Payment Plans Work: The Impact of Offering a 2 Payment Plan Option on a $138 Product
December 8th, 2006 by Mike Reining Read more about Checkout Design, Conversion RateIn this test we experimented with offering a payment plan on option on a product with an average price of $138.
The idea was to see if:
1. More people purchased when they saw a payment plan being offered
2. Did this cause more purchases to gravitate towards a payment plan rather than paying in full.
The product sold for $129 + shipping for a total price of $138.
The payment plan allowed the buyer to make the purchase for 2 installments of $69.
Predictably, offering the payment plan did cause a boost in sales.
What was less expected was:
1. The impact of the boost.
2. The percentage of buyers who actually used the payment plan.
The Experiment
Everything on the page remained the same except for the buy button.
Version A : Ordinary Buy Button
Total Views: 6567
Total Purchases: 64
Purchase Rate: 0.97%
Version B: Buy Button with Optional 2-Payment Plan Option
Total Views: 6457
Total Purchases: 84
Purchase Rate: 1.29%
The Results
The impact on sales was quite high.
With the payment plan, sales went up by 32.9%.
But here’s the unexpected thing.
Many people don’t offer payment plans because money received today is of course better than money received tomorrow. The thinking is that many buyers will bail on paying the second installment or that too many single purchase buyers will opt for the payment plan instead and lower present revenue.We found this to be false.
Of the 84 purchasers who saw the payment plan option — ONLY 13 people picked the payment plan. That’s a mere 15.5%.
Yet the overall boost in sales was 32.9%.
Could it be that by simply offering a payment plan — we established a higher degree of trust and credibility and thus encouraged more sales, even from people who did not wish to use the plan.
Last 5 posts by
- Do you think SEO in other languages is easier? — September 18th, 2008
- Google Slap — How to recover from a Google Slap? — September 8th, 2008
- What Does Your Order Button Say? — August 29th, 2008
- Will Knol Rule The Search Engines? First Analysis Indicates…It Just Might — July 23rd, 2008
- Are small businesses dropping out of AdWords? — July 23rd, 2008
About the Author
Prior to MindValley Media, Mike was the Head of New Ventures Strategy at eBay where he conceived of the strategies that led to the investment in Craigslist, the launch of Kijiji.com and the acquisition of Skype.
Mike has an MBA from Stanford and previously worked for the Boston Consulting Group. He is also a certified Google AdWords Professional.
Check out other posts by Mike Reining
If you want to see what's in the private lab...
Ask yourself... which part of your business would you most like to improve?
I can send you 7 tactics related to your specific needs, if you like.
You don't have to buy anything, just take it as a backstage pass into our private course.
Will it give you the solution you've been waiting for?
There's only one way to find out.
Mike Reining
Vishen Lakhiani
I’ve been reading your blog for a while and find it intriguing, but I’ve never seen any statistical analysis for any of your posts. In other words, are the differences *significant* and at what level of *confidence*? As a minimum, you should do a chi-square test with a p value (confidence interval) of 0.05, a normally accepted threshold for significance in the social sciences. Often lower values like 0.01 or even 0.001 are used in manufacturing; I don’t know what is accepted in marketing, but I would suspect it’s similar to other social sciences.
For example, if I do a chi-square test on this example, the results look like this:
Degrees of freedom: 1
Chi-square = 3.01695682127174
Chi square at the .05 level = 3.84.
Therefore, the conclusion is that your result has no statistical significance. In other words, the probability that your variation in results are due to random sampling error is high enough that a sociologist or psychologist would not consider these differences to be significant and would say that there is no conclusion that can be made from your data.
In manufacturing, typically a higher confidence level is used which would require a chi-square value of 6.64 at 0.01 or 10.83 at 0.001.
I haven’t analysed any of your other results, but I often have the feeling that I would get a similar chi-square value.
By the way, it isn’t that far off in this case — all other numbers being equal, 87 purchases would put you over the 0.05 level of significance.
Neither here nor there. The thing is that I just find that your methodology is not sufficiently rigiorous for me to treat your conclusions seriously without running my own chi-square test on your data.
Hi Tom
I’m an engineer by training and so I understand what you’re saying. You cannot compare ecommerce site optimization to manufacturing.
The avg ecommerce site owner would not have enough traffic to study thousands of leads. So split testing is done until we identify a case where there statistical significance of one version outperforming the other. The general measure we look at is 95% statistical significance.
This means when a site owner is 95% sure that say, version A is outperforming version B, he should jump to Ver A. Waiting for 99.9% statistical significance would be an utter waste of time.
Following the 95% rule — you can optmize much faster and see better results.
For more rigorous testing — some people would use Taguchi Testing (which does come from the world of Japanese manufacturing — but in cases where I’ve tried applying taguchi to my sites I’ve not found that it was worth the effort — reults took too long to come by and overal boosts where never as high).
Of course 99.9% would be a waste of time. That’s why social scientists (like me) use 0.05 (95%) and that is what I suggested in this case. However, when I did the chi-square caclulation to see if your results were significant with a p-value of 0.05 (95% certain that the results are not due to random variation), I found that your results were *below* the 95% confidence level in this article. Haven’t tested any other articles, but I have been suspicious of your numbers for a while.
Now I wonder what test you are using. Not a standard chi-square it would seem. I know there are other means of testing, but to me I just have to take your conclusions with a grain of salt since you do not say which tests you used and what level of confidence you have.
I don’t want to come off too strong. I find your tests fascinating and think it’s fantastic that you post them. There is a lot of food for thought there and I really appreciate it. I just think that if you are going to present these sort of results, you need to do more than just say “x% increase” without info on confidence levels and, perhaps, sampling error, though that might be really hard for the type of tests you run.
Hey, I see that in your latest post you include indications of significance. Maybe I’m the only anal reader who cares, but I do think it improves the quality of your blog. At least I appreciate it. I hope someone else does!
hi Tom
The algorithm we’re using is similar to the one available on splittester.com
Which test of ours did you chi-square? I’ll double check.
I did on the data in this thread using a standard chi-squre calculator. Some test do various adjustments from the chi-square and there are of course other means of testing, but chi-square is the most standard, so I just did that one for curiousity and left at that.
You can see my numbers in the first message.
I’m not sure what algorithm the split tester is using. I used the raw numbers (clicks) rather than percentages (CTR), though it shouldn’t make any difference
sales plan example…
It looks like you have really researched this.
One of the more interesting…