This topic has been on my mind a lot recently and naturally I fall back on the teachings of the legendary Avinash Kaushik. There are many multi channel attribution challenges that Avinash talked about:

• Impact of Online channels on offline stores/sales
• Impact of Offline marketing activities on Online traffic and conversions
• Attribution across multiple screens: i.e. conversions happen cross devices from laptop to mobile to tablet. Or even from smart TV to laptop to mobile etc…
• Attribution across multiple online channels. 

You could refer to his posts on Definition, Model, Reality Check, Tracking Online Impact of Offline campaign, Tracking Offline conversions. So when someone talks about Multi-channel attribution, I guess we need to define clearly what type of multi-channel attribution he/she wants to talk about. In reality, almost all enterprises face two or more multi-channel attribution challenges so deciding what problem to focus on is as important as solving it. One way to approach this is by attempting to answer these questions:

• What action we could take as a direct result of the multi-channel attribution analysis?
• What impact could this have on our overall marketing objective/sales? Does the estimated impact justify the efforts?
• Who do we have in the team to do the analysis once we have the data? Will this person be in the position to present data/convince other stakeholders to follow the recommendation?

Now, let’s look at different scenarios together.

1. Impact of Online Channel on offline store visit/sales

This is something that puzzles marketers for ages and there have been many efforts trying to solve this question across some online channels, but not holistically. The value of answering this question is quite obvious. If you know the impact of different online channels on offline store visit or sales, then you could try to up weight media spend for certain channel and observe store visit/sales. If the model is correct, you should see uplift.

Also you could conduct various tests to try to establish the “perfect” percentage split between different online channels to maximise the store visit/sales.

However, the fact is that your store visit/store sales does not depend solely on your digital activities, it’s affected by your other offline activities, your sales promotions, other conversations your prospective customers have about your brand (on social media or not), even what your competitors are doing etc… So rather than trying to get a perfect media split, it’s important that you feel comfortable working with an incomplete set of data.

In terms of ways to track offline impact, we have seen things like:

• Use dedicated coupon/promo code/phone number for your online channels and then track redemption/sales in store. To do this, you need to have a database of all Online only coupons/promo codes and dedicate phone numbers, your staff at store need to be trained to follow process of logging in the information and the ability to match all of this data back to each online channel (impression, click, cpc, to acquisition and revenue).
• Membership: using membership data & login data, you could track users online/offline behaviour similar to a bank tracking its customer transactions.
• Micro conversion set up for anyone who visit your store locator/contact us page online. The assumption is that the more people using store locator, the more likely they will visit your store
• Run Online campaigns in certain specific locations: and observe store visits/sales to the stores in these locations vs other locations. Many online channels allow you to target people at city level. You could refer to the Online to Store case study done by HP and Google when they use this concept. 
• Do survey at store or on-exit survey online to understand the behaviours of your audience etc… Avinash covered everything beautifully here.

2. Impact of offline marketing to Online performance

Online performance benefit from offline activities all the time. There is little doubt that major PR campaign, TVC campaign, sales promotion or other offline activities would impact website traffic/conversions. The only issue is that with major offline activities (TVC or PR or OOH), it’s hard for us to optimise the campaign on the fly due to high cost of production and a even more complicated booking plan.

If we buy TVC/OOH on a programmatic basis then its a different story all together, however, until that time, it remains challenging to change the offline media plan half way through the campaign or optimise bi-weekly.

So again, what you could do?

• Vanity URLs like abc.com/train, abc.com/print or QR code to differentiate if certain traffic to the site is due to an ad on the train or a PR article on newspaper.
• Organic search uplift for brand search terms during the TVC/Print/Out of Home campaign (we assume that after the first exposure offline, people would go online and search if they are interested). However, again this is very hard to read and get a correct figure because you don’t run offline campaigns in silo, you run that together with other online channels, social media channels, mobile channels so it’s hard to attribute the uplift to a single channel.

3. Attribution across multiple screens

This is one of the major problem facing businesses today. Consumer no longer used one screen to research/buy/share information about a product/services. It is very common for a typical consumer being exposed to certain ads on their mobile, later on do some research on their laptop and click on a paid search ad and then go to the store to buy. So how could businesses optimize their media spend situation like this? Which channel should get credit for the sales? How much credit should we give to our mobile display campaign, or paid search on desktop?

There seems to be no clear cut answer, at least for now. One of the main challenge is with data collection and identification? How could we collect all the data (from post impression, post click) and being able to know that it’s the same user who is being exposed across multiple channels or when that user takes a desired action? Cookie, which used to be very effective within the computer browser environment suddenly found itself hugely inadequate to address additional screens (Mobile, Tablet and even OOH).

• Login could be a method, however, we would face privacy issue i.e. who would want companies like Google/Facebook/Apple etc… to track their activities across devices and store their internet surfing behaviours?
• Identify users based on contextual information: like if the two devices are used in the same physical locations, same network at the same time (like every evening)
• Unique behaviour/browsing pattern: each one of us would have unique browsing habits/unique app usage pattern. If networks/sites could collect all of these information and do big data analysis, they could in theory identify unique individuals across sites/apps.
• Credit card/transaction behaviour: yup, this is one easy way to track users across multiple devices as well. If your site/app/network could record the same credit card/paypal account being used on two different devices, you could assume to a certain degree that it’s the same person
• Tradition panel based, survey type could be used of course. But here lies the issue of ad recall and users may not even notice/remember every things that they do.
• Click to call could be a way to bridge mobile ads with offline activity.

You could read more about this here, or firms that claim they could do this. I haven’t tried their solutions though.

4. Attribution across multiple online channels

This issue is no longer new and has been discussed/solved to a large extend by different platforms.

Many analytics solution offers this attribution capability, for example Google Analytics (even the free version). You could read a couple of articles like Attribution Modeling Overview, Multi-Channel Attribution Modeling: The Good, Bad and Ugly Models

First, Why do we need to care about this problem? Because if you run Paid Media campaign long enough, you would notice the following scenario:

• Search Engine Marketing (Paid Search or Organic) works best, and majority of the conversions (from last click non direct attribution model) happen because of brand terms. Yay! Let’s spend more money on Paid Search or SEO. 😐 or is it?
• Remarketing works best for Display, yay! fantastic, great work again. or is it?

To be honest, if you don’t optimize your Brand term campaigns for Paid Search or Remarketing campaign for display, they would still perform the best simply because of the way you measure & attribute conversions. There is no brilliance in showing brand terms driving 90% of all conversions for Paid Search programs or Remarketing drive a similar impact with Display. People who know you, search for your brand terms would obviously convert better than people who search using more generic terms. They are often in research mode when they use generic terms. As for remarketing, it feeds off other marketing activities by reaching out to people who have been to your websites before. You already pay certain amount of money to attract them to your site in the first place and now you just use remarketing to convert them.

I don’t want to make it seem too easy to run an efficient remarketing campaign or a brand term campaign for Paid Search. However, the effort is not huge. Anyone could do this. You don’t even need too many people or hire an agency to do this if this result is what you are looking for. By focusing all of your budget on brand term for Paid Search and Remarketing for Display, you do not reach out to new people, who are in consideration stage or further up the funnel, your sales would slowly plateau or “die” over time. So how to tackle this? How to reach people higher up in the funnel in an effective way? You need multi-channel attribution analysis, so that you could understand the impact of different online channels on one another, the multi-visit reality before sales, to assign different KPIs for different activities.

Second, it’s about How? I guess you need:

• Track all online channels (both post click and post impression interaction) including search, display, EDM etc…

• You will probably need to use a third party ad serving platform like DoubleClick Campaign Manager to traffick all of your display activities
• Use a third party platform like DoubleClick Search or Marine/Kenshoo to traffick all of your SEM activities
• Use pixel or JavaScript tracking on the landing page and along the conversion funnel. (DoubleClick floodlights for example)
• Store all data into a common data warehouse/Data management platform or have a way to connect post impression/post click cookie with conversion cookie
• De-dupe conversions across channels and do attribution as per custom rules set by user

If you happen to use Google Analytics for your data collection and analysis, Google Display network for your Display campaign and Google Adwords for your SEM, then Google Analytics could match post impression data, post click back to conversion (defined by Google Analytics and shared with Google Adwords). Linking Google Adwords and Google Analytics in that case would automatically give you post click data for your SEM activities using Adwords and post impression data from GDN so you are good there. You dont need to use third party ad serving platform like DoubleClick Campaign Manager. For other activities like EDM or the like, you could use URL tool builder to tag the destination URL with respective UTM parameters.

In the event that you are running display campaigns on more channels than than just GDN then you will need to use a platform that could do third party ad serving for both Display/Video and Search, many of the popular third party ad serving platforms could do this including DoubleClick. You could go ahead and traffick all other activities like EDM traffic using clicktracker. 

In real situation, what many enterprises face is the inability to connect conversion data with click/post impression data. Reasons for this situation are many like:

• Conversions happen offline
• Conversions/sales involve other teams like Call center/sales team hence information is not passed back internally across teams. 
• There are no unique identifiers for all leads generated online/offline through the CRM system until the point of conversion/sales
• etc…

For examples, leads are generated online, then the sales team/call center team would follow up and try to close these leads. When they do, they just update the internal CRM system and it does not connect to marketing platform to match conversion with click/post impression data. 

It’s not too difficult to solve this issue, you probably need a Data warehouse, a platform for your online marketing (like DoubleClick for example) and they connect with your CRM platform to share data. Then you will just need to assign a unique identifier for all leads that come through. This unique identifier could help to connect marketing activities (post impression, click data per channel) to sales/conversion data. You could do this by using DoubleClick Campaign Management (DCM) placement ID, which is dynamically generated per placement you set up, and similar approach for Search (using {creative ID} for example). This will help you to scale since you don’t need to manually create thousands unique identifiers manually upfront and tag each destination URLs of your online campaigns with these unique identifier.

Finally, after you collect all the data, what’s next? I guess you need someone who is great at analytics to analyse the data and provide actionable & impactful insights. This person would need to understand how your company works, how decision is made across different team, have the charisma and the human understandings to present data in an impactful way and convince all stakeholders to act based on his/her recommendation backed by data.

Anyway, that’s about it from me. Do you have anything to add? Anything that you want to comment on? Feel free to drop your comments below.

Cheers,

Chandler

P.S: England lost today in Brazil!