Shomei Contributes to Attribution Round Table

Brands face the attribution challenge
By Lynsey Barber, marketingmagazine.co.uk
A fragmented digital space presents a valuable attribution challenge. Here, leading marketers discuss a fascinating and vital subject
Click here to read the full article

http://goo.gl/qBmA7

Recent Study Shows Relationship Between Paid and Natural Search

The Search debate is often centred on the value achieving high Natural search page rank and thus mitigating the need for Paid bids for the same terms. Is it that simple? In a recent Study for a lead client Shomei explored the detailed relationship between Paid and Natural Search – Generic and Brand terms. Through utilising the Shomei attribution methodology several distinct relationships were identified.

Value – whilst value means different things to different Brands, for this study higher ‘basket’ value was the key metric. Our Brand delivered products for both domestics and commercial audiences. For commercial products Shomei identified that ‘generic’ (non Brand) natural term listings drove the lower value buyers compared to the same paid term listings, whilst for domestic products the exact opposite was true… For Brand terms both products showed higher values for natural Brand listings compared to paid Brand – despite those terms being universally present in Search Engine Results Pages (SERPS).

So what? For our example Brand this simple Insight allowed the Strategy for natural listing 'Search Engine Optimisation' (SEO) investment to be refined concentrating on improving higher value terms and therefore better SEO Return on Investment (ROI). In addition where budgets for paid search meant lost share, budget could be refocused on those terms that drove higher value – again delivering better ROI.

Conversion – When we explored further we also identified that as well as higher and lower value, for generics and Brand terms, Natural versus Paid, there was a correlation between propensity to convert. For example we identified markedly different propensity to convert between the Paid terms versus the same Natural listing, as well between paid and natural. This Insight was then used to further refine the Investment Strategy leading to further ROI gains…

In addition to the basic but significant investment optimisation and improved ROI at a paid versus natural level, Shomei delivered a term by term level report allowing each to be optimised ensuring that every pound spent drove the highest value clicks that had the greatest propensity to convert irrespective of whether it was at a paid level on a day to day basis or setting the longer term SEO investment Strategy.

Figure 1. The diagram above shows a simple RAG based model that can be used as the basis for investment and concentration of optimisation effort…

In addition to driving optimisation the derived Insight base led to a greater understanding of the behaviours of the audience, the ‘considered’ natural of the interaction as well as the core relationship with the Brand. This deep level of Insight facilitates better planning of all media channels, as well as creative and messaging development.

If you would like to know more about the Shomei methodology please download our white paper ‘The Beautiful Game’ or read the sections on Performance Intelligence and Media ROI Measurement.

Guidance on the use of Cookies…

Under the rules of ‘The Privacy and Electronic Communications Regulations 2003, an EU Directive, the use of cookies and other such technologies are regulated. These technologies store and access information on a ‘user’s’ device – mobile, laptop, PC etc., using cookies. A cookie is a simple and small file that contains information such as user IDs, campaign information, basket contents, preferences etc., etc.

So what’s the big deal and why is it all of a sudden a hot topic? In 2009 an amendment to Article 5(3) of the E-Privacy Directive was made that requires consent for cookies – in other words to store and subsequently access information of the users device. These changes were to be implemented by member States before 25 May 2011.

The basic tenant of the legislation is to protect user information, even if that information is not personally identifiable. Originally scoped to address concerns over technologies such as spyware, it now covers all such technologies irrespective of the motivation, need or intent. Another contributing issue is the lack of awareness of what cookies are and how they are used – hence the ‘desire’ for legislation.

So what are site owners expected to do? In short understand what cookies are used, provide suitably informative ‘usage’ information for users and deploy ‘specific measures to obtain consent’. Clearly where consent is not forthcoming then systems need to be altered to continue to function, which could be a very significant task.

Don’t think you’re not affected… The legislation covers all types of cookies such as Flash, web beacons et al, whether they are session, persistent or time limited, first or third party.

So what does consent really mean? The guidance from the ICO (Information Commissioners Office) essentially states: -

• Before a cookie is set information about cookie and information useage and storage must be provided and explicit consent obtained.
• The ICO know this is difficult, and therefore websites must be able to demonstrate that they are actively reducing the time before cookie information is provided and consent obtained.
• Use of persistent cookies should be considered and wherever possible time limited or session cookies should be used instead.

What about implied consent? Again guidance can be summarised as: -

• Any reliance on implied consent must be on the basis of ‘shared understanding’ of what will happen, i.e. a cookie will be set, what they do and that ‘they’ signify their agreement.
• The ICO believes that awareness of cookies is so low that reliance on implied consent is untenable at present.

So who has to give consent? One (frankly daft) part of the legislation is that consent must be gained from both the subscriber and the user. The subscriber is defined as the person that pays for ‘the line’ and the user is the person using the device. So in my personal case I pay for home Internet access and my children use various websites over that ‘line’… In practice it is near impossible to get consent from both, the ICO suggests that consent from one at least should be obtained.

The ‘subscriber’ issue is further complicated by determining whose consent is required – the user or the subscriber. In reality using the example above, as the ‘subscriber’ I can change the browser settings so that cookies are blocked. I have declined consent on behalf of the user if that user can’t change those settings. Now change this to an employer ‘subscriber’ and employee ‘user’ scenario – the employers wish to allow cookies should NOT result in unwarranted collection of personal data of the employee. So using the browser alone is out…

So are there any exceptions? In short, yes there are, but interpretation becomes harder. For example the sole propose of providing a subscription service or where it is ‘strictly necessary’. The point here is that an exception is where it is for providing what the user wants not what the website owner wants. So if I want to buy a product, a cookie can be used to allow my selected purchase to be put in the basket for example, but if that cookie is then used to analyse what pages were visited and which keywords clicked then no exception applies. The one other exception is compliance with the data protection law.

So who has to comply? Again this is not strictly defined, but the website has the primary obligation irrespective of whether the cookie is first or third party. So you need to know what is being stored and by whom and make that information and the facility to gain consent available to the user. With first party cookies the website has complete responsibility and for third party cookies that responsibility is shared with the third party.

What about Browser settings? The directive states: -

‘consent may be signified by a subscriber who amends or sets controls on the internet browser which the subscriber uses or by using another application or programme to signify consent.’

However the main issue is that browser technology today does not cater for differing types of cookie – they just have a global ‘enabled’ or ‘disabled’ control. As we have already highlighted certain types of cookie are permitted, browsers just won’t do – today. But in the future as browser technology is enhanced then potentially the browser can be used to signify consent.

So what could websites be doing now?

1. Audit use of cookies and how they are used – (what information is being saved etc.)
2. Assess whether under the ICO guidance you need to seek consent or not, or for that matter for which cookies you do or do not need consent
3. Publish information about use of cookies and what type they are – making that published information easily accessible to the user
4. Act now… implement consent gathering solutions – or at least have a plan and be working towards it…

When do websites need to comply?

The ICO essentially gave 12 months lead-time for compliance – which means compliance by 25 May 2012. Any website for which a complaint was received would need to demonstrate they are actively implementing measures to comply…

So what if websites fail to heed the ICO guidance? Well in short he can: -

1. Require organisations to provide him various information – see 1 above
2. Require websites to give an undertaking to a course of action – see points 1 through 4 above
3. Issue an enforcement notice…
4. Fine websites up to £500k

Supplementary issues: -

• Multiple websites for one organisation can be dealt with individually or collectively but it must be absolutely clear for what consent is or is not being given etc.
• Consent only needs to be collected once – unless the use changes in which another consent needs to be gained
• Consent does not need to be gained for each cookie but can instead be gained by function as long as suitable information is provided
• You need to provide for user to change their mind – either withdrawing or granting consent at anytime.
• The legislation covers more than just cookies – if an ‘alternative’ such as device fingerprinting is used the legislation still applies
• Analytical cookies are NOT excluded from this legislation
• The DPA still applies – this is additional legislation

Notice: - Please DO NOT take our word for it or take our advice as definitive or wholly accurate – it’s our interpretation… please seek your own compliance advice.

The relevance of Ad format and Creative messaging...

Too often I see Advertisers following a standard approach to Ad development and placement, quite simply they produce each creative in every relevant IAB standard format. They hand that over to their media Agency and they are trafficked across the media buy without any tangible criteria other than placement format.

Wrong... In tests conducted across various Shomei Advertisers we have seen a huge variance in both format and message. One such example is a well known cinema operator, whom as you'd expect was promoting film releases coupled with specific sales promotions. In analysing their approach of all creatives in all formats, we identified that certain formats worked better in driving immediate response and others drove subsequent site visits and sales - analysis being done using Shomei's partial contribution attribution models. We identified that using different creatives in different formats would deliver performance step change. The net net, of such changes was around 10% gain in conversion from display.

Wrong... You'd be wrong to assume that this behaviour is generic across Advertisers, formats etc., one rule fits all. For example I often hear broad brush statements such as 'an MPU is more engaging than a Banner'. Well in the example above that is certainly true, more viewers engaged with the MPU, when it contained film release trailers etc., compared to the Banner. However the reverse was found with the Banner format out performing the MPU when a ticket sales promotion was included. In combination the uplift of the 'right' message in the 'right' format was significant - it also reduced production costs!

Wrong... Simple click through (CTR) and post impression analysis using 'Last-click' attribution is not good enough to indenting (I don't know what you are trying to say here - for identifying?) such behaviours. Full partial contribution analysis using individual user level journeys is required. You can find out more about such analysis by reading our white paper 'The Beautiful Game'. There is also other posts on this topic - Why is multi-model Attribution important? or Creative Attribution and How to make display work…

Wrong... Attribution analysis is all about Channel level optimisation and one time performance step change. At Shomei we do deliver initial large gains and then on an ongoing basis we use our analysis approach to deliver consistent long term gains and Insight. Incremental gains such as those identified above are much more significant than those that Advertisers are used to achieving with 'Last-click' metrics and certainly out weigh the cost of systems such as Shomei. For more information on Portfolio Optimisation analysis please see: URL to be inserted...

If there are other Attribution related topic areas you would like us to address, please do add a comment and we'll happily oblige...

Google Launches Default SSL Search

So there’s a huge amount of buzz about Google SSL search and what it means. So what’s the issue and what does it really mean to Analytics and Attribution Management?

The long and the short of it is the loss of referrer information for Natural search clicks. Most Website Analytics and Attribution Management technologies like Shomei use data ‘attached’ to the user when they arrive at a site to show what keywords (search term) the user entered to find and click to a site. In techno speak it is called referrer data. With Google SSL search the keyword data is no longer available.

For SSL search to be enabled the user either needs to be a logged in with Google, or use the https://www.google.com URL. Whilst some Google sites do not yet apply SSL by default it will happen over time – with Google what they start rarely stops with just the .com site or for that matter logged in users.

Whilst we totally support Google’s move towards SSL (Secure Socket Layer – look for the ‘s’ in https – this denotes a SSL connection) to protect users privacy we have to question why removing the Keyword from the referrer data has any bearing. If you look at the following two sets of referrer data from a search to the Shomei site you’ll see that there is only ONE difference.

Non SSL search done at http://www.google.com: -

http:\/\/www.google.com\/url?sa=t&rct=j&q=shomei%20attribution&source=web&cd=2&sqi=2&ved=0CC8QFjAB&
url=http%3A%2F%2Fwww.shomei.com%2Fproducts%2Fmedia-attribution-management.aspx&ei=QOCmTq3nNMGu8gOYqYytDw&usg=AFQjCNHPqYxQLp
MlxMvePoyYFgumdQmJmg&sig2=cJc-_ba_xSRq9m7I8Qv6Qw

SSL search done from https://www.google.com: -

http:\/\/www.google.com\/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&sqi=2&ved=0CCQQFjAA&url=http%3
A%2F%2Fwww.shomei.com%2Fproducts%2Fmedia-attribution-management.aspx&ei=IeWmTtiuK8Ok8QPC1dy7Dw&usg=AFQjCNHPqYxQLpM
lxMvePoyYFgumdQmJmg

So you’ll note the ONLY difference is that the ‘q’ value used by Google to denote the search term in the second data set has changed to ‘ersc’ instead of ‘shomei attribution’, you need to ignore all the ‘&’ and ‘%2F’ and the random IDs etc., as these denote gaps, different variables etc.

How does that protect the user if they were anonymous anyway? All it does is weaken the ability of the site owner to see what Keywords are being used by anonymous users – even in Google Analytics all you’ll see is a row in your ‘Traffic Source – Keyword’ section of ‘Not Provided’…

So to be clear it does not matter if you use Google Analytics or any other web analytics system you will cease to be able to track the search term when Google SSL search is used. This also means that any Attribution Management system is similarly affected. We’ll need to wait and see how many search terms this affects and maybe the application of ‘pareto’ will allow a sensible modelling of keyword usage.

When the last cookie crumbles!

Cookie bombing and conversion scooping are just two examples of the jargon surrounding ‘last-touch’ and how such Attribution models can be manipulated inappropriately. The following post explains some such problems and several solutions that advertiser can adopt to mitigate such cookie manipulation.

Let’s try and keep it realistic and avoid the jargon and stick to simple examples. First a site based ‘re-targeting’ example. A notional digital advertiser deploys a range of media, focused on driving sales. Customer 1, searches for our advertiser’s product and then clicks on a paid search ad and arrives at the site, reviews the product but leaves without buying. Our advertiser has retargeting active and our customer 1 is added to the re-targeting list and is subsequently exposed to various banners for the product but does not click. He later returns directly to the site and buys.

I know you’re ahead of me… the last touch attribution would attribute all the value to the retargeting irrespective of there being no click. The original search would get no value attributed.  Okay so in this example you could argue that adopting ‘first-touch’ would work but did the retargeting do nothing? What if it was behavioural targeting that drove the first visit with a paid search click that preceded a subsequent visit and conversion event? Did the paid search deliver no contribution to the sale?

The above examples are fairly typical examples and typify the debate that rages about ‘last-click’, ‘last-touch’ type attribution and what’s the best approach. It also reflects the annoyance focused on the retargeting industry by the Affiliate community – not that I haven’t heard the accusation against the affiliates for merely cookie spamming the browser and claiming all the credit. The fact of the matter is that each channel has a role that contributes to the end result – the ‘fault’ is the simplistic attribution approach.

So what can be done?

Multi model, multi touch, causality based attribution combined with path analysis is the answer! Yes, don’t worry I’ll explain: -

1. Multi model – relates to the number of models against which a touch point is scored. Attributing the value of a sale is a direct response score and therefore other models such a ‘first-touch’, first-click, ‘last-click’ etc., should be used in combination to provide a clear view of the ‘overall’ contribution.

2. Multi touch, causality – relates to scoring all touch points that contributed to a given event, the touch points that are deemed to have contributed are those where the cause and effect can be measured – where it can’t the touch point should be considered data noise! Just because it’s tracked does not mean it contributed. In essence there should be a measured benefit before a touch point is accredited value – all such touch points are attributed the correct partial amount based on their partial contribution using a Recency and frequency complex attribution model.
3. Path analysis facilitates identification of the distinct combination of media types and the differences between them. For example paid search with and without a proceeding display impression etc., such as with retargeting. We need to identify distinct journey combinations and check that the ‘cause’ delivered a resultant ‘effect’.


Fig 1. Example Conversion Path using Attribution Media Management (click image to zoom)

In Fig., 1 above you can first see Paid Search with a Preceding Display Impression (in this case behavioural targeting) and below it a Paid Search Click only, in comparison you can see that where the impression is present the conversion rate is 1.68% versus 1.25% - so causality exists. Now if you look at the second extract you’ll see Display Click followed by a Display Impression (retargeting). In this case the conversion rate is 0%. Now I’m not saying this is absolutely typical – there is no such thing, each brand and media combination works very differently. You just need an attribution framework such as Shomei’s that can help you identify and value the cause and effect.

The Shomei attribution approach removes any debate about which media contributed. With a proven alternative, the last cookie should finally crumble! Click here for more information

Attribution, Profit and Customer value …

The ‘Profit’ and ‘Customer’ words are often ignored when it comes to Digital acquisition and media performance measurement. Why? Are marketers unable to get their heads around the difference between ROI in investment and the long- term bottom line contribution? I doubt that’s true, I believe it’s because current measurement systems, metrics and KPIs just don’t deliver – they are often derived from ‘one size fits all’ systems such a Google Analytics and are therefore too generic. The opportunity to optimise marketing efforts focused on bottom line contribution is often missing. It need not be!

Attribution platforms like Shomei address such shortcomings and deliver huge opportunity to increase customer value and profit: -

Attribution measure and KPIs

I’ve covered in detail many of the ways that moving away from the de facto ‘last-click’ standard to measures and KPIs based on an Attribution framework such as Shomei’s can deliver huge performance step change – I’m not cover it in this post. If you’d like more on that topic please do have a browse through other Shomei blog posts or visit www.shomei.com, here are a few posts that will give you a feel for what can be achieved: -

How to make Display work
Attribution based optimisation
Attribution ‘X’ Factor…

So if you’re using ‘last-click’ Attribution, the first positive step would be as I mention above, change it! But let’s assume you’ve now done so and achieved performance step change, we typically deliver c. 30% plus during this step. What else can you do beyond regular Attribution portfolio optimisation? Here’s a couple of options: -

Lifetime Value Optimisation

Lifetime value optimisation is for many Brands the next logical step. A customer segmentation typically represents groups of customers meeting certain criteria, for example purchase frequency, basket value, Recency etc., etc. Lets assume we have a simple segmentation that represents best, average, below average and worse clients, (Segments 1 to 4)… By integrating such data into the Attribution Framework there are a number of distinct advantages and opportunities.

One such example would be ‘differential acquisition’, in simple terms pay less for the worse clients (4s) – pay more for the better segments (2,3 and 4s). Adopting such an approach leads to significantly different media planning, for example some activity that historically might have been excluded based on an average CPA might now deliver a ROI if the media attracts a high proportion of your target segment 1 clients. Couple this with reducing the proportion of 4s attracted, given they might not even represent a ROI, you are suddenly achieving a rebalanced client portfolio... And maybe more importantly substantial Campaign growth.

Profit Optimisation

When such a segmentation does not exist or the data required is not available or indeed the purchase is not typically repeated or that frequent then concentrating on basket profit is probably the best approach. Here instead of focusing the customer attribute we focus on the composition of the basket and the profit generated by each item. Essentially this allows you to go beyond a blanket ROI, and deliver optimisation based on profit.

The above represent a couple of the simple approached – in real-terms there are many more ways that such profit and customer value data can be utilised within the Attribution Framework to deliver value. Hopefully the above is enough to get your thinking about the opportunities that might be in your reach.

So how does Attribution fit?

One of the major advantages of a Framework Attribution approach is that it delivers very clear funnel based metrics. So for each segment (be that customer segment, or product profit line etc.,) each media touch point can be scored. This allows the site, placement, ad creative, messaging and keywords to be optimised to either attract more or less of each segment at each stage of the customer journey as required to drive great customer value and profit. Couple this with segmented actionable Insight, the wider Strategy and planning can be improved before any ‘test and learn’ approach is adopted.

Moving to an Attribution framework like Shomei’s delivers low hanging fruit through media optimisation, followed by long-term growth and greater profitability. Isn’t time to move beyond, ‘last-click’ wins and deliver competitive advantage?