Shazeeye's Blog Thoughts on User Experience, Technology and Business

13Jun/140

Five steps to improve your conversion funnel

Dear Readers,

I apologize for the hiatus.  I almost forgot about my dear blog after having a baby (Sept. 2011) and starting a new job (May 2012). I'm back and want to update you on all the wonderful things that I've learned over the last 2 years.

Let's start with some best practices to improve your conversion funnel. A conversion funnel is a series of steps that users need to get through to get what they wanted on a website. For example, if users want to print coupons from your site some of the steps they would go through would be sign up (optional), select coupons, download and install software to print coupons (only for new users and not returning) and finally, print coupons.

Step 1: Define the "steps" or stages in your funnel and track it. Some funnels are complex but it is critical to define them. Some steps could run in parallel while others could be sequential. Some are optional while others are mandatory. Some steps are not required for certain segments (new vs. returning visitors example above) of your site traffic. In all these cases having a visual flow will help tremendously. Tracking these steps and how many get through each step is also very critical. For many sites it may be easy to turn Google Analytics on and get this data but in some cases such as the coupon printing one above it can be difficult to track activities that are not on the site. It was easy to track how many clicked on the print download button but difficult to track what steps took place after the click when the plugin was downloading on a user's laptop.

Step 2: The next step is to monitor the funnel to get a baseline while accounting for highs and lows due to seasonality, day of week or time of day. It is important to get a predictable baseline to make sure improvements to the funnel can be attributed to your efforts and not external causes.

Step 3: Prioritize and understand the drop offs in your funnel. Now that you have a baseline and know how many are dropping off at each point you can prioritize which one to tackle first. For example, we looked at users that added coupons to their credit card so they could use the coupon if they swiped their card at the store. We prioritized on the last step and targeted those that added the coupon to the card but did not use it. If we increased conversions at that point we would have maximum impact as we would increase our revenue if they used the coupons on the card.

Step 4: Understand why users are not converting. There are many ways to understand why users are leaving your site. The most simple and effective way is through a survey but you could also do some usability tests. To understand why users were not redeeming we sent an email to the users who added the coupons to the card but didn't use them with 3 simple questions. Why have you not used the offer? How likely are you to return to the site and use other offers (rate us)? Why did you rate us this way? Insights from our survey indicated that users had forgotten about the offer or didn't shop at that store recently. Thus to improve conversion we are considering alerts as a way to remind users to use the offer and them check conversion. In this case it was easy as we had email addresses to send a survey but if you don't you could pop a survey when a user is leaving your site to understand why. You could also follow up with some interviews to get more insight after you conduct a survey.

Step 5: Make changes and repeat steps 3 and 4. We will soon be testing if alerts will have higher conversions that no alerts (or the control group). There is lots of literature and best practices on A/B tests (see a presentation from KISSmetrics) specifically how many to test, what significance level to accept and how long to test. After you decide on making the change or rejecting the change you would go back and optimize on another step (assuming you have done everything for this step) and repeat steps 3 and 4.

30Mar/120

Indicating Interest Online Quantitatively

Attention and interest on the web are critical metrics and are an essential component that should guide any online strategy.  LinkedIn has done an excellent job in this area of indicating interest quantitatively. Let us look at a few examples:

Indicating interest in you/your profile by showing how many looked at your profile. Indicating interest in a job by showing how many people clicked on the Apply button.  Indicating interest in your connections by showing how many changed jobs in a year.

There are some other examples in the online retail industry. For example, Rue La La indicates interest in their products (clothing, accessories, home goods, etc.) by letting us know how many Ralph Lauren sweaters are left to buy thus indicating how quickly a product is getting sold. We also measure interest (though not shown quantitatively) by grouping stuff under most popular, most commented and most shared on various blogs and news sites.

The theme of the third largest social network, Pinterest (Facebook and Twitter are the top two) is centered around interest. Interest is indicated quantitatively through likes, repins and comments. We need to have a measure of interest by consolidating our online behavior (sharing, commenting, viewing, etc.). Let me know if you have any ideas on how to measure interest.

30Nov/110

Micromarketing: Location data to better serve your customers – Part 2 of 2

Location data such as using a zip code to find out how much revenue a grocery store can make is critical in your decision to decide if you want to open the store at that location. This is just one example of the powerful potential of micromarketing. Read an earlier post to get the details. Let's look at some more examples of how micromarketing can be used in defining marketing campaigns and identifying sales trends.

Identifying Marketing Campaigns based on Market Potential: Market potential is the estimated maximum sales revenue of a product during a certain time period. MapInfo Professional visually depicts the market potential of households who spend more than $150 per week on groceries for each block group (group of adjacent zip codes) in Orange County. The software also gives details on which customer segment will most likely contribute to the sales at the grocery store. For details on customer segments based on PRIZM groups read the earlier post. We see that White-Collar Suburbia have the highest market potential (count* penetration) of 21.1% and hence will be the target of a marketing campaign. This group is well described and is very specific so a direct mail ad campaign is suitable. As this group is family centric and enjoys a healthy and busy (both parents work) lifestyle we can tailor the campaigns to emphasize healthy foods and easy to make dishes that brings the family together. We can also identify the market potential by block group so say if Block X has high market potential we will place a billboard in that area to target customers. We could also use coupons to entice the White-Collar Suburbia that live outside the trade area (area where customers that visit the store reside - usually a 5 minute radius for a grocery store) of the grocery store to visit the store.

Using Point of Sale Data (data collected at cash registers) to Identify Sales Trends: AC Nielsen collects a lot of data from grocery stores and can show sales trends based on customer locations (zip codes). As seen in the image below we see market share and sales over a year for 2 brands of cranberry drink - Ocean Spray and Coca Cola.

For Ocean Spray we see that within a retailer’s trade area the retailer’s total market share for Ocean Spray’s SS Cranberry Drink is 38.6%, a decrease of 4.3 points from last year. This means that the retailer sells 38.6% of this brand SS Cranberry drinks in this trade area. When we look at the Total Sales we see that the retailer’s sales is down 14% while the remaining market increased by  2.9%.  This means its sales decreased by 14% or people could be going to another retailer with a better marketing campaign (possibly a discount) for this drink in the trade area. The total sales were $700,000+ which is significant. Thus this drink could be a cash cow (based on BCG classification) for the retailer with the right marketing campaign. Plus, the sales for Ocean Spray or the remaining market increased by 2.9% though the overall trend for sales of ocean spray was slightly down by 3.6%.

For Coca Cola within a retailer’s trade area the retailer’s total market share for Coca Cola’s SS Cranberry drink is 23.6%, a decrease of 14.9 points from last year. This means that the retailer sells 23.6% of this brand drink in this trade area. When we look at the Total Sales we see that the retailer’s sales is down 62.3% while the remaining market decreased by  23.5%.  Thus this drink is a dog for the retailer and should be dropped as its market share is less than 35% and its total sales % change is less than 5%. Plus, overall sales were $1450  which is nearly insignificant (less than 1k is insignificant).

30Nov/110

Micromarketing: Location data to better serve your customers – Part 1 of 2

Location data such as using a zip code to find out how much revenue a grocery store can make is critical in your decision to decide if you want to open the store at that location. This is just one example of the powerful potential of micromarketing. Let's go through an example of using location data to open a grocery store in Orange County. We will be using SRC's Allocate to help analyze the location data and MapInfo Professional to map the data.

Mapping propensity and density to determine revenue potential of the store: As we see in the images below, we use SRC's Allocate to determine the revenue potential of  a grocery store in Orange County (OC). We choose the retail store option as the input variable and the dollar per store as the output using the software. Data is also available for furniture stores, sports stores, etc. After the variables are input a map is produced (below) which can be interpreted as follows. For Orange County, the average grocery expenditure per house hold per month (propensity) across a block group (a group of zip codes) where darker green shades indicate  higher expenditure for groceries per household per month is approximately $5800-$16900/month for the darkest or most expensive parts. The hashed region shows total dollars spent for groceries per square mile per month in Orange County (expenditure density). The darkest hashed regions indicate people in OC spend a total of $18,000,000 to $103,000,000 per month on groceries. This data helps you determine if the revenue potential is close to what you expect and can help compute your approximate profit given all the expenses you will incur. It also helps you compare revenue potential across different locations to help you determine the ideal/optimal location for you.

Choosing a Store Location by Mapping Competitors Location Data: Using the Yellow pages we identify the zip codes of the competitors. For this example - a grocery store - let's assume it's Trader Joe's and Whole Foods. We identified 19 Trader Joe's and 2 Whole Foods store in the OC area and mapped their trade areas (area from where customers visit the store - usually a 5 minute radius for a grocery store) using the software.  The blue areas represent the Trader Joe's and the red and fluorescent green represent Whole Foods. This is mapped on the propensity and density map from above using MapInfo Professional. With this information we choose a location (in yellow) that is far from competitors and has good propensity and density. You will also check for magnet stores, customer demographics and traffic (info in next paragraph) and ensure that the information provided by these parameters will help drive your store's growth. You can also compute the break even demand (average retail demand per square mile) as seen below to inform your decision.

Identifying magnet stores, traffic, customer demographics and trade areas for the new store location: The software helps to draw the trade area for the new store location (for this example a 5 minute radius as seen in black) and can identify the magnet stores or stores that will help pull traffic (for example, drug stores).  It computes traffic - 32,800 cars/day. It also helps define the type of people in the trade area. Types of people are defined by PRIZM clusters (for details check  PRIZM Clusters) and gives you demographics and characteristics of the population you are likely going to attract. According to the report this store will attract 54% of people belonging to the PRIZM cluster defined as White-collar Suburbia. This group can be described as "upscale, college-educated baby boomers living in suburban comfort in expensive new subdivisions". For more details on this segment visit Experian's description. Now that you have such a wealth of information on your customers you can tailor your marketing message as well as grocery needs to better suit them.


31May/110

Learning Google Adwords

I started my first paid advertising campaign early this month with Google Adwords and would like to share some interesting things about it. My campaign is to promote my services in design and usability. My skills are varied and range from usability to design to marketing so I created many ads (example of one on the right) to target different users in the Bay Area. Let me walk you through the process.

1. Define the goal for your campaign and a budget. My objective was to promote my services online so that targeted people would visit my blog and email me to inquire about my services. I started with a $50 budget but you can start with $10 and see if the return on investment is greater than the costs.

2. Sign up on Google Adwords: Google helps you through this process and you can sign-up in less than 2 minutes.

3. Create a campaign: To create a campaign you need to create an ad (as shown above), identify keywords, define the regions where you know your audience is from (example: Bay Area) and define the cost per click (CPC). It is best to create more than one campaign to target different segments. For example, I make websites easy-to-use so I can target marketing people, design people and user research/usability people. I could also target industries such as healthcare, finance and retail. Let's review each step in detail.

a. Create an ad: Having decided to target an audience that wants better design I created 3-5 ads as shown on the right. Ads need to include keywords and a call to action. I created more than 20 ads and after trial and error narrowed it to the ones that I found to be most effective.

b. Identify Keywords: Keywords are the words people enter in Google search which trigger the appearance (or absence) of  your ad. With the help of the Keyword tool and the Traffic Estimator tool you can identify about 10 effective keywords for each of your campaigns. Keywords should have a high Quality Score (yellow square) and attract substantial traffic for a low estimated CPC.

4. Monitor your campaign: It is critical to monitor campaigns regularly and change them if needed. I stop campaigns that don't work and create new ones that I think will work better. Similarly, I'm tracking my most effective ads and keywords. You can also connect Google Adwords to Google Analytics to track your campaigns.  In the past few days I have got 6 clicks (right image). Eye Tracking and User Experience Design are effective keywords. Visitors spent an average 2 minutes on my blog and bounce rate was at 50%. Both bounce rate and average time on my blog through these paid campaigns show better numbers than the free traffic visiting my blog. I have yet to learn how to control CPC (as it's currently in auto mode) and will share my learnings soon.

Page 1 of 212

Switch to our mobile site