Bounce Rate, Leakage Rate, Abandon Rate

I was in a meeting with the Google Analytics team yesterday and heard a lot of these metrics. I had heard about the abandon rate and also leakage rate but Bounce Rate was something new to me. I had seen this in Google Analytics and had always interpreted it in a different way.

The Google gurus explained me that "Bounce Rate" has been termed as the "SEXIST" metric of website analytics. Bounce rate means indicates the % of visitors to your site who exit the site by viewing only one page i.e. they did not get to any other page of the site and left the site.

From a business application point of view, one would want to think that if a lot of consumers are exiting your site from a particular page, then probably there is something that needs to be fixed on that page. And if this page was the homepage or the landing page to which the online media is driving traffic then there is definitely a problem.

Some other applications:

For a retailer if this is the start shopping page and then people are leaving the site – big problem!

For a financial services or an insurance company – consumers leaving the homepage or the account login page – big problem!

For a branding website – consumers leaving the landing page to which media is driving to – your branding is not working!

How does Digital Ad Serving Work?

How do all these banner ads get placed and how does this work? First let's talk about what is an Ad Server, as per ZDNet's dictionary, an Ad server is defined as a "Web-based server that delivers banner ads and popup ads to the requesting Web pages. For companies that sell their own ads, the adserver may be an inhouse or co-located machine at an ISP, or it may be owned by an Internet advertising company".

The Ad publisher or the site where the ad will be placed contacts the Ad server to get the creative every time the page loads. Based on how the inventory for that particular placement has been sold, the ad server provides the particular creative to be published. This communication happens real time and sometimes slows down the page load. This is the main reason by sites like MSN, AOL do not want to use third party ad serving on their key homepage placements as they have over 100MM impressions in a day and if the page load is slow, they might lose consumers coming to their site.

The reporting part of the Ad server then captures the number of impressions, clicks, post-click activities, post-impression activities and all the other standard metrics based on how and what things have been tagged.

Cookie Window, what is it, what is the optimal length?

A lot of times we come across the term – “Cookie Window” or media folks talking – “There is a 30 day cookie window on this campaign”. What does that mean? This means that consumers who clicked on a digital banner or a search result have 30 days to return to your site and take the action that is being tracked. If the consumer had clicked on the ad, it will be shown in post-click activity and if the consumer had not clicked on the ad, it will be tracked as post-impression activity.
The optimal length of the window should be chosen based on each individual campaign or business, it should be tested and then the correct decision should be made. Typically, a click to conversion or action lag report is available from the ad server like DART or Atlas which shows the time tag between the click and the conversion or action. This is very important for DR (direct response) businesses which are trying to either sell something on their website or have a clearly defined action as then they can adjust the cookie window based on that.


As shown from the chart above, this is a clear increase in the number of actions up to day 10 and then there is a decline through day 15 and after that there are minimal actions. Thus, looking at this chart a cookie window of 15 is the optimal length for this business.

Launch Campaign, Measure and then….

I have worked on a lot of online campaigns, majority of the times the campaign is launched, and then the analytics team typically starts reporting on the KPIs (Key performance indicators) on a weekly or bi-weekly or monthly level.

The key question is what does the strategy or the marketing team does after getting the results. The marketing team should look at the reports carefully at the most granular level (placement or creative) for display campaign and then start optimizing based on the results. Sometimes combination of metrics needs to be taken into account for example the cost, engagement rate, interaction rate, conversion rate etc.

It is very important that the optimization takes place sooner rather than later. The optimization process should be establish and then followed week over week. I have seen that this is very important for campaigns whose key success metrics are ROAS or ROI. As the campaign manager is trying to generate highest revenue for each dollar spent.
If the campaign’s purpose is to increase the branding/engagement then it is important that the placements driving higher Click-through or Interaction rate should be served higher impressions. The relevant publishing sites should be used. The capability of Ad networks like Ad.com, Specific Media, and Context Web should be fully utilized to get lower priced media and target the relevant consumer targets.

Some portals like Yahoo! are also able to use Behavioral Targeted placements to better target the consumers to help increase the brand awareness and also increase the conversion rates.

The key is to start optimizing sooner rather than later..

Importance of Setting up correct Tracking – Always test before going Live

In my previous posts I have stressed about setting up tracking correctly. I think I found an example to help explain that today. If tracking is not set up correctly, one may end up reporting metrics which are “Too good to be true”.
What if one fine day you pull a impression click report from your Ad server, example DART and get 1,000 impressions and 500 clicks. You will be like, 50% Click-through Rate. This ad must be a rock star! Every other person who was exposed to this ad, clicked on it. When I last checked the average click-through rates, I think they were in the 0.1%-0.2% range. Where did this 50% come from?
First thing I do is contact my Ad ops team to check if the tracking was set up correctly, did the publisher implement the tags correctly. To me, this seems to be a case where either:
1. Impressions were not being counted correctly OR
2. Clicks were being counted 50x
3. I have also seen instances where for every impressions a click was being counted – somehow the click tag was firing with every impression
Most of the times, it ends up being a tracking issue or the site did not implement the tags correctly.
These issues are usually an easy fix but what if there was a one day placement (the ad was supposed to be live for only 24 hrs) and there was an error. The data reported would be incorrect and there is no time to fix this issue. We all know that some of these one day placements at major sites like MSN, AOL, Yahoo!, CNET, CNN etc cost a lot of money – a few hundred thousand dollars per day. This is a lot of money and missing out tracking on so much money is never good.
So, again please always make sure the tracking is set up correctly and ask your Ad ops team to test before going live.

Measuring Multi-Channel Campaign

It is very common for an advertiser to have a campaign that targets consumers across various channels - DRTV, Email, Display, Search, Billboards etc.

One of the most important thing while measuring from all of these channels is setting up the correct data requirements, tagging etc so that accurate data is available at the desired level to be able to measure the success metrics.

Each channel would also have a different success metric for example in Display, it is very easy to attribute conversion up to the placement level but in case of DRTV, it is very hard. If the campaign involves 5 networks with 2 creatives and running at all day parts - then the number of unique 800 numbers would be very large. Sometimes, it is possible to get them but it can be a challenge even with huge advertisers.

The case explained about DRTV applies to all the offline channels - Billboards, Print media etc

Online is so easy to measure and is as accurate one can get in measuring marketing efforts.

While looking at a holistic picture, it is possible to identify the peaks and valleys based on what media went live at what time. It helps establish the cause and effect relationship across media channels. Sometimes, we see a 1:1 relationship but sometimes it is not worth using a particular channel.

In my opinion, using Email to acquire consumers is not a very good idea, they should be used to maintain a relationship with the consumer. One example could be maintaining a relation with a patient or a car manufacturer could keep sending service updates about the car.

Aarav Prakash

I have been not posting anything on the blog for the last few days, as I am busy taking care of my newborn boy - Aarav Prakash :)

He came on April 27, 2008 at 1334 hrs.

It is funny, I keep thinking do I need to start creating a dashboard trending his number of diapers or the amount of food he is eating :)

He is showing a day over day increase in both of those metrics...there seems to be a 1:1 co-relation.