Visit Our Blog
Anatomy of an Optimization Recommendation
With all the companies springing up (and some that have been around since ’98) to provide Conversion Rate Optimization services, I think it’s time to talk about what makes a good optimization recommendation. After all, if you’re going to spend hard-earned budget on CRO, shouldn’t you have some expectation of what you’ll be getting? And, shouldn’t you know how to avoid getting ripped off?
These are some of our thoughts on what makes for a good online marketing optimization recommendation. But before I even start, note my use of the word “recommendation,” because it’s crucial. The best way to optimize is to undertake a program of continuous improvement, and you can’t do that by completing an ‘optimization project.’ A project has a beginning, an end, and a finite duration. A recommendation is relatively small, track-able, accountable, and can be tackled one-by-one in an ongoing manner (unlike a project). Optimization isn’t a site redesign, in other words. It’s better to get 1 recommendation per week than 300 recommendations in a project document!
Without further ado…
#1 – Good CRO recommendations are data driven: This should go without saying, but without analytics data to back up assertions, I’d question the validity of any recommendation. And without a foundation of data, how would you possibly measure the success (ROI) of the implementation? For example, I might not like how you’ve designed your call to action on a particular landing page, but that’s useless. I can, however, use baseline data to demonstrate how your call to action has room for improvement, and measure the impacts of any changes you make.
#2 – Good CRO recommendations aren’t “canned”: The world is full of best practices and common web design patterns, but there are plenty of exceptions. For years, online retailers copied whatever Amazon.com did in terms of design, but didn’t see any improvement. Each case was proof that “best practices” don’t work for everyone. Good recommendations are specific to the intended site, and don’t make any dangerous assumptions.
#3 – Good CRO recommendations isolate the ‘problem’ from the ‘solution’: Make sure any recommendations you get document a specific problem (a challenge that your visitor is encountering) before they suggest a solution. For example, I might tell you to test a few versions of a landing page headline. That covers the potential solution, but doesn’t adequately educate you on the problem. If I put that recommendation into context by telling you that the current headline likely doesn’t resonate with the buying stage of the prospect who’s landing on your page, your chances of implementing that recommendation and seeing a positive impact will be much higher.
#4 – Good CRO recommendations encourage testing, but don’t require it: A good CRO recommendation should encourage or allow formal testing (i.e. split or multivariate) but shouldn’t require it. Some tests just aren’t technically feasible, realistic, or efficient. Who wants to wait 3 months to get results? Not very many of our clients, I assure you. And, different company cultures are more ready for formal testing than others.
#5 – Good CRO recommendations estimate Return on Investment (ROI): Knowing an estimated ROI is the way to make prioritization of CRO easy. You do the “low-hanging fruit” work first, then work from there. We think about ROI in terms of the balancing act between the level of effort required to implement a change and the impact that change will likely have on conversion rate. To target your highest ROI items first, you start with the highest impact items that require the least amount of effort, and work your way toward the recommendations where the potential impact only slightly exceeds the level of effort.
#6 – Good CRO recommendations are tied to goals and KPIs: It gets repetitive sometimes, but it’s still useful to re-state the goal every time we make a recommendation. It ensures that we never get distracted from what our clients pay us for. If you’re paying someone for CRO, and your goal is to get more paid membership sign-ups, and the recommendation drives more “free” memberships, is that really a good use of your budget?
#7 – Good CRO recommendations are explicit, yet flexible: Despite the temptation to “solve” our clients’ online problems, we have to let the site owner solve their own problem, in their own style. I could tell you that the only way to overcome a challenge is to create an interactive flash application, but that would immediately limit everyone’s imaginations and problem-solving skills. Instead, I tell you the challenge, give you specific options, and guide you in your implementation as needed. You can follow our recommendations to the letter, or you can interpret and tweak as needed. Who knows: maybe the solution is actually just some really good copywriting 
And we’re not afraid to learn some things from our clients along the way.
#8 – Good CRO recommendations reference examples and tools: Just like any academic pursuit, or any job search, it’s good to have references. A good CRO recommendation should provide reference materials such as screenshots of sites who have overcome similar challenges, free/paid tools to aid the effort, and rough mockups or visual aids. Any reference material that helps the implementer do a better job makes the recommendation better, and increases the chances of success.
What do you think of my list? Did I forget anything? Have any of you worked with vendors that didn’t cover all of these points in their work?
- Nobody wants to read your sh**!September 21st, 2009
- Redesign? Ask The Right Questions!September 18th, 2009
- Cross-post: How to Improve a Product Page, Part 1September 15th, 2009
For More, Subscribe To Grok!
- Capterra: The Conference Sept 13-15, Wash., DC
- "Recession Marketing: From Pre-click to Post-click" (webinar) Sept 16, 12pm EDT / 9am PDT
- "Securing the Sale: Turn Your Site's Browsers Into Buyers" (webinar) Sept 17, 2pm EDT / 11am PDT
- Excellence and Innovation conference Sept 28-30, Orlando
- The SEM Conference
Sept 23-24, Oslo, Norway - eConsultancy Peer Summit Oct 8, New York City
- The Bolo Conference Oct 12-14, Scotsdale, AZ
- Google GAAC/WOAC annual meetings
Oct 13-16, Googleplex California - eMetrics Marketing Optimization Summit Oct 19-23, Wash., DC
March 19th, 2010
4:28 pm
Well written and thought out article, Brendan. I just got all of my GA straightened out which I thought was correct to start with. I am now able to track all my PayPal order through my shopping cart process.
Plus I now have 2 advanced segments setup now. One segment is for audio listeners and one segment for non audio listeners, which to me is highly important when you sell Audio Bibles and think everyone comes to my site over others because they can listen to my audio samples online before they buy.
Well I have learned a lot of things about people who come to my website and still continue to learn every day. It is a step by step process of making your site better each day.
March 19th, 2010
11:10 pm
I agree with everything you said here especially with recommendation backed up by data. This is a recommendation you can trust because it is with evidence. I also agree with continuous improvement. there’s no perfection, only continuous improvement.
March 20th, 2010
4:27 pm
Very good article – I would agree with Andrew that the most important part is that CRO recommendations are data driven, although in my experience that is often not enough when a CEO decides he wants his Call to Action to be a purple button. I’ve often butted heads with executives who want to see “industry benchmarks” for comparison with the data (apparently execs really loved “canned” recommendations rather than fresh!)
March 20th, 2010
7:31 pm
Thank you for this interesting article. I also can agree with Andrew and Gabriele and I think it really is a step by step process of making my website better each day like the first comment author said.
March 21st, 2010
2:17 am
#1 Couldn’t be more right on, and it has to be #1. I was having coffee with an aspiring artist the other day who needed some advice on blogging. She has been blogging for years and when I asked her what her bounce rate was she didn’t know what that meant. Go figure she had never heard of analytics either haha….
March 21st, 2010
9:01 pm
1 good recommendation certainly beats 300 average recommendations. After all, execution comes after and there is just not enough resource to work on too many recommendations for optimization anyway.
March 21st, 2010
10:35 pm
I agree that a good optimization program is one that is continuous. The whole idea behind optimization is to contunually progress. Also agree that 1 good recommendation is far greater than many average recommendations.
March 22nd, 2010
4:47 am
My conversion rate is very average
Hopefully it will start increasing soon once I have made some changes!
March 22nd, 2010
6:03 am
Good article, some good things to put into practice here.
Out of interest re: #4 – I understand the point here perfectly, however one issue I have sometimes is how to do this exactly. How sure can you be that whatever recommendations you make will work? Are you not then just relying on best practice and/or and educated intuition which as #2 indicates isn’t always for the best?
March 22nd, 2010
7:11 am
According to Analytics, my CTR is only .5%, is that ok?
March 22nd, 2010
11:24 am
@Ron Michaelson: Segment that conversion rate until you see numbers that aren’t average. I guarantee you’ll learn something!
@Matt: You can’t be sure your idea will be better than the control; you can only be confident that you can run a relatively quick and scientific test with valid results.
@drupal themes: a clickthrough rate must have a beginning and an end point. .5% is ratio of people clicking through from where to where?
March 22nd, 2010
11:50 am
Hmm I think you misunderstand my question. Or I misunderstand your answer. Or maybe even the original point…
You state that while testing should be encouraged it shouldn’t be required, which I understand allows for those sites with low traffic, corporate cultures that won’t do it or other issues which prevent them testing. So – if you make recommendations how can you or they be sure that those recommendations will work when you CAN’T have a quick & scientific test with valid results?
It’s a tough point when working with clients who want CRO or something like it but don’t have the traffic to even do a simple 2 way A/B test. We can highlight areas of the site that may need work based on the stats & customer’s requirements, and make recommendations for alternatives based on a little education & intuition, but in this situation what use are they unfounded? What actions other than testing can result from the recommendations?
March 22nd, 2010
1:48 pm
@Matt: There is never a guarantee that a recommendation will “work.” There are two approaches to take when you can’t test: 1) use a repeatable optimization methodology (Ours is dubbed Persuasion Architecture). 2) Monitor visitor behavior after the change is made (analytics, surveys, KPIs going the right direction, etc.)
March 22nd, 2010
7:44 pm
I don’t see how #4 can stand as it is:
“#4 – Good CRO recommendations encourage testing, but don’t require it”
I believe if you’re truly working on conversion rate optimization you HAVE to test. I don’t think all tests have to take 3 months either…
March 22nd, 2010
8:27 pm
@Chris Guthrie: I should’ve made the distinction between what I call “formal testing” (e.g. split and multivariate) and what you might call ‘serial testing’ (e.g. make a change and compare pre and post data points). I agree that you HAVE to test to optimize, but sometimes we conduct ‘serial’ tests because the test duration estimates discourage formal testing.
March 22nd, 2010
8:53 pm
@ Drupal
0.5% is the minimum CTR for programs like adsense, so honestly it isn’t very good if you are talking about ads. If you are talking about converting customers to purchase, then it depends on traffic and profit margins.
March 22nd, 2010
9:05 pm
I think you should know about this well as conversion rates are what every online site is hoping for. If you know the mechanics of this then you are on your way to a successful venture.
March 23rd, 2010
3:18 pm
You really have to watch out for the “canned” CRO companies as there are many of them that are out there. I’ve had a number of companies call me and promise this or that. I did some research and they are a scam. They will promise you anything and give you nothing. Great article.
March 23rd, 2010
6:05 pm
I agree with this “Good CRO recommendations are data driven”
March 23rd, 2010
10:18 pm
I think rule # 7 needs to be higher–and this starts in the sales process. The minute the sales person says “We solve…” is the instant that it stops being the client’s problem and becomes your problem to solve. Not good
March 29th, 2010
10:20 am
[...] [...]
March 29th, 2010
4:51 pm
I always look deep into CRO reference examples for making sure that the listed content is true and is good to try
April 1st, 2010
10:56 pm
a good optimization program is one that will continuous.thank you for write this CRO information
April 4th, 2010
4:12 pm
“#4 – Good CRO recommendations encourage testing, but don’t require it”
Not sure if I agree, because testing imho IS necessary to find out what works best for your particular site. However, it does not have to take several months. If you got a lot of traffic on your site, steady traffic that is, maybe you can cut it down to as little as one month. In the long run I really would not care about “wasting” a couple of months on testing if the result of the testing is an conversion increase on several percent..
April 6th, 2010
8:39 am
We can highlight areas of the site that may need work based on the stats & customer’s requirements, and make recommendations for alternatives based on a little education & intuition
April 8th, 2010
3:20 am
So very true. Better to have some ongoing recommendations than alot at one time. There are many things that should be considered a journey instead of a destination
April 9th, 2010
7:37 pm
So true regarding feeding recommendations one at a time. I would rather make one change that made an impact that recommend 300 that sound smart but never get implemented.
April 14th, 2010
11:37 am
Yes it is better to slowly meter out recommendations because it gives the web owner more time to think about and implement changes as well as learn from them at the same time. It makes sense.
April 19th, 2010
2:20 pm
@drupal
Your conversion rate is going to vary depending on a lot of factors. This article shows you some great way to improve them so take them to heart and take action to improve your page.
May 3rd, 2010
5:36 pm
@White smoke
The conversion rate depends very much by the niche. The strategy must variate if you want the same results for the different niche. A good SEO / CRC expert will know exactly what strategy must use.
May 10th, 2010
11:04 am
Number 2 is key here. Good CRO recommendations are bespoke after a high degree of analysis and testing.
May 16th, 2010
3:00 pm
I did some research and they are a fake. They will promise you anything by which you become satisfied and give you nothing.
Anyways I have to change something in my own plan to get more.
Great post.
Keeping up to write more and more article.
May 23rd, 2010
5:31 pm
I will just add to your point #7
It is important that the recommendation comes with some form of education, because most often, the lack of knowledge is the main reason for the poor performance.
With keeping the owners informed, future changes to the structure or marketing should take these recommendation ahead of time as opposed to after the fact.
May 31st, 2010
3:18 pm
Very good points Brendan. I especially like #2 because most people want that magic bullet, but it always seems to be something different that gives the big lift. Great work
July 17th, 2010
6:15 am
I think testing can be overdone. Often sites do not have enough transactions to truly test lots of different elements and wrong conclusions can be drawn from small amounts of data.
For someone who has done lots of conversion optimization I would think it makes sense just to do the obvious things that feel right and are logical.
July 19th, 2010
3:18 pm
Yes it is better to slowly meter out recommendations because it gives the web owner more time to think about and implement changes as well as learn from them at the same time. It makes sense.
July 20th, 2010
3:43 am
Good CRO is always data driven.you need to have suficient data as well as patience.
I am saying this because i have worked with many companies to improve CRO but most of them failed.
July 20th, 2010
4:44 am
We can highlight areas of the site that may need work based on the stats & customer’s requirements, and make recommendations for alternatives based on a little education & intuition
July 20th, 2010
4:45 am
So very true. Better to have some ongoing recommendations than alot at one time. There are many things that should be considered a journey instead of a destination
July 20th, 2010
4:45 am
Number 2 is key here. Good CRO recommendations are bespoke after a high degree of analysis and testing.
July 29th, 2010
2:06 pm
@What is the Bible
I don’t agree with you on earning more money by more visitors, They aren’t exactly same thing.
August 7th, 2010
5:00 pm
@aldeol
could you explain us how you can gain that result
October 9th, 2010
12:10 pm
You mention ROI needing to be projected accurately, that has been very difficult for me as a marketer. I’m almost synical about it because I’ve struggled with it. But I’ve learned that clients expect realistic ROI and that you be consistent about being right. So I guess I should shop whining.
October 22nd, 2010
4:17 am
Internet marketing is just one part of an entire business approach. It does require an on going strategy and it is an important part of our new approach to communicating and capturing our client and audiences interest. I dont believe any SEO company, and I say Company will ever suggest that they have the entire solution, but a good organization will insist that they become part of your market strategy and that will require commitment and working together in partnership with your other strategies.
November 4th, 2010
10:29 am
Better to have some ongoing recommendations than alot at one time. There are many things that should be considered a journey instead of a destination
November 12th, 2010
7:56 am
I think you should know about this well as conversion rates are what every online site is hoping for. If you know the mechanics of this then you are on your way to a successful venture.
November 15th, 2010
3:38 pm
For someone who has done lots of conversion optimization I would think it makes sense just to do the obvious things that feel right and are logical.
November 17th, 2010
6:05 pm
I agree that a good optimization program is one that is continuous. The whole idea behind optimization is to contunually progress. Also agree that 1 good recommendation is far greater than many average recommendations.
.
March 21st, 2011
11:41 am
I should’ve made the distinction between what I call “formal testing” (e.g. split and multivariate) and what you might call ‘serial testing’ (e.g. make a change and compare pre and post data points)!!
April 11th, 2011
10:39 am
1 good recommendation is much more better then 1000 average… that’s my experience
August 9th, 2011
11:46 pm
Yes I must agree with what you wrote that “best practice” don’t work for everyone. It reminds me of the growing of the Facebook-like sites, plus mimicking its design, yet what they try is nothing but futile!
January 23rd, 2012
1:47 am
With all the companies springing up (and some that have been around since ’98) to provide Conversion Rate Optimization services, I think it’s time to talk about what makes a good optimization recommendation. After all, if you’re going to spend hard-earned budget on CRO, shouldn’t you have some expectation of …
January 23rd, 2012
1:39 pm
@ Federal Trade Commission
Expectation of……what? This is a great topic to bring up. Please tell us more about what you are thinking.