Factors Affecting Link Pricing Options

[randfish]

I was hoping to solicit advice from people on what factors you feel are important when considering renting or buying a link from a website. The top 3 are obviously:

1. Traffic
2. Focus of Visitors
3. Click-Through Rate

However, I'm hoping that for this thread, we can speak more about the link popularity issues that affect purchase price. Not everyone pays for links, but there are obviously industries where it is neccessary. I'm hoping to develop a tool that can automatically measure as much of this as possible and return a range of pricing that would be fair:

My list so far:

1. PageRank of Site (poor measurement, but probably still worthwhile)
2. PageRank of Page
3. Site Position in Top 50 Results for Primary Term (TLD)
4. Page Position in Top 50 Results for Primary Term (Page specific)
5. Number of External Links on Page
6. Site Flavor from Google (shows theme)
7. Date of Cached Snapshot of Page (shows spidering frequency)
8. Primary Topic of Page Extracted via Yahoo! API (Then conduct C-Index with target term)
9. Alexa Rank (again, poor measurement, but probably worthwhile)
10. External Links to Site (Using Yahoo! LinkDomain Search)
11. External Links to Page (Using Yahoo! Link Search)
12. Internal Links to Page vs. # of Internal Pages
13. Type of Link (customizable text, directory listing, banner/image, etc.)
14. Location of Link (content section, advertising section, navigation area, footer, etc.)

Looking forward to your ideas!

[bragadocchio]

Hi Rand,

I'm one of those who attempt to avoid renting or buying a link when possible, but I do like the idea of exploring this concept. Lots of great ideas in your post, already.

When we do what we call "media buys," one of the main factors we consider is how well the sites match up together. How well does the site where the link will appear fit into the marketing segment that we want to attract?

Another important consideration is how the link will appear. What options are there upon the page? How prominent will it be? How much control is there over the appearance and placement?

The measures that you are talking about seem to be reasonable ones for understanding what might be a reasonable price for the cost of a link. I don't know how Google would react to the use of pagerank in any calculation involving the pricing of a link. I think I've seen the aftermath of something like that before, and it wasn't pretty.

By the way, how would you go about attempting to calculate the page rank of a site? I can see a couple of ways of trying, but I'm not sure which would be the best way.

Site and page positions for primary terms. Would that be across a range of search sites? It sounds like a reasonable idea, though it might be slanted towards sites that are heavily branded, and their main Primary Term might be the brand that people recognize.

For frequency, you would probably need to capture date of cached snapshot over a period of time. It's possible that you could confuse a not so frequently but very recently cached page with one one that gets cached frequently.

The c-index comparison might tell you how well the targeted keyword phrase was chosen, and what type of semantic connectivity there is between the words that make up the phrase, but I think I like it better as a way to decide between phrases that share a keyword.

What types of things would I add that might be able to be determined programmatically? Here are a few that I thought of:

1. A possible boost for pages that might be linked to from .gov and maybe .edu pages.
2. Location of link within site - (home page versus interior page.)
3. Age of site (I'm not sure how you would measure that, but it seems like it should count in here somehow.)

Like to see some other ideas.

[Michael_Martinez]

NOTE: No comment on PageRank.

3. Site Position in Top 50 Results for Primary Term (TLD)
4. Page Position in Top 50 Results for Primary Term (Page specific)

I would probably only go for Top 20 myself.

5. Number of External Links on Page

Location of those other links, and purpose would also be important to me. For example, many of my pages have cross-promotional links to other parts of my network. Some of them are so common now I am sure people largely ignore them. So, I wouldn't be concerned about seeing my link to next to the domain's cross-promotional links. But if I were paying for linkage, I would want some exclusivity, so I would not want to see links going to competitive sites on the same page, or at least I would want my link to have a favorable position either in the body text or in the right-hand margin (which produces more clicks than left-hand).

6. Site Flavor from Google (shows theme)

I'm not sure that would matter to me.

7. Date of Cached Snapshot of Page (shows spidering frequency)

You would have to see two such snapshots, preferably three, in order to establish any sort of frequency-of-crawling footprint.

8. Primary Topic of Page Extracted via Yahoo! API (Then conduct C-Index with target term)

Waste of time.

9. Alexa Rank (again, poor measurement, but probably worthwhile)

Another waste. Alexa represents such a small user base (10-11 million users) that their rankings don't have any relevance to what is actually happening on the Web. And most of the top 10,000 sites, from what I have seen, are manipulating their rankings anyway. It's largely artificial.

10. External Links to Site (Using Yahoo! LinkDomain Search)

I would use every means possible to look at those links, including Google's reference check (since the link check is no good).

11. External Links to Page (Using Yahoo! Link Search)

Same as number 10.

12. Internal Links to Page vs. # of Internal Pages

I would refine this, if possible. I would want to know if the page was as well-linked as any other page on the site. If there were better linked (internally) pages, then why pay for a link on this page?

13. Type of Link (customizable text, directory listing, banner/image, etc.)

Agreed.

14. Location of Link (content section, advertising section, navigation area, footer, etc.)

Agreed.


But I would also look at the page itself. Simply analyzing it indirectly doesn't tell you anything about its usability and aesthetic qualities. Is it a page people would want to visit and use? I think that would be important.

[travis]

Mate,

Your an idea's man.

Maybe you could look at the depth of the page in which the link sits.

Is it in the bottom of the seventh directory or on the front page?

That may be (over the analysis of 100s of thousands of URLs ) , on average, related by chance to the length of the URL string.

Or you could measure the number of / in the url.

It may be measurable through other means, but given the speed of these calculations, quick and nasty numerical calculations will server you better.

I have cancelled your tools a number of times because they take too long and the time indicator bears no resemblance of the truth. (Need a different flash animation for each tool.) I honestly fell asleep the other night watching these little flower effects on your site, so if the seo gig does not work out, you could always use that tool in insomnia treatment.

Probably a bigger market anyway.

Having said that, when I eventually do get the results, they are usually first class and very informative, so your tools and research are at least, very exciting for designers in our situation.

So in finishing my little pontification, I would favour those data calculations which give you faster results.

[randfish]

Hey gang,

Thanks for your thoughtful input and advice. It is very much appreciated and really does figure into the development of these tools. That said, there are a couple issues I should comment on:

#1 - Speed: Right now, the tools are pretty fast, only 2.5-5 minutes per tool... however, since we instituted a queueing system, we can have 40 or 50 people running tools simultaneously, meaning you have to wait the equivalent of 2.5 minutes X 50 people...

We are purchasing a new, very fast server to handle more requests at the same time, so the queue can process up to 3 or even 4 users' tool usages at once. The timer issue is something we can't quite figure out - when we set it forward, folks complain that it's too far off one way, and backward, we get complaints the other way. Right now, we've found a happy balance of people complaining that it's inaccurate on both sides, so until we can find a better way (which I promise we're working on), it is what it is (and it's free).

#2 - Link Pricing for Link Popularity
Rather than attempt to figure out what matters to a human being, we are only trying to assume automatically those things which a search engine could assume (also has to be automatic).

The cached snapshot issue sucks, but I don't think users would want to get a result that says "thanks, now come back in 2 weeks and we'll check the cached snapshot again for a few days and let you know." Folks are going to want instant answers, even if they're somewhat inaccurate (which is why I'll probably make the responses a range rather than an exact number).

I think the Alexa system can work, because it's very hard to game at the top of the charts (top 5-10K). We can use a logarithimic calculation (which we do a lot of) to make sure that being ranked 20,000 in Alexa gives a very similiar number to being ranked 500,000 (but #7 still gives a hefty boost).

On the internal links thing, the percentage of pages on the site linking to a given page should be a relatively good measurement, you just take a site:url.com command and divide the results by a link:page.html site:url.com command and voila!

PageRank - I'll just steal from the toolbar, as we have done. It's fairly bad, but still somewhat useful, so I'll probably just make it a very small factor.

Bill - I really like your ideas about .edu and .gov as well as home page vs. internal page. Age of site is fairly calculable from registration date or through Wayback Machine - we scrape both of these for other tools.

Thanks again for all your help and keep the comments rolling. I'm really excited to have something to offer, which we can then test, tear apart and re-build.

[Michael_Martinez]

I think the Alexa system can work, because it's very hard to game at the top of the charts (top 5-10K). We can use a logarithimic calculation (which we do a lot of) to make sure that being ranked 20,000 in Alexa gives a very similiar number to being ranked 500,000 (but #7 still gives a hefty boost).

Trust me, that is precisely the range where it is being MOST gamed.

People don't care if they get into the top 100,000. They care if they get into the top 10,000.

[sanity]

Pretty much agree with all the factors you've suggested so far. One other thing I look for is trustworthyness. There's a lot of sites out there that look ok but how sure are you you'll get the link after you've turned over your hard earned cash...

[bragadocchio]

Trustworthiness is a tough one, but a good one.

Honestly, I lifted a couple of my suggestions from concepts in the latest Google patent application.

The use of links from .gov and .edu sites - which tend to be viewed as authority or trusted sites, was one that they suggested. Why the IRS or the Federal Trade Commission link to a commercial web site? My presumption, and the presumption that the people who wrote that patent likely had was that the site was an authoritative and trustworthy source of information. Though that could be a bad assumption, I'm willing to venture that the more links like that there are from .edu and .gov sites, the more trustworthy a site probably is.

I also noted their use of age of a site via domain registration information or another source as some way of gauging trust. A site that has been around for a few years under the same ownership is more likely than not to be a legitimate business.

It's possible that some other ideas might be gotten from the search engines.

Another one that came to mind, but not from what the search engines might be looking for, was checking to see if there was a valid digital certificate on the site for secure socket layer transactions. That might also be an indication of trust. Many of these things, by themselves may not have a tremendous amount of value, but adding them up may provide a bigger picture.

[sanity]

Good points Bill. I tend to check domain registrations too. How long has it been around, how long is the registration - are the details hidden.

What I also do now is email and ask some questions, if I don't get a response my questions have been answered.

[Bobby]

I am sure Enron had a few .edu and .gov links - not sure that was an indication that those guys were to be trusted

[glengara]

Sort of surprised how little topicality featured...

[randfish]

Glen - topicality will be used twice - in C-Index with primary term and with site flavor from Google Labs tool

The tool is finished now, but I'm tweaking and tweaking, as the formula is not ready for prime time. I also need to add in bonuses for .gov, .edu, .mil, etc.

Thanks for all your input. I'll post the tool here when it's finished and hopefully you can all playtest it a while.

I won't be launching it until we have our new servers online, so tool time can be cut down - the popularity of the tools lately means that we're running into report times over 30 minutes, which is no fun.

[travis]

No fun at all.

You could pop-up your latest news articles while you wait. The patent one was good. I was showing a new legal client that page today. Lawyers love patent documents.

Anyway, look forward to seeing the new tool, and the new servers !

[Peter (IMC)]

Though all those factors can be used it´s going to be way too much work to monitor them all to find the best sites to advertise on. I think that there are just a couple of factors that determine price:

Pagerank of the page the link is.
Location of the link in the page.
Anchor text + description allowed.
Number of links in the page (outgoing or internal really doesn't matter)

Those 4 are the main factors I think. All others are of little importance in most cases as they are reflected in the above 4. It also speeds up the decision process.

[randfish]

Hmmm... Peter, I'd have to disagree.

#1 - PageRank and it's flow (through outbound or inbound links) is really not a big concern. Some of the best links are on "hub" pages with many, many outbound links.

#2 - Page Topic is critical and isn't contained in any of those 4

#3 - High rankings are a great indicator that a page will pass good link value on to you.

I do really appreciate your input though. If we all agreed all the time, there'd be nothing to discuss!

[Peter (IMC)]

That´s what forums are for,.

I know that the factors I described don't describe the whole of all link factors, but they are enough to,... no I just noticed that I forgot one factor,.

Relevance of course is a must!

That one you have to add to the list. It will cover things like page title and all as well. My main point was that trying to cover every little angle may lead to too much researching and not enough decisions on which links to choose. But I guess that depends on how many links you would want to buy.

Personally I never buy links because I can't afford it and also because they only come into play when a high PR is a necessity for ranking well. In most cases PR isn't that important.

[glengara]

*..because they only come into play when a high PR is a necessity..*

I'd never buy links for PR purposes, I might well for decent anchor text from a relevant site though...

[Peter (IMC)]

*..because they only come into play when a high PR is a necessity..*  

I'd never buy links for PR purposes, I might well for decent anchor text from a relevant site though...

And how much would you pay for it? Buying links is never done for PR purposes, but if you want to determine a price, PR can be (and generally is) used.