6 Keys to Digital Success in Attribution Modelling


Today I am going to share with you the ‘6 keys to success’ in attribution modelling. Learn my tricks and tips and your success in attribution is guaranteed.

  1. Understand Customers Purchase Journey
  2. Understand the concept of ‘Missing Touch Points’
  3. Fix Data Integration issues
  4. Go beyond Google Analytics
  5. Understand not all “touch points” are equally valuable
  6. Use Proportional Multi Touch Attribution Model


1. Understand Customers Purchase Journey

Understanding customers purchase journey is the no.1 requirement to fixing the attribution problem. If you fail to understand your customers then you have already lost the attribution battle. Well not just the attribution battle but almost every other battle in the business and marketing world.

I have broken down this journey into following steps:

  1. Interview your client, customer support and the sales people.
  2. Understand the decision making process
  3. Understand ZMOT
  4. Use data from ‘Google Think’
  5. Do Proper Market Research


Interview your client, customer support and the sales people

Your client and his staff know more about their business and industry than you ever will. So it is imperative that you leverage their industry expertise and get a better understanding of their target market.

Ask following questions:

1. How do you define your target audience (age-group, gender, education, income, likings, ethnicity, lifestyle etc) and where majority of them live?

2. Who are your best customer types in terms of revenue generation and why?

3. What kind of relationship you want to build with your target audience?

4. What are the desires and expectations of your target audience?

5. What is the level of product use? Are your customer loyal to you?

6. What are the most common objections raised by your customers?

7. Who are the actual decision makers (who has the final say)?

8. What are the buying triggers?

(A trigger is an event that causes a person to get into a serious buying mode.

For example you might have a vague interest in going to Seattle. This might have caused you to browse the web for hotels, flights etc. But an upcoming Mozcon event could act as a trigger that makes you seriously look for hotels and flights. )

Ask as many questions as you like. I have just given you few examples.


Understand the decision making process

Start your analysis by understanding the decision criteria framework of your customers.

In case of fast moving consumers goods (like toothpaste, milk, soap, vegetables etc which are bought frequently) least amount of consideration and evaluation is involved before making a purchase.

It is highly unlikely for someone to visit half dozen reviews websites, product comparison and coupon websites just to make an informed decision on buying a toothpaste.

Often in case of FMCG products, people already know what they want to buy. They already have set decision criteria in their mind (like certain model or make). So they just go to the store and make a purchase straightaway.

However they are many products/services which require lot of consideration and evaluation before a purchase is made as they are bought occasionally. Often in case of such products,

the decision criteria framework cannot be established quickly because of ever changing specifications which results in a  long sales cycle.

For example:

People don’t buy a new car ever week. So when they do think of buying a new car, there are not really sure what they are looking for in the new car.

So they need to do lot of research just to determine the ideal specifications of their new car. Once they have determined their ideal car’s specifications they have established their decision criteria framework.

Once the decision criteria is set, consideration and evaluation begins to take place and ZMOT occurs.


Understand ZMOT

There is no attribution without understanding ZMOT

Zero Moment of Truth or ZMOT is the moment which occurs after the customer has been exposed to your brand, but before a purchase is made.

It is the moment when the customer do research and make decision about buying your product by going back and forth between various digital and non-digital channels known as ZMOT sources:


Since 84% of all shoppers use ZMOT sources in the path to purchase (source: Google Think Insight),

you need to find ZMOT sources and optimally allocate your marketing spend across them.

You need to determine that if you invest in multiple ZMOT sources, then how much incrementality does each ZMOT source can bring to your company’s bottomline.

ZMOT is the most powerful moment in a customer journey to purchase as it shapes the consumer’s purchase decision. And

 If you fail to understand your customers at ZMOT, you have already lost the attribution battle

To know more about ZMOT and its importance, please read this amazing article: How to use ZMOT to increase Conversions and Sales exponentially


Use Data from ‘Google Think’

Google Think Customer journey to online purchase is an excellent tool to understand how certain acquisition channels assist conversions in your industry and how the length of the customer journey impact the order size.

For example:

Before you create your own attribution model, you need to understand the role of acquisition channels in your industry:


As seen in the graph above, within the tech industry, direct and ‘other paid’ interactions often act more as a last interaction.

The organic search acts both as assist interaction and last interaction.

Whereas display, social and email act more as an assist interaction.

Without such understanding, you could negatively impact your brand awareness if you dump campaigns which play key role in initiating conversions.


Do Proper Market Research

By ‘proper’ I mean, investing considerable amount of time and resources in conducting the research.  Market research is an excellent way of understanding your customers.

Conduct surveys, do A/B testing (see what works and what doesn’t work), hire a market research agency.

Buy market research and industry reports from companies like Experian and read them from page to page. The level of insight that you will get from such reports is unparalleled.


2. Understand the concept of ‘Missing Touch Points”

You need to understand the current technical constraints involved in ‘attribution modelling’:

The multi-channel funnel reports that you see in your Google Analytics account show attribution only across ‘digital channels’ and that too for only one device and one browser.

This report doesn’t take into account the offline touch points (touch point means ‘exposure’ to a marketing channel like exposure to a newspaper ad or TV ad).

Google Analytics identify a person through his/her browser and device in multi-channel funnel reports.

So if a person visited your website via ‘chrome’ browser on a desktop PC and then later converted via ‘safari’ browser on an IPAD, then Google Analytics will report that they are actually two people who visited your website.

The first person visited your website via ‘chrome’ browser on a desktop PC but didn’t make a purchase. The second person visited your website via ‘safari’ browser on an IPAD and made a purchase.

Clearly this is not true but that’s how you will be reported of the customers’ behaviour on your website 🙁

So your attribution modelling will be based only on known touch points.


Another example

If a person saw an ad on TV and then later converted via paid search ad on a desktop PC, then Google Analytics will report that a person clicked on a paid search ad and made a purchase. GA will completely ignore the role of the TV ad prior to conversion.

So again your attribution modelling will be based only on known touch points.

You can now realize that

you get distorted picture of conversion paths from your multi-channel funnel reports.

Needless to say, if you are heavily involved into multi-channel marketing both online and offline, your conversion path reports could be way off the mark 🙁


3. Fix data integration issues

Data integration is the key to fixing attribution issues

You should always aim to minimize the number of missing touch points in your conversion path by integrating as much data as possible from different data sources.

These data sources can be (but are not limited to):

  1. Google Analytics
  2. Google Adwords
  3. Google Webmaster tools
  4. Google Merchant Center
  5. Bing ads
  6. Kissmetrics
  7. Qualaroo
  8. Facebook Insight and other social analytics data
  9. Compete
  10. Survey Monkey
  11. Phone Calls data
  12. CRM data
  13. Point of Sale (POS) data
  14. Data from Customer Support
  15. Financial data and data from other departments.

Once you have integrated all the marketing and business data in one place, you can quickly track various aspects of your marketing campaigns, analyze the overall performance and above all take timely decisions.

Data integration can help you correlate all of your data with business bottomline impacting metrics like revenue, cost, gross profit etc

Without proper data integration, you will always get SILO view of your marketing campaigns.

You need to create a robust data integration system in order to carry out any meaningful analysis. In fact

if you are a big organization then it is completely pointless to collect and analyze big data without proper integration.

You have to invest in data integration technologies if you are really serious about carrying out attribution modelling.


Start by using Universal Analytics and then gradually move to custom built applications.

Universal Analytics (UA) provides more ways to collect and integrate different type of data than Google Analytics (GA).

Through UA you can integrate data across multiple devices and platforms. This is something which is not possible with GA.

Consequently UA provides better understanding of relationship between online and offline marketing channels that drive sales and conversions than GA.

Source: Beginners Guide to Universal Analytics – Creating Custom Dimensions & Metrics

Eventually you have to use customized applications because no single tool/software alone can minimize all of your data integration issues.


4. Go beyond Google Analytics

Attribution is much more than Google Analytics.

It easily goes beyond Google Analytics. Don’t limit yourself by just using the GA attribution tools.

You are not limited to just what you can achieve through Model Comparison Tool and Multi-channel funnel reports. There are lot of attribution modelling softwares out there which can provide much more robust attribution solutions than Google Analytics.

You should definitely consider them esp. if you are a big company. In the end only custom made attribution tools can minimize attribution issues as they provide:

1.  Robust data integration capabilities

2.  Robust data forecasting platform

3.  More flexibility in terms of creating attribution models and applying credit rules.

4.  Customized solutions as each business is different


5. Understand not all “touch points” are equally valuable

The exposure to a marketing channel during the path to conversion is known as interaction, touch or touch point. So if you are exposed to one marketing channel in your path to conversion then your conversion path includes only one touch point.

Similarly, if you are exposed to 4 marketing channels in your path to conversion then your conversion path will include 4 touch points.

Now consider I follow this conversion path:


Here I am exposed to 7 different acquisition channels before I made a purchase. Since each of these exposure is considered as a touch point, there are 7 different touch points in my conversion path.

Now let us see how credit for the conversion is distributed to different touch points under different attribution models:


In case of first touch attribution model 100% credit for conversion is attributed to the first touch.

So according to first touch model, my reading of the blog post gets all the credit for conversion. But this is not true. As you can see 6 other acquisition channels have also played an important role in my path to conversion.

In case of last touch attribution model 100% credit for conversion is attributed to the last touch.

So according to last touch model, direct traffic gets all the credit for conversions. Again this is simply not the case as there are 6 more channels in play.

In case of Linear attribution model all touch points get equal credit for conversion.

So according to linear model, all 7 touch points are equally important in my path to purchase. But this is also not true.

I read product reviews and went to product comparison website before making a purchase. These two touches were more valuable to me than the exposure to the blog post, display ad, PPC ad and organic search listing as they played a very important role in my purchase decision.

Had I not been satisfied with the product review or pricing, I wouldn’t have made the purchase in the first place.


In case of time decay model the touch points which are closest in the time to conversions get more credit.

So according to time decay model, my exposure to organic search result and visiting the website directly should get more credit than to my exposure to product reviews and pricing.

Again such type of credit distribution is not accurate as had I not been satisfied with the product review or pricing, I wouldn’t have made the purchase in the first place.

The time decay model is the modified last touch attribution model. Though it is crappy but less crappy than the last touch and other GA models. 

One issue worth pointing out about giving more credit to last touches/interactions:

Last interactions act as a last point of contact prior to a purchase. By the time a customer experience last touches to  marketing channel(s), a purchase decision has already been made.

So it doesn’t really matter which channel closed the sale when it comes to allocating budget. 


In case of Last Non Direct click attribution model 100% credit for conversion is attributed to the last non-direct click.

So according to this model, organic search gets all the credit for conversion. So this model also give incorrect picture of the conversion path.

In case of Last Adwords Click attribution model 100% credit for conversion is attributed to the last Adwords click.

So according to this model, paid search result gets all the credit for conversion. But this is also not a true representation of my buying behavior.


In case of Proportional Multi touch attribution model the credit is distributed to touches in proportion to their contribution in the conversion.

The acquisition channel which assists the most gets the maximum credit for conversion and maximum resources are allocated to it regardless of it being the first touch, last touch or middle touch.

All other touches would get credit in proportion to their contribution in the conversion path.

So according to this model, exposure to product review and product comparison websites get more credit for conversion than all other touch points as they played key role in the decision making process and my purchase journey.


6. Use Proportional Multi Touch Attribution Model

I am suggesting this model not because I developed it but because it is incrementally better than all the attribution models out there.

Following attributes make this model by far the best attribution model in the world (ok that’s an exaggeration 🙂 )

1. Proportional multi touch is the first generation of real world attribution model.

Because of that property it has the ability to provide truly complete picture of the conversion path followed by customers. It is the first generation of truly multi-channel analytics modelling tool.


Source: Selecting the Best Attribution Model for Inbound Marketing

 2. Proportional Multi Touch takes into account your business model, marketing objectives, sales cycle, customers’ activities and seasonality as it allows you to assign credit to different marketing channels/touch points (both online and offline) in proportion to their contribution in the conversion process.

Thus it provides more flexibility than linear, position based and time decay models ever will in terms of credit distribution.

3. The usage of proportional multi touch model goes beyond Google Analytics as it is not just a digital model.  Because of this property it faces less issues of ‘missing touch points’ than the traditional Google Analytics models.

Other articles you may find interesting

Quick announcement about my new book on Web Analytics and Conversion Optimization

The role of maths and statistics in the world of web analytics is not clear to many marketers and web analysts. Not many talk or write about the usage of statistics and data science in conversion optimization. That is why I have written this book Maths and Stats for Web Analytics and Conversion Optimization to fill this knowledge gap.

This expert guide will teach you, how to leverage the knowledge of maths and statistics in order to accurately interpret the web analytics data and take business and marketing decisions which can quickly improve the bottom-line of your online business. Read first few chapters of this book for free here.


Himanshu Sharma

Certified web analyst and founder of OptimizeSmart.com

My name is Himanshu Sharma and I help businesses in finding and fixing their Google Analytics and conversion issues.
  • More than ten years' experience in SEO, PPC and web analytics
  • Certified web analyst (master level) from MarketMotive.com
  • Google Analytics certified
  • Google AdWords certified
  • Nominated for Digital Analytics Association Award for Excellence
  • Bachelors degree in Internet Science
  • Founder of OptimizeSmart.com and EventEducation.com
I am also the author of the book Maths and Stats for Web Analytics and Conversion Optimization If you have any questions or comments please contact me