Profile

Cover photo
Richard Fergie
Works at Keyfund
Attended Christ Church, Oxford
Lives in Alnwick
777 followers|11,857 views
AboutPostsPhotos

Stream

Richard Fergie

Shared publicly  - 
 
Simulating different split test methodologies:

Suppose you are trying to maximise the number of clicks. You start off with two versions and evenly split impressions between them until each has 1000 impressions.

Next you perform some kind of maths to determine what to do next. Depending on what the maths tells you, you either continue to run the test or you pause one variation and introduce another one. Then you wait 1000 impressions and repeat...

After 10,000 impressions how many clicks will you have? 

Obviously, the answer depends a bit on what maths you do. The chart below shows the result of simulating 100 runs of the above game with six different strategies:

1: If 99% confident then pause the worst ad and add a new one
2. Same as 1. but with 95% confidence
3. 90% confidence
4. 80% confidence
5. Just pick the advert that is performing the best when you look (best observed CTR)
6. Cheat (because this is a simulation we know which version is actually better - so pick that one)

The results of the simulation show that, as long as you can keep up by producing enough new variations, waiting for statistical significance does not improve outcomes.

This result is very surprising to me. I'm sure it is at least partly because my simulation is overly simplistic; if you can precisely define what aspects of real life I'm missing here then I will have a go at coding it up.

Want more stuff like this? I'm starting a mailing list/newsletter http://www.eanalytica.com/subscribe/
6
5
Tom Smith's profile photoAaron Levy's profile photo
Add a comment...

Richard Fergie

Shared publicly  - 
 
The boat being rescued in http://www.bbc.co.uk/news/world-asia-25573096 is part of a project trying to follow the route of Douglas Mawson.

Mawson's Antarctic journey was overshadowed (in the UK at least) by Captain Scott's which happened in the same season.

On a trip to reach the magnetic South pole one of Mawson's two companions fell in a crevasse along with most of their food and many of their huskies. Mawson survived and made it back to their basecamp; much of the journey being undertaken alone.

This would have held the record for the longest unsupported polar journey until the "Footsteps of Scott" expedition in the 1980's but the use (and consumption) of dogs as well as the eating of seals compromises the claim. I think having people die also compromises this record.

Mawson's own book "The Home of the Blizzard" is good, but his reserved writing style does not do justice to his story. "Mawson's Will" by Lennard Bickel is a more readable dramatisation including a cringe inducing passage where Mawson's companion Mertz bites off his own finger to prove he can still continue (this doesn't make much more sense in context either)
1
Richard Fergie's profile photo
 
Looked up the finger biting bit in the book; he does it because he doesn't believe that his finger is that badly frostbitten and wants to prove to Mawson that it is still alive - it isn't.
Add a comment...

Richard Fergie

Shared publicly  - 
 
I'm seeing a lot of talented SEOs either moving on from SEO entirely or moving to jobs where SEO is no longer their primary focus.

Firstly, this might not be an actual pattern - there are various reasons why the sample of people I know about is not representative.

But if this is actually happening then I see two reasons for it:

1. SEO is becoming more like normal marketing and less about finding geeky hacks. I'm not saying that this makes geeks worse SEOs than they were in the past - but it does remove some of the thrill.

2. (not provided) is making it much harder to measure SEO. Although there are alternatives, anyone who tells you that SEO is as measurable now as it was a year ago is talking bollocks. This makes SEO less fun IMHO.

The (not provided) one I think is particularly interesting. I would certainly not be working in PPC if it wasn't measurable; had similar changes been introduced here then I would have some scary career changes to think about.
6
Pritesh Patel's profile photoDavid Iwanow's profile photoRavi Sodha's profile photoMontse Cano's profile photo
11 comments
 
I see a lot of people moving to start their own projects but see a lot of people with no idea on SEO & Digital marketing moving in to fill roles as there is a limited market of people they have or are willing to adapt to a changing SEO space. I know that several have either given up on clients or lost interest in the space and want to do something fresh
Add a comment...

Richard Fergie

Shared publicly  - 
 
The Moz guys have been very generous by posting the raw data used to generate their correlations for the 2013 ranking factors survey. You can get it yourself at http://moz.com/search-ranking-factors/methodology

Everyone keeps saying "correlation is not causation" which is true. What they don't say is that it is possible (to a certain extent) to infer causative relationships from data where at the moment we only look for correlations.

People have developed algorithms for this so I thought I'd run one of these over the Moz data and see what comes out.

The result is a graph http://www.eanalytica.com/files/pc-moz.png which shows which of the variables in the survey are thought to cause other variables.

I just dumped the raw data into this without any refinement and this is why some of the results seem a little odd. Here are some comments:

1. Many of the factors don't appear as nodes in the graph.
2. But "position" does (thank goodness - otherwise the whole exercise would be pointless)
3. Onsite factors rather than offsite factors dominate

I can (and given time/interest) will improve the analysis in the following ways:
1. Some variables have obvious causative relationships that I can manually include in the graph. For example the and exact match domain causes the keyword to be contained in the domain. Adding these kind of obvious edges will change the shape of the graph
2. I can also force edges not to be included in the graph. I'd do this for edges out of "position" into onsite factors. Can't really do it for offsite factors as these often correlate with "traffic" which is caused by "position"

Thoughts? (Both on the current graph and the usefulness of this type of analysis)
3
1
Will Critchlow's profile photoRichard Fergie's profile photo
2 comments
 
+Will Critchlow don't suppose you know who I can contact at Moz about getting some help understanding their data dump?

The fields are named things like "anchor_text_exact_appau" and "anchor_text_exact_appiu"; I can't figure out what these are meant to mean
Add a comment...
Have him in circles
777 people
Dan Sharp's profile photo
Yuliya Kravchenko's profile photo

Richard Fergie

Shared publicly  - 
 
It has come to our attention that a number of users, instead of clicking a bookmark to the site or typing https://www.nhseportfolios.org into the address bar of their browser, access our site by searching for terms such as "NHS ePortfolio" in search engines such as Google or Bing, and then clicking links in the search results page these sites choose to provide.

When navigating to the site in this way, the results page provided by these third party sites can often include links to alternate instances of our main site in which users login credentials do not work, but which look very similar to our main site, causing confusion/frustration to users.  (Please note, these results vary on a per user basis and are not consistent in their content).

This situation arose previously, in May/June 2013, when Google substantially changed the algorithm used to prioritise search results.

These alternate sites can be easily identified as they have different addresses (e.g. http://qa.nhseportfolios.org) which will appear in the address bar of the browser when accessing the site and because all but one of the alternate sites now contain the following information message on the homepage:
This is not the main NHS ePortfolio site - you may have reached this site by mistake.  You are viewing xxx.nhseportfolios.org, were you looking for www.nhseportfolios.org?

Whilst the contents of the results pages in the search engines are not
controlled by NES, we have contacted the two major global search engines - Google and Bing - to remove links to alternate instances of our site to reduce this possible source of confusion.  We are currently investigating the options available to us to ensure that the non-inclusion of alternative sites is made permanent.

 In the interim, please ensure that all trainees/users you are visiting the main site at the following address: https://www.nhseportfolios.org

If users continue to experience/report login issues, please first confirm that they are visiting https://www.nhseportfolios.org and, if problems persist, please provide a full-screen screengrab where possible to assign with identifying the issue.
1
Add a comment...

Richard Fergie

Measurecamp  - 
 
Here is an old post where I try to explain some of the ideas I was talking about in my presentation http://richardfergie.com/attribution/attributionexplanation.html
3
Add a comment...
 
A little update on what I've been busy with recently.

I'm doing some data work for a small charity. They are really, really into trying to make the most of technology in this sector (which is great).

Currently their data input/output system looks like this: 

Web frontend <-> Database

Which is simple in concept and is working well for us at the moment. But it is causing some problems that I worry will grow in the near future. There is a moderately complicated database and a very complicated website.

This really slows things down because anything that interacts with the database (i.e. everything I'm likely to work on) needs to be built into the web frontend.

The website is built in C# which is another barrier (along with it's complexity) to using volunteer coders - I have met many who want to help, but none who want to write C# in their spare time.

So what I think things should look like is:

                                iOS app
                                       |
Web app 1 <--- Database <---> Web app 2  
                                       |
                              Web app 3

So each app can be small and do one thing well. Adding new functionality is easier because it can be more self contained.

Do you have any experience with this kind of setup? 
Do you know anyone who has an experience of this kind of thing?

I have many further questions and would love to bash out the pros and cons of various solutions 
1
Add a comment...

Richard Fergie

Shared publicly  - 
 
An amazing opening sentence that is just slightly too long for Twitter:

"I always know it's going to be a tough day when I'm knocking back Cliff Caffeinated Shot Blocks still hours away from the bergshrund to stay awake"

http://www.alpineexposures.com/blogs/chamonix-conditions/11005353
1
Add a comment...
 
Following on from a Twitter conversation with +Craig Sullivan I wrote this blog post http://www.eanalytica.com/blog/machine-learning-and-feature-selection/

Any thoughts?
1
Add a comment...
People
Have him in circles
777 people
Dan Sharp's profile photo
Yuliya Kravchenko's profile photo
Education
  • Christ Church, Oxford
    Maths, 2004 - 2008
Basic Information
Gender
Male
Links
Contributor to
Work
Skills
Bow hunting, nunchuck, computer hacking
Employment
  • Keyfund
    Director of Impact, present
  • E-Analytica
    present
Places
Map of the places this user has livedMap of the places this user has livedMap of the places this user has lived
Currently
Alnwick
Contact Information
Work
Email