Three Tips for Choosing the Colors for Data Visualization

Colors can add to your data visualization and make the story stand out or they can distract the audience from the main message. It all depends on what colors you chose and how you use them. In this post, I am listing three things to keep in mind when creating your next data visualization. I am not guiding you specifically on what colors (hue, value and chroma) to use but three things to keep in mind when selecting the colors:

1. Use conventional colors:  Generally red color means negative or a bad value and green means positive or a good value.  Choosing the default colors as presented by your visualization tool might not convey this meaning in all the cases.  Say, for example in Tableau, when you choose the default “Red-Green Diverging” color from the pallet, the red represents the smallest number while green represents the largest number, and it changes from red to green for the values in between. This generally works fine.  Where this becomes an issue is when smaller numbers are actually better than the larger numbers. For example, when you are showing Cost Per Click, lower is better so it should be Green while higher numbers are bad, hence they should be Red. Leaving the default will show low CPC as red, while higher CPC will be green, this will convey the wrong message to someone who is just glancing at the visual. To fix this issue in this case, all you have to do is just check the box, titled “Reversed” to reverse the colors, now Red will be used for higher value.
2. Don’t use too many colors: Using too many colors can distract the consumer of the visualization from the main message. Minimize the number of colors you use in your visualization. (See guidelines for selecting colors section in “Expert Color Choices for Presenting Data” paper by Maureen Stone)
3. Make main message/data to stand out:  Make sure the color combinations you chose enables you to make the main message to stand out. For example, white on black will stand out while grey on black might be lost.

Subscribe to my mailing list

* indicates required

Email Address *

First Name

Last Name

5 Data Quality Issues To Watch Out For

We all know that wrong data leads to wrong analysis and hence can cost a lot of money.   In this post I am highlighting 5 issues that lead to wrong data and hence wrong analysis.  Make sure you take care of these issues before spending any time on analyzing and putting your presentation together.

1. Manual entry by end users – When you rely on the end users/customer to enter the data in the free form you will get lot of variations of the data. For example, when you ask them to enter the city, you can get variations such as Redmond, Redmond WA.  Redmond (Seattle), Remond etc.  Doing any analysis on the city level will pose an issues since you have several variation of the same city. Unless you have thoroughly cleaned and accounted for every possible variation you will not have the right analysis. In order to avoid such issues, wherever possible provide the choices via drop down or auto-fill rather than letting users type in the answers.
2. Manual entry by someone in the organization – Though this is more controlled than letting end users enter the data, you still have issues similar to number 1 above.
3. Excel sheets (calculations) – This becomes an issue when you have lot of calculated columns and some of them are dependent on the other calculated columns and sheets.  One simple mistake can cascade to multiple columns and can mess up all your analysis.  Whenever possible, do not rely on calculated columns (prepared by someone else) in excel sheet, just use the raw data and do your own calculations so that you can stand behind them.
4. Data Imports, connections and processing – In most companies the data resides in various places – databases, excel sheets, flat files Hadoop, 3rd parties etc. For you to get a 360 views you will need to collect and combine the data for all the sources. The data corruption can occur at several places including but not limited to mapping wrong keys, missing some key data, writing wrong queries, importing partial data etc.  You should always verify the data that you are getting and make sure that it is clean and complete. Since there are several owners of the data this is not always an easy task, particularly in the large organizations.  There is not much that you can do on daily basis to verify the quality of the data, but make sure you understand the underlying data sources and have good understanding of the process that is used to combine the data.  Make sure that you in loop on any changes that are being made to the data sources and the process. You can not afford for your raw material (data) to be of a low quality.
5. Visualization tools – You expect these tools to work, don’t you? However be careful and double check your calculations before you present your data.  I recently had two issues that made me believe that there are many of you who will run into these issues and might present the wrong analysis.
   • When you use averages makes sure that you are using the right columns.  Average of an average is not the same as average by summing the values in individual rows.  Recently I ran into a situation where CPM (Cost Per Thousand Impressions) was already calculated in the excel sheet (see Issue no. 3 above).  When that data was brought over in Tableau, the analyst, in an effort to find average CPM, used the calculated column to compute average CPM.  Everything looked good on the surface but it was wrong since it calculated average of an average.
   • When using a map make sure you use the correct values of Longitude and Latitude. I saw an example where the map showed up perfectly, however a quick quality check showed that the average value for a state was more than the individual values of all cities in that state, which is not possible.  On further investigation we found that the issue was with the way Longitude and Latitude were used to render that map. Once the issue was fixed, everything worked fine.

I would love to hear from you if you have encountered these or any other sources of data quality issues.

A Simple Fix To Grow Your Email List

Are you loosing your email subscribers right after they are signing up? If yes then you might be making a similar mistake that I encountered recently. Here is what happened to me:
I submitted my info on coolinfographics.com to download a free chapter of the book. After I filled the form, I got a message that I will get an email with a link to the download.  I eagerly waited for the email to arrive. As soon as I got the email, I opened it and quickly clicked on a very prominent and long link. I was expecting to see a page where I will either be able to download the chapter or my download will automatically start. Instead I got the following message:

What???? What just happened?
Turns out that the long prominent link in this email was not the link to download the sample chapter, instead it was an Unsubscribe link.  So I accidentally unsubscribed from the mailing list I subscribed only few minutes ago.  I am sure, I am not the only one who has done that. I wonder how many subscribers is this site is losing due the design of this email?

Here are two mistakes in this email

1. The unsubscribe link is bigger and more prominent than the actual download link (main Call to Action).  My eyes directly went to the long link and without even thinking twice, I clicked on it.
2. Transactional Emails do not require you to have an unsubscribe link (See http://www.spamresource.com/2009/12/is-unsubscribe-link-required.html). Since this is a transactional email so you don’t need give an option to unsubscribe.  You can argue that it is a best practice to provide an Unsubscribe link. Sure if you want to provide an option then make it a small in font and length.

Fixing the problem:

Simple solution is to remove the attention from Unsubscribe link and make the main Call to Action (Link to Download) stand out, see below for a revised version:

  

What do you think? Questions? Comments?

BTW: If you are interested in data visualization then go ahead and check the sample chapter or buy the Cool Infographics: Effective Communication with Data Visualization and Design at Amazon.

5 Reasons Why Your Display Advertising Is Not Working

Are you one of those advertiser, who is struggling to understand why your Display Advertising is not providing the desired results?  If answer is yes, then this post is for you. Below are 4 most common reasons why your display advertising might not be working.

1. Ad Views: According to a study by Sticky,  77% of ads are never seen by people. Even when the ad is considered viewable, meaning it is within viewing area, only 55% ads are actually viewable. Which results in a very lower click through rate, the average banner CTR is about 0 .1% and declining.
2. Spider and Bot Ad Clicks:  Spider and bots, instead of humans, make up a significant amount of clicks on the ads. All these spiders do is click on an ad, land on your site and then leave causing millions of dollars in fraudulent clicks.  As a result you will either see a very high bounce rate on your pages and/or mismatch in the clicks reported by ad network and visits reported by your Web Analytics solution.  In 2012, a start-up reported that about 80% of their clicks from Facebook ads were by spiders. Another study found that 20% -90% of clicks on some sites were via spiders. I also showed an example of a bot in my post, 4 Reason Why Your Bounce Rate Might Be Wrong
3. Fat Finger: Over 35% of the ad clicks on Mobile are by accident, again causing high Bounce Rate.
4. Mismatched Landing Experience:  Make it a seamless and consistent experience from your banner to conversion. Users don’t have time so make it right the moment they land on your site. For example, If a banner ad promotes “Free Trial” then make sure landing page make it easy for user to sign up for the free trial. Don’t expect the users to click through to your site to find where the “Free Trail” page is.  Mismatched landing page and ad experience leads to High Bounce Rate and Low Conversion Rate.
5. Site Speed: Slow site speed breaks visitors flow from a display ad to your site. If it takes too long for the page to load then the visitor will be gone before she sees the full page.  In this case you will see a clicks but not visits and/or high bounce rate.

21 Metrics for Measuring Online Display Advertising

In this post I am listing the 21 metrics to measure the success of your display advertising.  Most of these are also applicable, with some variation, to other forms of advertising such as Paid Search, Social Media Ads, Print and email. I will cover these other channels and mediums in the future posts.

1. Impressions – It is the number of times your ad is displayed. The number by itself does not hold much value but it is a metric used to calculate other metrics and KPIs. Keep in mind that an impression does not mean that someone actually saw the ad, it just that the ad was shown on a web page/app.
2. Reach –This is the number of unique people (generally identified by cookies) that were reached by your ad. This number is always lower than the impressions because your ad is generally shown to same person (cookie) multiple times.
3. Cost – The total cost of running the ad campaigns.  This is calculated differently by different tools and organizations. Some use actual media cost while other use a fully load number that includes the agency cost, creative cost etc. Whichever number you use, be consistent in your approach. If you are going to do comparisons with CPC models such as Paid Search then I suggest using the actual media cost. Most of the publicly available benchmarks are based on actual media cost and are expressed in CPM (explained later in this list).
4. Engagement Rate or Interaction Rate– This applies to the Rich Media Ads, where a user can interact with the ad without leaving the Ad unit/widget.  Engagement Rate is the percentage of interactions per impression of the ad unit and is calculated as (Number of Interactions/Total Impressions)*100%.
5. CPM – This is the cost for 1000 Impressions of the ad unit. Display advertising is generally sold on CPM basis. (For more information on CPM, see  Cost of Advertising: CPM, CPC and eCPM Demystified).
6. Clicks – Number of clicks on an ad unit that lead to a person leaving the ad unit.  Keep in mind that a click does not mean that a person landed on the intended destination of the banner ad click. There are multiple factors that could lead to a click but not a visit to the destination (I won’t cover those here but am happy to discuss over email or a call).
7. CTR (Click though rate) – It is the number of Clicks generated per impression of a banner ad. This number is expressed as a percentage. CTR = (click/impressions)*100%
8. CPC – Cost per Clicks is the cost that you pay for each click.  Generally, display advertising is sold by CMP (see above), you can easily convert the cost in to Cost Per Click to compare it against other channels such as paid search. Cost per click is the effective amount you paid to get a click.  It is calculated by dividing the cost with number of clicks.  CPC = Cost/Clicks. Sometime this number is also referred as eCPC (effective Cost per Click).
9. Visits – As stated above in the definition of clicks, not every click turns into a person landing on your destination (generally your website). Visits measures the clicks that did end up on your site.  (For more definition of visits, please see Page Views, Visitors, Visits and Hits Demystified)
10. Visitors – Visitors metric goes one step ahead of the visits and calculates the number of people (as identified by cookies) who ended up on your site as a results of the clicks on the banner ads.
11. Bounce Rate – Is the percentage of visits that left without taking any actions on your site. It is calculated as Number of Visits with one page view /Total number of visits resulting from the display ads. (Bounce Rate Demystified for further explanation).
12. Engaged Visit Rate – Generally this is opposite of bounce rate (though you can have your own definitions of engagement).  It measure the quality of the visits arriving from your display advertising. You can calculate Engaged Visits as  (100 – Bounce Rate expressed as percentage).
13. Cost/Engaged Visit – This is effective cost of each engaged visits. It is calculated as total Cost divided by number of engaged visits.
14. Page Views/Visit – Page views the number of pages on your site viewed by each visit. With a lot interactions happening on one single page, this metrics is losing its value. However, for now, it is still a valuable metric for ad supported sites.
15. Cost/Page View – As above, this is valuable metrics for ad supported site to figure out the cost of generating on extra page view.
16. Conversions – Conversion is defined as the count of action that you want the visitors to take when they arrive from you display ads. Some examples of conversions are – purchase, signup for newsletter, download a whitepaper, sign up for an event, Lead from completions etc.
17. Conversion Rate  – This is the percentage of visits that resulted in the desired conversion actions.  Conversion Rate = Total conversions/visits*100. If you have more than one conversion actions then you should do this calculation for each one of the action as well for all the actions combined.  In case of Leads, you can take it one step further and calculate not only the “Leads Generation Rate” (Online Conversion Rate) but also Lead Conversion Rate, which is, Leads that convert to a customer divided by total leads generated.
18. Cost per Conversion – This is the Total Cost divided by the number of conversions achieved from visits coming via display ads.
19. Revenue – This is total revenue that is directly attributed to the visits coming from display advertising. It is pretty straightforward to calculate in eCommerce but gets a little tricky when you have offline conversions.
20. Revenue per Visit   – Shows the direct revenue achieved per visit originating from the display advertising. It is calculated as Revenue Generated from Display Ads divided by the total Visits.
21. Revenue per Page – This is useful for ad supported business models. This is sometimes expressed as RPM (Revenue per thousand impressions of ads) = (Total Ad Revenue/Number of page views) * 1000

Note: In addition to Clicks, you can also looks at View Through and calculate your other related metrics by view through.  View Through is sum of all the cookies that visited a page that showed your ad on it, and then landed on your site. The assumption, in this calculation, is that you landed on the brands site because of that ad exposure.
 Where can you get these metrics from?

• Impressions, Reach, Cost, Engagement Rate, Clicks, CTR and CPC data is available from your agency or ad server tool.
• Visits, Visitors, Page Views, Bounce Rate, Engaged Visit Rate, Conversion, and Conversion Rate are available in your Web Analytics tool.
• Revenue is available in either your Web Analytics tool or other offline sales database.
• Cost/Conversion, Cost/Engaged Visits, Cost/Page view and Revenue/page are calculated using data from multiple tools.

Questions/Comments?

---

Digital Analytics Jobs: Manager, Digital Attribution Science	MarketShare	Los Angeles, California
Manager, Digital Attribution Science Apply Now!(https://hire.jobvite.com/j?cj=oIQEYfwU&s=Mashable) You will be a key leader on the team that d... more Occupation: Other | Job Type: Full-time
Director of Analytics	Blue State Digital	New York, New York
The Blue State Digital Strategy Team is dedicated to identifying and implementing the most effective ways to fulfill our clients' changing digital adv... more Post Your Web Analytics Job

4 Reason Why Your Bounce Rate Might Be Wrong

Bounce Rate is generally defined as single page visits. These are the visits that leave the site without going any further than the landing page.  A very common approach to landing page analysis is to start with Bounce Rate and see if it is a problem before diving deeper into the page/site. However, the number that your tool is providing you might be wrong. Relying blindly on the number provide by your Web Analytics tool might lead you into the wrong direction. Below are 4 reasons why your Bounce Rate might be wrong

1. In Page Actions – If you have certain actions, such as video plays or windows that popup with JavaScript then the chances are that you are not tracking them as valid site interactions. In this case, even though the visitors will take one of the desired actions i.e. watch the video or click on a link to launch the popup window to fill a form etc., your Web Analytics tool will count these visits as single page visit, thus inflating your Bounce Rate.
2. External Links – If you have a lot of external links on you landing page e.g. “Like Us On Facebook”, “Buy this book on Amazon” etc. then you are purposely taking visitors out of your landing page but you might not be counting these clicks as valid site interactions thus inflating your Bounce Rate, as explained above.
3. Profile Configuration:  If you build a tracking profile of only select few pages on your site then any views of pages outside those select few are counted as “external links” (see above). For example, if you have 3 pages on your site, home.html, products.html, services.html but your profile only tracks home.html and products.html, then a click to services.html from any of these pages will be counted as an external link and hence counted as bounce if those were the only two page views that happened in the visit.
4. Spiders and Bots – For a long time Spiders and Bots were not a big issue for JavaScript based Web Analytics solutions, as very few spiders/bots executed JavaScript tags but now more and more of spiders/bots execute JavaScript thus inflating your visits counts. Many of these spiders only execute one page on the site, thus inflating your Bounce Rate as well.  Spiders and Bot seems to be an even bigger problem when major source of your traffic is Banner Advertising (Display Advertising) or Paid Search Ads.  See below for an example of a bot that might be messing up your Bounce Rate.

Now that you know that your Bounce Rate might be wrong, do your own calculations and come up with the right number before you start redesigning your Landing Page.  Recently, I came across a situation where all of the above applied. The Web Analytics tool reported that the landing page had over 90% bounce rate, after adjusting for above factors, we ended up in 50%+ range, which is still a little higher than industry average but not as bad as it initially looked (see average bounce rate).  A Bounce Rate of 50% calls for different analysis and actions than a Bounce Rate of 90%.

Read More Bounce Rate Posts

3 Tips for Expanding Tweet Reach and Engagement

Twitter feeds keep flowing with over 9000 tweets every second (Source: http://www.statisticbrain.com/twitter-statistics/). Unless your followers are constantly watching their twitter stream the chances are that your tweets will not be seen by a lot of them.  Below, I have listed three tips that will ensure that your tweets/content gets noticed and reaches most of your followers.

1. Tweet Same Content Multiple Times in a Day  – There are a lot of people like me who log into twitter few times a day (or after a gap of few days) and then and do a quick scan of timeline (or search).  They go back few hours in their twitter timeline and if your tweet did not happen to be in the timeline or searches at that time then they never see it.  Keeping this in mind, you need to tweet same content multiple times in a day to make sure your tweets are in the timelines of most of your followers when they login to twitter. In my post “Best Time to Tweet”, I suggested following timeline to tweet (this is just a suggestion and you should figure out your own timeline based on your followers and goals)
   1. Tweet at 9:00 AM PST (If all of your follower are in one time zone then tweet at 9:00 AM in your time zone).
   2. Tweet again the same message at 1:00 PST (4:00 EST) – (you might skip this if your followers are local.
   3. Tweet again the same message at 4:00 PST( If all of your follower are in one time zone then tweet at 4:00 PM in your time zone).
   In addition, I suggest adding another one later in the evening.
2. Tweet Same Content on Different Days – If you have content that is evergreen then it make sense to tweet it again. Just like above, it might take few tweets over few days/months to get your tweets noticed by your followers. Additionally, tweeting your content again after few days/months will put your content in front of your new followers and those who might have missed it previously.  However, going overboard with such strategy can potentially cause issues with some of you long time and ardent followers as they will see the same message over and over again. I use this strategy to tweet my old blog posts, which results in new retweets and followers. (I do it automatically- more on this in future).
3. Add Images to your Tweets – Late last year, Twitter started showing full images (instead of a link) in the timeline, just like Facebook does. A study (http://blog.bufferapp.com/the-power-of-twitters-new-expanded-images-and-how-to-make-the-most-of-it) showed that tweets with images got 150% more rewteets than tweets without images.  I suggest, you start including images in your tweets, when it make sense, and do your own tests to see how images affect the engagement with your tweets.

Follow me on twitter at @anilbatra