Marketing Attribution Modelling – determining the optimal mix

Digital marketers have access to a wealth of data about every marketing communication sent out (Impressions shown, Videos Seen), every user action generated (Clicks, Visits) and eventual outcomes drove (Sales, Leads, Downloads, Registrations). On one hand, this is great because never before have marketers have had access to such a humongous amount of detailed information about each and every marketing message and how it affected their target customers. On the other hand, this leads to a very different problem presenting itself – that of understanding which of the messages actually influenced the outcome and which ones did not.

This is the question attribution modeling attempts to answer. The choice of attribution model will define the value that a marketer sees in a specific marketing action. Consequently, attribution models allow a marketer to determine optimal budget allocation., i.e. given a $ 100, what would the optimal mix be in order to drive a specific outcome.

This blog will take a quick look at some of the attribution models that are used in the market today, look at the pros and cons of each model and comment upon how these models are likely to evolve over the medium term.


Part I – Attribution models


There are 3 broad types of models which are used to address this question –


1) Simplistic Attribution – Also called Single Source Attribution, these type of models will generally attribute the desired outcome to a single event. This constitutes of models like –

– First Click Attribution, where the sale is attributed to the marketing channel or event that was responsible for the first click which brought the user to the site or

– Last Click Attribution, where the sale is attributed to the last channel/event that preceded the desired event.


2) Fractional Attribution – These type of models allow for the credit for an outcome to be shared across multiple channels/events. This constitutes of models like –


– Linear Attribution. where the credit for every outcome is distributed equally to all actions (Clicks or impressions) preceding it. i.e. if a user clicked on a SEM ad, on a display banner, and on an email link before buying a product, the credit for the sale is given equally to all 3 (SEM, Display & Email)


– Time Decay Based attribution – This model gives greater credit to channels/events which take place closer to the point of sale i.e. if a person clicked on a SEM ad 15 days before the sale and then clicked on an Email 30 minutes before buying, the Email Ad would be given vastly more credit than the SEM ad, though the SEM ad too would be granted some credit based on certain assumptions made about the average time to purchase that are made by the marketer.


– Position Based Attribution – This model grants a fixed amount of attribution to channels in each stage of the funnel e.g. the channel bringing in the first click would be attributed 40% of credit of a sale, the channel for the last click another 40% and all intermediate channels/events share the remaining 20%. The choice of 40%, 20%, and 40% is a marketing choice made based on the perceived optimal split between first, last and intermediate events (impressions, clicks)


3) Algorithmic modelling – These models do away with fixed weights for first and last click events that the previous models rely on. Instead, these models use proprietary algorithms to determine how to attribute credit across events/channels.


Part II – Pros and Cons


Great, so now we not only have access to a trove of data but also have a multitude of ways in which to infer value from this data. This brings us to the question of the relative merits and demerits of each model. A quick check is marked below –

attrinution models comparison



Part III – Models in the market and how these are likely to evolve


The market today largely relies primarily on Last Click attribution models since these are easier to get alignment on across stakeholders. Needless to say however these leave a significant amount of value on the table for most businesses since the budget allocation decisions are likely to be skewed in favor of events/channels which work in the final stages of the customer journey. Quick explanation for the uninitiated –

Let’s say, a customer goes through 4 stages as part of a purchase funnel – Awareness, Intent, Decision, and Action.

Awareness – Customer is made aware of the offering

Intent – The customer feels the need to acquire the offering

Decision – The customer evaluates alternatives and makes a decision

Action – The customer actually takes the necessary action (to acquire the offering/product)

Different channels have different strengths and perform differently in each stage of the funnel. For e.g. Display might work great in the first phase of the funnel (awareness) but might not fare as well in the final stages. On the other hand, email and search would perform significantly better in the final stages (Decision, Action) once the customer has been made aware of the offering.

To place emphasis entirely on the last step of the funnel (Action) would be equal to saying that the Goal-Scorer is the single most important person in a Soccer Team! Read an article that takes on this question.

While we are seeing the emergence of position based models, these leave a lot to be desired since weight choices are arbitrary to a large extent. Also, the models do not factor in cross-channel effects. For instance, reports from studies by the Atlas institute[“Where can you find your customer”] have shown that users who are exposed to both Displays and Search convert 22% better than users exposed to Search only. However, this channel-specific behavior is hard to incorporate into fractional models.


Algorithmic models offer to simplify a number of the issues that are associated with their more primitive modelling counterparts, but have 2 broad areas that need to be addressed before they can really take wing –

1) Statistics related know-how within organizations – Unless organizations are in a position to understand and challenge the approach used and results that are generated by Algorithmic attribution, there will be little buy-in.

2) Ecosystem-related constraints – Some of these models rely on using clickstream data. Still, others rely on impression level information being available in order to credit an outcome to a channel/event. The ecosystem today is not transparent enough to always be able to provide this data resulting in data gaps which would be difficult to fill.


To summarize – While there is a lot of data available and there are approaches which allow for us to interpret this data, there are a number of subjective decisions which will influence how a marketer optimizes his spends. These choices form the competitive edge because a marketer who understands how best to use the budgets and channels at his disposal is likely to be the one who outlasts his competition.


References –

Infographic: Retargeting ALJ’s customers

Financial Services is yet another industry where Retargeting has generated great results for marketers. ALJ, a leading automobile finance company in Saudi Arabia increased their online leads by 20% in just 2 months with Vizury’s retargeting! Read on.


If you want to reuse this infographic, please write to for the terms of use.

Mobile App Retargeting Is Here!

At the outset, I have a question to ask advertisers. How much of your marketing budget do you spend in driving app downloads? And, how much of it is driven towards engaging with those who already have the app?

The fact is that current mobile advertising technology vendors are focused on driving app downloads for marketers. There is very little or no focus on re-engaging with users who have already downloaded and used the advertiser’s mobile app. Hence, most hard-earned app downloads result in a one-time usage at best and in the eventual uninstalling of the app.

The current mobile retargeting scenario

The diagram below illustrates the various challenges in effectively re-engaging with the existing mobile web or app user base:

Mobile blog image
Android Mobile Web: Android mobile web users can be targeted using cookies and messaged. But usually, the inventory pool available on Android mobile that can support cookie-based targeting is very limited. Most publishers do not support cookie based buying on their Android mobile web inventory as user-based targeting is a new frontier in mobile ad-tech. Also in most geographies, 50% of the mobile inventory is inside apps and cookie-based targeting is not possible on this in-app inventory

iOS Mobile Web: 98% of the iOS users use Safari browser for accessing mobile websites. Third party cookies are blocked on Safari browser and hence user targeting becomes tough. Privacy compliant solutions for iOS Mobile Web user targeting are still in early stages of development.

Android Mobile App: Cookies do not work inside apps and hence the user targeting in Android apps is usually via a device id called Android_ID. However, this id is not privacy compliant as the user cannot opt out of this Android_ID based messaging. Google is introducing a new Android Advertising id that solves this privacy issue by allowing users to disable access to this id. However, this new id is still in very early stages of adoption across the mobile ecosystem.

iOS Mobile App: Cookies do not work inside apps and hence the user targeting in iOS apps is via a device id called IDFA (ID For Advertisers). This id is user privacy compliant as the user can disable access to this id. However, the IDFA is still not widely adopted in the complex mobile advertising landscape

Vizury has pioneered a solution for the retargeting mobile web users called Vizury MobiConvert and launched this product in a beta phase around 7 months back. We have seen great results for advertisers from this product and hence did a wider market launch of this product in October 2013. Now more than 50 advertisers across the globe use MobiConvert and are enjoying great results.

The next frontier: Retargeting Mobile App Users

As the next step in the evolution of Vizury MobiConvert, we are now ready to enable retargeting of mobile app users. With both Android and iOS supporting privacy compliant mechanisms for app user targeting, we believe that the ecosystem is now conducive towards app retargeting. Hence Vizury has launched the Vizury Event Logger API for advertiser apps. This is a super light piece of code that advertisers can put into their mobile apps and then Vizury will be able to retarget the app users with a personalized message based on the products seen inside the advertiser’s app. We hope to kick start this new personalized messaging for app user retargeting as we get ready for the new year. Please get in touch with your Vizury Account Manager or email us at to know more on how to get started with Vizury mobile app user retargeting.

Retargeting and Remarketing guide for Marketer

Retargeting in holiday season

Who does not love a holiday season? E-commerce industry loves it too; it welcomes its customers with great offers and deals. During the Black Friday event in Brazil, we saw a significant increase in sales across major players. The sales grew by ~9 times during this period for one of the major electronics e-commerce websites.

Below are a couple of graphs highlighting the spike during Black Friday event:


So how can re-targeting be used as an effective tool in comparison to the other digital marketing channels during the holiday season?

1. Use of historical data to show relevant products and create segments based on user-behaviour

It is interesting to note that the evaluation period for the customers on this website for the sale period is around one week; where we observe an increase in website traffic. During this time, the potential customer is in research phase and still deciding on which products to buy. The sale may not be great during this week. However, the users who have spent considerable time researching are very important. This is where retargeting comes into play.

Any user who has seen products on the website and spent time on researching about the products is more likely to buy during sale period. Retargeting acts as a trigger to this action and gives the final push to the decision making process by showcasing customized banners(based on past browsing history)

Vizury creates customer segments based on multiple user-behaviour attributes and focuses on users with highest probability to buy, thus maximizing the ROI for the client. For instance: One of the important attributes could be the Average order value (AOV) .It can be clearly seen from the below graph that AOV increases during the sale period. Vizury prioritizes users who have abandoned cart and have a high basket value.


2. Push Products

By definition, these are the list of products which clients want to promote aggressively (usually products with high discounts and products which are introduced during the event).

Performance of push products during sale period:


Around 200 client recommended products contributed to more than 45% of the website sales during the holiday season for a major e-commerce website. Any marketer would want to spend the maximum $ on the marketing activities of these products. Vizury’s push product algorithm ensures that such products (based on their category) are prioritized over other products and are shown more often on banners. We also chase users visiting these products aggressively.

3. Category Analysis

Distribution of sales by category:


Customers associate a certain website to products of a particular category. For e.g. Website XX has the highest discount on products of category YY. Thus, top selling categories on the website generally has the highest sales during such an event. We prioritise the products of these categories over others in combination with focusing on categories with highest % increase in sale.

4. Target wider population

The user intent to purchase on the website varies inversely with the time since dropped off. However, during sale period it makes sense to target and bring back the customers who have not visited the website recently. The effectiveness of targeting these customers increases by showing the push products in our banners.

5. Use of Social Channel

Social media has been the new nitro-boost that re-targeters were waiting for.

  • Social media like FBX helps in engaging customers for a longer period than other marketing channels.
  • Social media has a great influence on buying behaviour of potential customers. The CVR on social media is significantly higher than many of our conventional marketing channels.
  • Social media has shot up the overall unique reach by ~20-30% in most campaigns.

6. Communication about the website offers through customized banners

  • Vizury helps in showing customised offer messages in banners to which in turn increases the CTR.
  • Banner theme matches the overall website theme during the sale period; to ensure higher brand recall.

Customised Black Friday banner examples:



To Summarize – During the holiday season, re-targeting offers multiple levers to marketers to invest in the right customers and display relevant products. When used effectively, the contribution of re-targeting (as a marketing tool) to the website sales increases many fold. We look forward to a great sale period during the upcoming holiday season while continuously innovating to help our clients succeed.

Merry Christmas and Happy holidays! We will see you next year.

Happy new year and a 2013 summary of Vizury news

A very happy new year to all our friends and readers from everyone at Vizury. 2013 was a great year for us and we hope it was the same for you as well. And also, here’s to all hopes that 2014 will be even better!

Below, I have summarized the top 5 things that happened to us / we did well in 2013. For this we have to thank in equal measures, every Vizurian, our customers, our leadership, our investors, and of course, all our families and well-wishers.

(In no particular order)

    • We went mobile and we rocked it! Our mobile retargeting solution, MobiConvert, was launched in 2013. We hope you caught our mobile product manager, Shiju’s blog here and also his article here on mobile retargeting. Since its launch, MobiConvert has also grown to include the ability to place retargeting ads within mobile apps. Also, customers have really taken to it with great interest and excitement. A Japanese customer’s results from their mobile retargeting campaign were also very encouraging.
    • The business grew fast and furious! Our revenue multiplied almost 3X, we increased our customer base almost 40% and customers on an average spent 70% more with Vizury solutions than they did in 2012. Phew! Much of it was thanks to aggressive geo-expansion, increasing our product portfolio and also implementing operational innovations to improve campaign throughput. It was business growth at breakneck speed but there’s not a minute of it we did not love! A roller-coaster ride we are looking forward to repeat and beat this year.
    • There are so many more Vizurians now! We started the year with about 90 Vizurians and now the number has more than doubled. Vizurians in offices across the globe (other than Bangalore) have gone up more than 150%! In Bangalore we’ve expanded space to occupy the entire building (where we had just 2 floors to begin with) and now it’s common to see Vizurians at the lunch table exchanging introductions.

There you go. I’ve constrained myself to 5 things that happened in 2013, along the lines of the 5 predictions I made for retargeting in 2014 in the ClickZ article.

Now, can you predict 5 things that Vizury will do in 2014?

Vizury’s core values

(Co-authored by Arjun A.V.)

Culture takes over when the CEO leaves the room” – Frances Frei and Anne Morriss

You will agree that the value system and beliefs of an organization determine its growth path and the core values of an organization are reflected in the way its business is run. Also, as one treads new milestones in business and grows as a team especially at the pace of Vizury (our global headcount has grown by >50% in the last 12 months) – the team is bound to include individuals from various backgrounds. This makes it essential to define a set of values and beliefs that act as the vertebrae that makes individuals function together as a team.

We always believed that a common set of values was the driving force behind Vizury’s culture and so we set out to define it. We knew it would be challenging and yet of great significance to us. It might have been easier to use generic values shared across leading businesses as our core values, but this would not serve the purpose. The intention was to list down the values that all of us at Vizury uphold and will continue to stand for.

Core values are relevant only when employees believe in them, identify with them and represent them in their day-to-day work and interactions. Bearing these thoughts in mind, we came up with a four-stage process to define Vizury’s core values. A couple of articles helped us shape this process, namely “Building your company’s vision” and “10 steps for developing your company’s core values”.

STAGE 1: Identify key personal values

Employees are the building blocks of an organisation and the personal values of its employees collectively form the core values of the organisation. So, we selected a group of ten employees from various functions at Vizury and asked them to share the core values that they personally believe in and strive to live by along with a brief description for each value.

These values were then scored based on how frequently they were mentioned by this group. Through this activity, we were able to identify the top seven personal values.


STAGE 2: Test and distill shortlisted values

We then asked our managers to evaluate each of the shortlisted values from stage 1 based on how well their teams lived by these values:

  • Is the value essential towards being a ‘role model’ within the company as well as to the external world?
  • Will the absence of the value make an employee a poor representative of Vizury?

Views or ideas on any new values were also solicited. Each value was scored based on the number of votes received for both questions. This was necessary to understand the relevance of each of the shortlisted core values to Vizury employees and teams. In this step “Customer Centricity” emerged as a new addition to our shortlisted values.


STAGE 3: Ranking the values

In order to validate the list and identify the top four core values, a larger group of employees were asked to rank each of the shortlisted values from stage 2 based on their relevance to Vizury and anyone working here. The scores were aggregated for each value and new values were also solicited along with a rank. We could list down four core values based on these rankings as shown below.


STAGE 4: Feedback from everyone

This was the most crucial step in the process. We shared the short-listed values from stage 3 across the entire organization and solicited feedback. We received many enthusiastic responses to this outreach.
Most employees identified with the values shortlisted and also suggested additional values which they felt were indispensable as Vizurians.

“Customer Centricity” which was eliminated after Stage 3 was recognized by many as a key value that was missing in the shortlist! So “Customer Centricity” was back on our list along with “Team Work”!


After the exercise which took more than a month to complete, we were able to define the core values that are an integral part of Vizury’s work culture. All of us at Vizury relate to and understand each of the values. Everyone at Vizury has had a role to play in choosing and defining them and we look forward to adding many new Vizurians to our team who exemplify these values.

Our Core Values


Vizury helped leading online booking site in Japan attain 4X higher conversions on mobile

Mobile retargeting is an important marketing avenue for advertisers today, especially in markets like Japan where the number of mobile users is ticking up at neck-breaking speed. One of Japan’s leading hotel and ryokan booking website wanted to retarget a niche segment of users based on parameters such age, and frequency and recency of visits.

Vizury MobiConvert utilized privacy compliant user profiling techniques to identify users of the desired age group with recent visits to the advertiser website. Through RTB optimizations, we retargeted these users with highly pertinent ads resulting in a 4X higher conversion rates. Here’s our story.

mobile retargeting in japan


10 cool ideas that can make Facebook Dynamic Ads work for you

Every marketer is eyeing Facebook ads as an important channel to reach out to potential customers. Facebook now supports dynamic ads through Facebook Exchange (FBX). This means you can now show dynamically customized ads on Facebook based on available customer data. While it gives you the power to super-optimize your ad spends, it also comes with the risk of not being able to unleash its true potential.

For the benefit of those who have not used Facebook dynamic ads yet, here is some brief background. The Facebook ad format is noticeably very different from the traditional flash based banner ads. Unlike a flash-based, dynamic ad where you serve a pre-designed .swf generated on the run-time during ad serving, a Facebook dynamic ad is served using four different elements each of which need to be customized at the bidding time itself. These elements are the 100×72 pixel image, the max 25 character Title, the max 90 character Body and the Landing page URL.


Vizury directly integrated with FBX a couple of months ago. As we scaled our business on Facebook ads, we realized this inventory is quite different from the traditional banner ad inventory and so the marketing strategy for this channel ought to be markedly different as well. Here is an attempt to capture 10 ideas which can help you harness the maximum potential out of your dynamic ads on Facebook.


(1) Crop out the whitespace

You have just 100×72 pixels for the image. It is a small size and hence your real estate is precious. You probably cannot afford to have a lot of whitespace in the image. Crop out the whitespace before scaling it to 100×72 size so that you get the maximum attention from your customers.



(2) Remember that Facebook ads auto-refresh

Unlike traditional banner ads, Facebook ads automatically refresh themselves at a frequency of a couple of minutes. The average Facebook user spends considerable time on the site, which means you run the risk of showing a lot of ads to the same user back-to-back and burning your impressions planned for that user really fast. You want to have Facebook-specific impression frequency caps in place to avoid this.

(3) Don’t rely on cookies, start storing user data in your systems

Remarketers rely too much on the cookies to store and read user data to customize their banner ads. However, this method will not work well with Facebook. Here you have to send the ad content along with the impression bid itself, so you do not have access to user cookie data when you customize the ad. Ensure that your back-end systems also store user data so that you can use it to customize Facebook ads.

(4) Have a preview mechanism

Since the text content of the ad copy gets customized during run-time, there are chances that you end up showing some unwanted text by mistake. E.g. If your system has stored price data along with currency, your customized ad might show something like “Jewellery at Rs. Rs. 500”. You do not want that. Have an internal preview mechanism so that you can verify your customized ads much before they go on to the users’ Facebook page.

(5) Try multiple themes and see what works

Facebook ads are as much about text advertising as they are about banner advertising. With a fully customizable title and text, you can try out various themes as different campaigns and see which one yields better results for you. Themes can be about discounts, prices, collections, etc. For example, you can start following customized campaigns for t-shirts “Buy t-shirts at Rs. 799”, “30% Discount on t-shirts”, “Buy t-shirts of size L” where price, discount, and size are fetched in the real time. Compare their performance and scale the best one. However, ensure that your title, body, image and landing page provides a coherent message to the user.

(6) Send responses for multiple advertisers

Another unique feature of Facebook inventory is that it sends you a single bid request for all the ad slots of a page viewed by a user. This is very different from traditional banner ad inventory where every request comes only for a single ad slot, even if there are multiple ad slots on the page. So, if you are a 3rd party marketer working with multiple advertisers, you can utilize this opportunity to send responses for multiple advertisers. All these ads will participate in the same auction and none/some/all of your ads can win ad slots on that page. This way you can create a lot more ad inventory for yourself. Just ensure that you do not send multiple responses from the same advertiser as Facebook does not allow that.

(7) Think if you really need to track ad-slot position

Facebook allows you to record the ad spot position won for every impression. It is fascinating to see which slot is high performing and which is not. However, you need to realize that currently, this information is not actionable. Facebook also themselves decide which slot position you will get and you cannot control it in any way. So there is no point in wasting your time bothering which is the best slot position and why am you are not winning it more often.

(8) Ensure correct # of likes

Facebook ads have the option to show the number of likes on the advertiser’s Facebook page. If you activate this option, Facebook automatically figures out the advertiser’s Facebook page based on the domain of the landing page URL. This approach can work fine for many advertisers while it might not work correctly for some others. The best way to ensure it works is to figure out the Facebook page id of the advertiser and force feed this id to the advertiser ad group settings. This will ensure that number of likes being shown is always correct.


(9) Make your systems partial-customization-ready

FB dynamics ads are generated in the real-time using customized variables from your system. Since values of these variables can vary significantly based on the product, you may run into situations where it will not be possible to customize all the elements of the ad for certain products. Wherever full customization is not possible, make your systems ready to perform partial customization. To enable this, perform all the validation checks at your end and if validation fails for any element, replace only that element with some generic value that suits all your products. For example, a title “Buy a Versace Gown for Rs. 10000” is invalid because of 25 characters limit. In such cases, your system can replace it with a default title such as “Change your wardrobe!” This way, all other elements still remain customized and increase the probability of a click.

(10) Utilize Image for the best



The first thing a customer notices in a Facebook ad is the image. The image has the real power to attract clicks. For certain verticals like Flights where you think image customization is not possible and you end up showing generic logos, you can actually customize the image as well. You can build a system for auto-generation of a text image in the real time. A customer looking to fly from Bangalore to Delhi on 14th Jan is much more likely to click on the ad based on the customized image.

The gist is that Facebook dynamic advertising is altogether a new channel and marketers should not confuse it as an extension of their regular banner advertising campaigns. We need to start thinking about this channel with a fresh perspective.

If you find this blog useful, have questions or comments do use the section below or email us at

How user behaviour is shaping mobile retargeting?

As mobile takes centrestage in a user’s purchase life cycle, marketers have quickly shifted gears to maximise user engagement through mobile assets. Hence, mobile retargeting has found easy acceptance among marketers who have witnessed the benefits on the desktop and are looking to monetise their mobile marketing ad spends. But mobile is different from desktop retargeting. Yes, there are a few technology-based differences that need to be addressed. Apart from these, one key factor that impacts mobile retargeting is the ‘changed’ user behaviour. Here are some important factors that determine why the mobile user is different and how your mobile retargeting strategy can address these behavioural differences.

Vizury Mobile Retargeting

Access multiple mobile channels and mobile devices
Unlike the desktop, you have multiple assets on the mobile – website and apps. You will know that both of them have different eco-systems and function independently at a single user level. However, users expect a seamless interaction across these assets and hence it is essential to identify users as they traverse between mobile web – mobile apps and collectively assess user behaviour data displayed across both channels. Also, many times multiple mobile devices (smartphones and tablets) are used to access assets. A strategy combining mobile web and app retargeting that employs accurate device identification techniques can identify users across mobile marketing channels.

Mobile – desktop users
Mobile web traffic is increasing at a neck-breaking speed and the rate of conversions through mobile are on the rise as well. However, many users prefer to buy on the desktop and use mobile for the pre-purchase research. Ease of use(click vs touch), display size, payment security qualms – there are various reasons for this. However, understanding this behaviour points out the need for an effective multi-screen retargeting strategy that will trace the user’s journey across devices and channels to ensure accurate attribution to various touch points.

Secondary screen syndrome
For a multi-screening user, mobile is mostly the secondary screen. The primary screen might be a television or even a desktop. In such cases, the user will have shorter attention spans and tends to switch between the screens in the blink of an eye. Engaging this fickle user is a challenge that can be achieved with precise targeting, accurate recommendations and crisp ads.

Shorter visits and longer purchase cycles
Mobiles are accessible anytime. It is what keeps people occupied while waiting for a bus or an elevator to arrive? Most smartphone users resort to a quick browse/game in these short waiting periods. While this ‘anytime access to user’ is the prime reason for most marketers to have shifted focus towards mobile, it also means that a single visit to your website/app might last just a few seconds. The user might re-visit the website/app many more times before making a purchase decision. Higher number of re-visits and longer buying cycles when compared to the desktop are typical of mobile users. Mobile retargeting solutions with Deep Linking capabilities will ensure that a user who clicks on your ad will land on the product page and not the home page. Apart from reducing the user’s browsing time, Deep Linking also enhances the user-experience.

It’s a personal device
When you reach out to a user on a mobile device it is important to understand that most users treat mobile as a personal device. A user who loves personalised ads on the desktop might not feel the same about personalised mobile ads. This is because most smartphone users, especially in emerging ecommerce markets perceive mobile as a device primarily meant for personal communication. So it is essential to tread carefully and avoid invasions into user privacy through privacy compliant retargeting practices.

To conclude:
User behaviour has a huge impact on mobile retargeting and it is essential to evolve a retargeting strategy that addresses them.

  1. Users are increasingly multiscreening, you must include a cross-channel, cross-device or even better a multiscreen retargeting strategy for effective user engagement.
  2. Crisp ads with accurate recommendations work best and assure great results.
  3. Deep Linking helps reduce user browsing time and enhance user experience.
  4. It is important to ensure privacy compliant retargeting.

Authored by: Akshatha Kamath

Original article published on Digital Market Asia

Image courtesy: Digital Market Asia

Overview of Recommendation Systems at Vizury

Recommendation systems work on the premise that when past performance data is fed into a computer the resulting output is a prediction of the future performances. Developing an effective recommendation system is the biggest challenge in retargeting. The impetus lies on determining the most preferred product/service for an individual user and we often deal with such large and diverse amounts of data that the “one-size-fits-all” approach does not work.

Need for a recommendation system

During the initial days of retargeting, ads were customized based on a few fixed parameters that were common to the campaign and did not necessarily reflect the user’s interests and preferences. However, for effective retargeting, our ads have to be engaging and grab the user’s attention through product/service offerings in the ad that appeal to the user. With advancements in web technology, retargeting has evolved and it is possible today to customize our ads for an individual user based on his/her interests and preferences.

Implementation of recommended systems

Vizury has been experimenting with various recommendations systems for verticals like ecommerce, hotels, flights etc. The first experiment was developed on the principles of collaborative filtering (CF) and worked on the assumption that if there is a huge affinity for two products among users visiting an advertiser website, a user who has seen one of the products is likely to be interested in the second product as well. This is one of the most common algorithms developed in this field and it performs really well. There are other algorithms which are either variants of CF or are proprietary to Vizury. These algorithms work better than the previously used technique and have resulted in a boost in online sales for advertisers and at the same time have enhanced the user’s online experience.

Identifying the best performing experiment

With the recommendation system live on multiple campaigns it is extremely difficult to manually identify the best experiment for each campaign. We have developed the Homing Algorithm to automatically identify the experiments that perform well for a specific campaign. The algorithm takes into account the business model of the campaign, analyzes the data and changes the weights of each algorithm for an individual campaign. We are running the initial experiments with the Homing Algorithm and expect the results to flow in soon. With this development, we also hope to deliver better returns for our clients. Watch out this space for the results of our experimentations that we promise!


Useful links