serial testing

Here’s a common question: “How do you increase conversions when you only get a small amount of traffic?”
The first answer is, go get more traffic. [pullquote]The closer your conversions are to zero, the closer your conversion optimization efforts will be to guessing.[/pullquote] You can do statistical optimization using split testing if you have enough conversions, but this usually comes with more traffic.
The second answer is to get more conversions so you can do conversion optimization to get more conversions. Which came first, the conversion or the optimizer?
This last point is, of course, the proverbial “rub.”
Here’s how to get started if you are running low-traffic websites.

Get Accurate Data

Be sure your analytics is setup properly. I offer an analytics setup checklist to help with Google Analytics. You’ll want to avoid blind spots such as overlay windows, tabbed content, and subdomains on separate analytics accounts.
You’re going to need a good source of data when you start picking things to test.
Compare your analytics data to a secondary dataset. Compare lead conversions to your CRM. Compare transactions reported to your accounting system. Your analytics should be within 15% of reality. Don’t be afraid to install a secondary analytics package to verify your main analytics setup.
[sitepromo]

Get Some Qualitative Data

Low-traffic websites need to get more qualitative data. Right now, the one-stop-shop for qualitative data is HotJar. It offers click-tracking, session recording and feedback surveys. For alternatives, check out the ConversionDashboard.com.

Low-traffic Websites Use Serial Tests

If you don’t have the conversions to do split testing, you’ll want to do serial testing. This simply means making a single small change to your site and letting it run for at least two weeks. Since you have solid analytics (see above), you can see if there is an improvement in performance.

Measure More Than Conversions

There are some predictive metrics that you can use to guage the performance of your serial tests.

  1. Bounce rate on landing pages
  2. Click-through-rate on key internal pages
  3. Add-to-cart for ecommerce sites
  4. Form completion percent
  5. Abandonment rate (basically the opposite of the last two)

Time on page, time on site, and pages per visit are to be taken with a grain of salt. Increasing these may correlate with lower conversion rates.

Start with the Message

[pullquote]Nothing works until your value proposition is strong. I recommend testing changes to your value proposition.[/pullquote]
I’ve done hundreds of free strategy consultations over the years. Most of the time, I ask the consultee to tell me about their business. Typically, I get a concise, clear statement of the offering and value.
Rarely does this clarity appear on the website.
Sit with a copywriter and tell your story. Then, don’t edit them. Whatever they come up with, try it.
You should also test:

  1. Headline
  2. Call-to-action button text
  3. Pictures. If you can’t write a meaningful caption for an image, change it.
  4. Add sub-headlines
  5. Add bulleted lists

Don’t bury the lead. A great headline — called the “lead” — is the core of a strong value proposition. Often the headline that would best “grab” a reader is buried somewhere in the copy.
Find the headline that gets visitors to read your value proposition, and you’ll have the cornerstone of conversion in place.

Look for Big Wins

You’re going to have to find what we call “big wins.” This means that your change increased conversions by more than 50%. Rich Page wrote on low-traffic testing. My comment on his post was as follows:

You can also split testing with less than 100 conversions. You just need really big wins. If you have a treatment with 20 conversions and another with 40 conversions, a 100% difference is something you can probably bank on, even with such small numbers. However, if one treatment got 20 conversions and the other got 30, that 50% increase is too close to the margin of error and shouldn’t be considered an improvement (even though it feels like a win).

Technically, it’s OK to make a treatment with, say, a 30% increase the new control. Just know that you’re not likely to continue to see such an increase with small transaction amounts.

Ditch Your Outliers

You’re going to have to eliminate “outliers” in your data. Outliers include extreme orders in ecommerce sites and rushes of leads from activities such as email blasts and bursts of word of mouth.
For an ecommerce site, you should look at orders that are one or two standard deviations away from the mean.
So, what does that “mean?”
Here is two weeks of daily sales data for a site that gets about one sale per day.

There are two obvious outliers: One day with no sales in the first week, and one with $160 in sales the second week. Statistically, a 16% increase is irrelevant, but the point is driven home when you calculate the standard deviation range.
For this data, an outlier will be lower than $27.90 or higher than $86.89.

When we remove outliers we see a drop in sales of six percent. This is statistically uninteresting as well, but illustrates how outliers can affect results.
If you’d like to see how I calculated the min and max, download Example of Outliers-Conversion Scientist-low-traffic post.

Don’t Let it Run

Split testing can be done on low-transaction sites. However, don’t let the test run for more than, say, six weeks. The results just aren’t reliable. There are too many other variables mucking with your data over such long timeframes.

Always Be Testing

Just because you have few transactions per month doesn’t mean you can’t be learning. In fact, not learning may well be the reason you have few transactions per month. Never stop trying things, and use good data to decide what you keep and what you throw away.
[signature]
Feature image by Shaun Garrity via Compfight cc and adapted for this post.

How important are images to your landing page? The formula we use in our Chemistry of a Successful Landing Page includes the element “Image” as a necessary component. At the heart of this is the need for the visitor to imagine owning the product or service. That’s right, even services.
For some, it’s difficult to “show the product.” If you’re offering an expensive software solution or consulting service, how do you communicate what it will be like to own that? Screen shots, flow charts and explainer videos are typical go-to solutions.
Lazy designers drop happy, smiling people on the page. Avoid this business porn.
At the other end of the spectrum is the visual product or service. Photographers, artists, decorators and designers have a portfolio of past work to help visitors imagine buying from them.
Vacation Beach Portraits is such a visual business, and they have some test results that offer some insights. I love it when small businesses take up testing.
Vacation Beach Portraits takes family portraits of tourists to the Orange Beach and Gulf Shores areas of Alabama. The beautiful white beaches and sunsets over the Gulf of Mexico offer an ideal setting.
The folks at Vacation Beach Portraits tried testing a landing page against their home page, a blog filled with samples of their work.

Vacation Beach Portraits HomePage-Selections

The Vacation Beach Portraits home page was full of delicious images showing off the work.


familyportraits_vacationbeachportraits_com

Then landing page features a prominent call to action and portfolio video.


Vacation Beach Portraits HomePage thumbThe home page was a long scrolling collection of pictures from recent shoots. Load time can significantly decrease conversion rate on pages like this. However, though lazy-loading of the images allowed me to start viewing images immediately.
The landing page, built using Unbounce, provided an explainer video with samples from their portfolio. It is shorter and features a bulleted list of benefits as part of the copy.

Serial Test

This local business will have few transactions each month. Therefore, Jason Odom of Vacation Beach Portraits did tests in series.
From May 1-15, he sent his search traffic to the landing page.
From May 16-31 he sent his search traffic to the home page.

image

Comparison of visits to inquiries shows a 42.1% increase in conversion rate for the home page. However, this is not statistically valid. Source: ABTestGuide.com


Given the relatively low number of clicks and inquiries, the two pages converted at the same rate statistically. When testing low-traffic sites, we are looking for treatments that beat the control by large margins — 50% or 100%.
In this test, the home page generated 42% more inquiries and 105% more paying clients. Neither of these results was statistically significant, though. The sample sizes were just too low.

Why Didn’t the Landing Page Outperform the Home Page?

Anytime we hear that people are sending “store-bought” traffic to their home page, we roll our eyes. We are almost always able to improve conversions by sending visitors to a landing page.
In this case that didn’t happen. What’s the deal?
Two hypotheses emerged from this test.
1. The long page full of gorgeous pictures found on the home page is what visitors want.
2. The clear call to action found on the landing page kept it in the running.
For their next test, we recommended either adding a bunch of these big gorgeous pictures to the landing page, or adding a call to action button at intervals down the home page.
The quality of the images in the landing page video was lower than the full-width photos found on the home page.
When someone decides they want an amazing family photo like those shown, a button with “Schedule Your Photo Session” is exactly what they will be looking for.

Other Considerations

There were some additional hypotheses we felt would improve the performance of these pages.

This font is pretty, but very hard to read.


We felt that the script font used on the home page was hard to read, recommending a serif print font instead.

Beach Clothing Color Ideas is at the bottom.


The navigation on the site was not particularly logical. The very helpful navigation item “what to wear” seems to link to anything but topics on what to wear. Every link on a site should keep its promise.
Making the phone number more apparent my close the time it takes to book a client from the web or landing page. We find that adding the phone number to the headline (yes, the headline) will significantly increase calls without depressing form fills.

Advice for Businesses with Visual Offering

If you have a visual product, you should leverage this with high-quality, high-resolution web images. Don’t be afraid of long pages. Visual visitors know how to scroll and will appreciate the wealth of stimulation.
However, don’t forget the calls to action.
You never know when someone has seen enough to buy. Lace a buttons or links among your images. Keep in mind that the buttons or links are going to have to compete visually with the images, so make them pop.
The button or link will go to a more traditional landing page or product page that handles objections, allows selection of size, color or format, and asks them to buy.
In almost every case, use captions. These are the most read copy on most pages and are a great place to include a call to action. Tell them what they are looking at, even if it is obvious to you.

Results From the Follow-up Test

This is the busy season for Vacation Beach Rentals, and their landing pages are already converting very well for them. We won’t know the results another test for some time. Subscribe to the Conversion Scientist by email to find out the rest of this story.
[sitepromo]
[signature]

Did you know that all of the changes you’re making to your Web site are largely wasted? Really.

If you were to put just a little discipline behind your site changes, you could learn scads about what your visitors like and what makes the cash register ring online.

I address the power of Serial Testing in my Conversion Sciences column on Search Engine Land.

 

Gather close, dear friends and hear my words of warning. Gird yourself, for my tale would make any enlightened man or woman stagger back in disbelief. I do not share such news with you lightly. It is not my errand to harm or discomfort your mental or psychological well-being.

 

Resist the urge to sever the very Internet connection with which you receive my message. I shall soon relieve your stress with a simple technique that will counter the malevolence that I describe.

 

I reveal my knowledge only because you may in fact be a person possessing an enterprising mind. If this is true, I know you will see in my tale of woe and tragedy the shadow of immense opportunity.

Read my Search Engine Land Conversion Science column…