals your team scores for your budget.

However, if you were to start assessing players based solely on their goal scoring ability, you’d end up with an unbalanced team full of strikers and probably let in a lot more goals than you’d manage to score.

The same story goes for cross channel media buys; each channel has a unique position on the team and serves a unique purpose, however they’re often measured on their ability to achieve the end goal independently of the other channels.

Buying your media in this way is attractive because it uses simple metrics and makes for great reporting, but if you’re still measuring prospecting based on your conversion pixel, you could be missing a lot of opportunities from further up the path.

As a simple example we looked at a client which provides us information about the channel each conversion was attributed to.

Using this data we examined how likely an ad is to get attributed depending on its position in the conversion path using a last click attribution model.

The results are hardly surprising.

Our takeaway was that even if the early path activity generates a lot of interest by making the user aware of the brand and potentially making them search around for it, the channel is placed at a disadvantage by being measured on only those conversions which weren’t somehow picked up by all the other channels available closer to the conversion.

This kind of measurement is similar to valuing a defender based upon their goal scoring capabilities.

However, if you were to measure all channels according to the part they play, rather than their ability to generate conversions by themselves, you’d get much better clarity into the actual effect of the advert irrespective of how close to conversion it occurs.

You’d also find out that brand spend can be measured against and optimised towards interim metrics such as digital GRPs, brand search uplift and a myriad of other easily recordable actions.

Considering the path in this way allows you to start valuing each stage in the path to conversion as part of a larger whole, where media isn’t bought in spite of your attribution model and different channels start to work together as a team to fill various sales functions instead of crowding round the conversion event.

All this leads to greater ROIs and a larger market share, but it does require some thought to pull off. Below I’ve outlined a couple of questions which come to mind when planning a buy like this, it’s not exhaustive so please feel free to comment below if you want to discuss further.

Do I need to change my campaign structure?

Campaigns are already set up to match user purchase intent based upon classical definitions of the user’s path to conversion; ever notice how well the brand, prospecting, retargeting structure matches the AIDA model of purchase behaviour?

At each stage of their path a user exhibits different behaviours, little indicators that your advertising is working before the conversion event occurs. Using these metrics to help measure a campaign means that brand and prospecting spend can be measured and optimised towards the actions they were set up to generate.

Can I track multi-stage conversions?

Yes. Using pixels which are currently in place on your sites you can radically improve the measurement of your campaigns, potentially saving many thousands of pounds in media spend which favours last click models.

The use of multi-stage conversions raises an additional question of double counting as you measure effectiveness based upon new metrics in addition to conversions at the end of the funnel.

This can be counteracted by simply assigning a small proportion of the conversion to the events which preceded it, allowing you to set targets which relate back to your overall goal to all stages in the path.

This is a small step towards the creation of a more accurate and sophisticated attribution model which would ultimately assign a value to each touch point based upon the actions it contributed towards.

However simplified, this model has the potential to open up large portions of your media spend to optimised buying techniques which were previously not even considered for early path channels.
 
If you’re struggling with how to measure your early path campaigns, first think about what they’re trying to achieve and start from there, you might be surprised by the money you’re wasting.
 
Try to remember that your different media channels are part of a team, working together to score that all important goal.

Considering each one in the same way as the others overlooks the part they play in that team and could lead to unbalanced media buying and slowly eroding margins as your pool of cookies eventually deletes itself.

If you think outside the penalty box you’ll see there’s a large part of the game that you could be missing.

Harley Norrgren

Published 12 March, 2013 by Harley Norrgren

Harley Norrgren is Head of Analytic Products at Infectious Media and a contributor to Econsultancy. 

1 more post from this author

Comments (11)

Comment
No-profile-pic
Save or Cancel
Avatar-blank-50x50

AZ

Very good point. I've heard lots of discussions about this topic previously but still think it can't be stressed less. Especially for e-commerce site, I can hardly expect a customer walks through the full purchasing journey the first time they click through an ad, as much as 'impulse purchase' is concerned here.

over 3 years ago

Harley Norrgren

Harley Norrgren, Head of Analytic Products at Infectious Media Ltd

Hi AZ,
This is very true, and despite diffrent length paths to conversion depending on the product, I still think that there's a case for more advanced measurement throughout all types of media buy.

Can you tell me how your own experience has been in helping advertisers to adopt more complicated attribution measures?

over 3 years ago

Avatar-blank-50x50

David Jarvis aka DJ, Business Director, London at cxpartners

There should be more use of football (soccer?) analogies within digital marketing. That is all.

over 3 years ago

Avatar-blank-50x50

AZ

Hi Harley,
No solution is perfect, but I always advise clients to use Google multichannel funnel as a start, which is a pretty good tool to relate various traffic sources to conversion goals.

over 3 years ago

Avatar-blank-50x50

Peter O'Neill

I am sorry but it is not possible to reliably track multi stage conversions due to the use of multiple devices and people deleting cookies. I agree on the issue with last click, you can't give credit to only players who scored the goal. But the data will always have holes in it - like you can only see parts of the football ground.

I gave a presentation last year at eMetrics on this problem, going on to describe the real questions being asked by business and my recommended solution - which is not alternative attribution models. As David should appreciate, the entire presentation uses a football analogy.

http://www.l3analytics.com/2012/12/16/campaign-attribution-is-broken/

Cheers

Peter

over 3 years ago

Harley Norrgren

Harley Norrgren, Head of Analytic Products at Infectious Media Ltd

Hi Peter,
Thanks for the comment!

I had a flick through your slides and I definitely agree with you, there are some serious holes created by cookie deletion and multi device usage that cannot be ignored.

It's hard to pick up from your slides how you propose to define new KPIs/ways to measure them to combat this. Could you please explain further for me?

I think that multivariate Y, i.e. interim metrics, in an attribution model can increase its ability to combat cookie deletion by decreasing the time between an ad exposure and the event it's trying to measure. Are you saying that last click is preferable to any other solution whatsoever, despite the marginal gains offered by more advanced methods? If so can you demonstrate that the bias introduced by cookie deletion is large enough to substantially undermine any attempt at monitoring multiple exposures?

Other quick thoughts are that if you have a conditional probability of a user having deleted their cookies given the number of observations you have against them, you can set your attribution model to only include those users who pass a particular confidence threshold. I think there would be bias questions here too, what are your thoughts?

Thanks
-Harley

over 3 years ago

Avatar-blank-50x50

Peter O'Neill

Hi Harley,
First, my bigger concern is ppl using multiple devices, not deleting cookies. But you do get the same end result in that it appears to be a new visitor.

My simple example of how attribution can lead to incorrect strategies is as follows:

-> a website has a high proportion of visitors researching before purchase - 80%
-> 75% of these researchers do so on their work computer before purchasing at home (first visit on that device)
-> any web analytics tool (assuming the visitors were not forced to login in order to research) would display 80% of purchases occurred on the first visit
-> the business strategy would be based on converting visitors on their first visit (as that is what the data informs is visitor behaviour) with this being a sub optimal strategy (if true behaviour was known)

A key point with the multiple device/cookie deletion issue is that you cannot even calculate the scale of the problem based on data, unless all visitor login. If you have one observation on a visitor, there is no way of knowing if they only visited once or if they are simply using a different device.

You have described how visitors would display different behaviour at each stage of their path towards conversion. My approach is to define the objectives for each campaign based on the behaviour they are intended to generate - different campaigns target different stages. Then to evaluate each visit individually - did that visit meet the objectives for the campaign that drove the visit? This might be viewed as the same thing as last click attribution but it is not, it is a different approach.

Clearly I could write a lot more detail here but going to stop now - have I answered your initial questions?

over 3 years ago

Avatar-blank-50x50

David Jarvis aka DJ, Business Director, London at cxpartners

It's wrong to assume data analysis for digital marketing activity is the solution to poor performance.
It's also wrong to assume that not doing data analysis will lead to better results.

So where do we go from here?

over 3 years ago

Harley Norrgren

Harley Norrgren, Head of Analytic Products at Infectious Media Ltd

Hi Peter
It seems then that we are in complete agreement.

How far have you gotten in estimating cookie deletion/ multi-device usage rates amongst users, and what techniques have you tried. We've been exploring the area for a short while, as it's implications lie far bayond the realms of attribution, and would be interested in getting your thoughts.

Thanks again
-Harley

over 3 years ago

Harley Norrgren

Harley Norrgren, Head of Analytic Products at Infectious Media Ltd

Hi David,
Thanks for your question. In my opinion it's about trying to understand the limitations of your measurement system with a view to incrementally increasing it's accuracy; this will ultimately allow you to optimise towards something at least a little more accurate.

I don't see data analysis as a replacement for performance, rather a tool that works alongside to help you unlock a little more from your campaigns.

What are your thoughts?

Thanks
-Harley

over 3 years ago

Avatar-blank-50x50

Adrian

Yes, a tool that can understand and attribute on a more complex algorithm the importance of interactions.

Exactly Kenshoo/Marin Software are the ones for PPC / any type of campaigns that can use URL tagging (going to display, banners, newsletter etc).
I believe in this moment that only SEO cannot be tracked properly as importance, all the other types of interactions can be monitored properly like that.

Indeed a tool that can monitor all the channels properly + digital attribution would receive a large amount of interest.

over 3 years ago

Comment
No-profile-pic
Save or Cancel
Daily_pulse_signup_wide

Enjoying this article?

Get more just like this, delivered to your inbox.

Keep up to date with the latest analysis, inspiration and learning from the Econsultancy blog with our free Daily Pulse newsletter. Each weekday, you ll receive a hand-picked digest of the latest and greatest articles, as well as snippets of new market data, best practice guides and trends research.