rfergie

20 ms·

151.
by rfergie 14y ago
> We actually think our ads can be as helpful as the search results in some cases. As someone who works managing PPC adverts I have been referred to as "Google meat based matching algorithm"
152.
by rfergie 14y ago
There is a difference between etags and cookies: I know how to delete cookies and I don't know how to delete etags. I think this is an important distinction and is one of the reasons I was disgusted with Kissmetrics using etags to identify
153.
by rfergie 14y ago
Stochastic Solutions have several papers about this problem: http://www.stochasticsolutions.com/papers.html In a subscription business who can be saved, who will leave regardless and then the tricky group of people who will leave if they
154.
by rfergie 14y ago
Look at http://www.getinvisiblehand.com/
155.
by rfergie 14y ago
As far as I know there are two main reasons for banning performance enhancing drugs in sport: 1. The health of the athletes. High level athletes will absolutely fuck themselves up for a shot at winning. Pro cyclists have used EPO to increas
156.
by rfergie 14y ago
All the Facebook campaigns I have run use url parameter tagging to track clicks (i.e. for Google Analytics append &utm_source=facebook...) to the url. I think this is fairly standard practice so it seems unlikely to me that they'd have
157.
by rfergie 14y ago
A bit off topic - how are you finding SnowPlow? I love the idea but I worry that it is no where near ready for the mainstream (more because of a lack of skill amongst web analysts then anything to do with the product)
158.
by rfergie 14y ago
I think the concern is that the hash function may converge.
159.
by rfergie 14y ago
All you are saying is here is that you got it wrong in the same way for all parts of the test. Which may remove the bias (I'm not sure about this) but it doesn't inspire confidence.
160.
by rfergie 14y ago
My take on bandit algorithms vs A/B testing: If you have a question where the answer can be used to inform other decisions (e.g. Which is more important for our customers, price or support level?) then an A/B test will give you a clear answ
161.
by rfergie 14y ago
Stop giving the rest of us a bad name
162.
by rfergie 14y ago
>Every URL being sent through GMAIL is being crawled by Googlebot This is demonstrably false. Email a gmail address with a URL that only you know about. Then observe your server logs - you will not see GoogleBot at this URL
163.
by rfergie 15y ago
For someone without much cash, expeimenting in a new market I would give the following advice: -Don't use broad match to begin with. -Stay off high volume terms - instead make up the numbers by adding a large number of long tail terms. -Con
164.
by rfergie 15y ago
1 pint of beer == 1 slice of bread (ish, in terms of carb content) But, a quick experiment will show you that 6 pints of beer /= 6 slices of bread
165.
by rfergie 15y ago
I don't think these are direct messages. If you follow me and I @-reply someone who you do not follow then that tweet will appear in my public timeline but it will not appear in your home view.
166.
by rfergie 15y ago
I agree. If you look at it on a CPM basis it is possible to get insane value out of social. However, the main way I've seen people do this so far is with viral content. Producing something that can go viral is very difficult for some brands
167.
by rfergie 15y ago
I completely agree. It looks like it is working and agencies/whoever is running social don't look any deeper because why would you when everything is fine?
168.
by rfergie 15y ago
I work with some brands on trying to measure this and in many cases I'm not sure it does work. The typical process works like this: 1. Brand creates facebook page 2. People who really, really love the brand connect with the page. 3. The bra
169.
by rfergie 15y ago
The point I am trying to make is that people's actions and feeling about advertising are very different. You are absolutely right that people consume advertising and react to it. I think you are wrong to conclude that this means they love a
170.
by rfergie 15y ago
You are correct about the usefulness of advertising but I think you are wrong about people's perception of advertising and its influence on them. Most people will not admit to being influenced by advertising and they'll rarely admit to clic
171.
by rfergie 15y ago
Most people do not use Adblock. Although that might be different if everyone knew about it
172.
by rfergie 15y ago
2007 or 2008 I think. Unfortunately I don't have any email from that time so I can't check or find exactly what he said
173.
by rfergie 15y ago
I once wasted some of rms's time by emailing him to ask if using Google went against his principles. He was polite enough to reply to me and say that it was ok.
174.
by rfergie 15y ago
The Dunning Kruger effect is talked about a lot on here. More recent research casts doubt on their original study. I first heard about this on [Overcoming Bias] where they talk about a 2006 study which showed "on difficult tasks, where ther
175.
by rfergie 15y ago
The robustness of Haskell is one of the things that improves productivity; less time is spent hunting bugs.
176.
by rfergie 15y ago
Couldn't that be modelled as 4 bandits: 1. Original colour/Original text 2. Original colour/"Buy Now" 3. Black colour/Original Text 4. Black colour/"Buy Now" A bit of a curse of dimensionality for anything except the simplest cases, but tha
177.
by rfergie 15y ago
Have seen exact on a 7 and broad on a 10. Discussion about probable reasons why in the comments of http://www.rimmkaufman.com/rkgblog/2010/09/28/negative-keywo... Because of some negative keyword trickery on the broad match version we dec
178.
by rfergie 15y ago
This is definitely not the case; I've seen different QS on different match types for the same keyword.
179.
by rfergie 15y ago
I also doubt that quality score takes only integral values. QS as reported in the interface is a bit like tool bar page rank; it can tell you if you're doing really well or really badly but at the end of the day your competitors probably ha
180.
by rfergie 15y ago
This is a pretty good resource for someone who just wants to know the basics. In my opinion the best high level resource for quality score information is http://www.clickequations.com/blog/ The guy who writes there also has a book coming

More ›