Uncertainty: Google’s Weapon Against Paid Links

by geek on October 11, 2010

Google says don’t buy links or penalties may result
If you buy or sell links that pass PageRank, your site’s rank in the search results may be negatively affected. So says Google Webmaster Central. The only paid links that Google tolerates are those that are flagged as such using sanctioned methods. Sometimes a ranking penalty means a site plummets in the serps, other times it means a site is completely removed from the Google index. Indeed, there are scores of sites whose owners believe rankings have suffered as a result of transgressing this explicit rule.

Despite Google’s warning, a thriving paid link ecosystem exists
Links are highly valuable. An SEOmoz survey reveals that SEOs consider external linked keywords, external linking site popularity, and external linking site diversity (in that order) the top three most influential variables that determine a site’s rank. Given the tremendous value of external links, it’s no wonder that an entire paid link ecosystem has evolved, one which thrives on the throngs of site owners who are willing to risk Google’s wrath in order to make gains in the serps.

Indeed, every search marketing firm providing “link acquisition” in its cadre of offerings knows full well the massive scale of text link buying and selling. Some vertical-specific search marketing agencies have their own network of publisher sites, some agencies have direct access to multiple networks, some use brokers who have even greater reach, some structure deals (guarded by NDAs) between key authoritative publishers and their own clients. Some sites sell links directly to site owners in an opaque fashion, while others market their wares transparently, making no bones about their pay-for-links service; Yahoo Directory, Business.com, and Best of the Web all openly sell text links. At times it seems as though everybody’s doing it.

Some sites get penalized for buying links, others don’t. What gives?
from the exceptional cases of big brand domains (who will forever be ensconced in their cushy top ranks no matter how many purchased links point to them), it is difficult to predict whether a website will be penalized by Google for purchasing links. The apparent lack of rhyme or reason to doling out penalties is inscrutable to many site owners and SEOs who regularly see flagrant violations by their competitors going unpunished.

Why the inequality? Is Google just so incompetent that it can’t spot offending sites in a reliable way? Does it really need site owners to manually report violations in order to properly police? You’d think that a company that sets the gold standard in indexing the internet and displaying relevant results would be capable of spotting shady links. Indeed, SEO guru Rand Fishkin says he’s got technology that could ID 90% of pages containing such links.

I’ve got a theory about why Google doles out justice in an inconsistent way. My theory is that inconsistent enforcement of penalties breeds confusion and uncertainty, and uncertain SEOs are less likely to buy than those who are certain.  I have an argument in favor of that last bit of my theory, that uncertain SEOs are less likely to buy than those who are certain (I’m not going to argue for the first part of my theory, that inconsistently penalizing leads to uncertainty). I’ve stripped my argument down to some premises that rely on several assumptions about those who are deciding whether or not to purchase links. I’ll refer to such deliberators as SEOs (they include both site owners who do their own SEO and 3rd party SEOs who work on behalf of another site).


1) The first assumption is that every SEO is a rational agent trying to maximize her own utility (in the serps).

2) The second is that SEOs are in destructive competition for maximal utility. That is, each SEO’s utility will increase when her competitor’s utility decreases and decrease when her competitor’s increases. For simplicity, I’m going to assume zero-sum competitions: one SEO’s utility change results in an equal change in the opposite direction for her competitor.

3) Third, I assume that all SEOs believe that all SEOs have access to affordable links (links that would net a greater utility given the cost were some other assumption about whether Google punishes true).

4) Finally, I assume that SEOs believe that, all things being equal, links increase a site’s rank.

That “all things being equal” bit in number 4 is important, as I’d like to consider the decision about whether to buy links, regardless of how other factors such as number of backlinks, anchor text linked, backlink profile, link cost, and quality of site linking, contribute to serp ranking. How one evaluates these additional factors in her deliberation over purchasing links is, of course, crucial to her final decision. However, in order to isolate the effect of only purchasing links, these factors will be lumped into an “all things being equal” bucket. Link cost is perhaps the most contentious variable in the bucket, however in the real world, a company can garner links for as little as one dollar or as much as one HD TV with satellite service. For our purposes, let’s just assume that, for any site, all SEOs believe that there are affordable links that could benefit that site’s SEO.

SEOs as Game Players
I’m going to focus on the strategies of SEO1 and SEO2, both of whom are SEOs at each other’s throats for superior rank in the serps for the same term. In game theoretical analysis, a payoff matrix, such as PM1, PM2, and PM3 below, is often used to show the utility that each player – in this case an SEO1 and her SEO2 competitor – realizes on account of acting (or failing to act) a certain way. The number to the left of the comma is the utility realized by the player in the row on account of acting (or failing to act), and the number to the right represents the utility realized by the player in the column on account of her (in)action.

ASSUMPTION 1: All SEOs believes that Google will not punish suspected link buying by any SEO
On the assumption that Google won’t penalize an SEO’s site for suspected purchased links to that site, I contend that SEOs will engage in link buying to improve their site’s position in the serps. This, to me, doesn’t seem to be a contentious claim. If SEOs were to believe all they do now about the importance of backlinks (ie nothing is more important than backlinks in determining rank) and believe that Google didn’t penalize for buying links, then they would think that purchasing a certain kind of link would have a beneficial effect in the serps on the site for which the purchase was made.

Suppose SEO1 and SEO2 both believe that AcmeLinks.com sells links for -1 and that any site to which AcmeLinks.com points a paid link will benefit by 2 points of utility. Where the cost of a link = -1 and the expected utility from the link = 2:


Payoff Matrix in case Google doesn't punish

  • Suppose SEO1 buys a link.
    • If SEO2 doesn’t also buy a link, then SEO1 will be out -1 for the cost of the link but up 2 for a net gain of 1, and SEO2 will go down -1 on account of SEO1’s gain. Bottom left quadrant.
    • If SEO2 does buy, then her purchase offsets the -1 loss from SEO1’s buy and both remain at 0. Top left quadrant.
  • Suppose SEO1 doesn’t buy a link.
    • If SEO2 does buy a link, then SEO2 sees a net gain of 1, while SEO1 loses the same. This is the reverse of the bottom left quadrant. Top right quadrant.
    • If neither buys, then the utility values remain unchanged (0 for each). Bottom right quadrant.<.li>

For both SEO1 and SEO2, buying a link is the strategy that will yield the least risk (0) and greatest gain (1), while not buying is the inferior strategy as it yields a highest risk (-1) and least gain (0). Hence, on the assumption that SEOs believe that Google will not punish SEOs for suspected link buying, SEOs will buy links.

Two caveats regarding the epistemic states of the SEOs:

  1. So long as each SEO believes Assumption 1, and believes that her competitor is like herself (ie believes that her competitor believes that Assumption 1 is true), then it doesn’t matter which value is assigned to expected utility earned from the AcmeLinks.com’s link, so long as the believed net gain from a purchased link is greater than one.
  2. It need not be the case that each SEO believes that the expected utility gains (for each SEO) are the same as her competitor believes they are. For example, SEO1 could believe that SEO1 will gain 20 and that SEO2 will gain 3. And SEO2 could believe that SEO1 will gain 2 and SEO2 will gain 2.

ASSUMPTION 2: All SEOs believe that Google will punish suspected link buying by any SEO
Note that this assumption does not reflect how most SEOs think about link buying today. Most SEOs today aren’t sure if purchasing a link will do them ill or well. They know of the numerous cases in which Google does not punish link buying; perhaps some of those cases involve their competition. However, Assumption 2 is that SEOs locked in a bitter battle over position are not uncertain; they do believe that Google will punish for suspected purchased links. Clearly, if an SEO believed that buying her site a link will diminish her site’s position, no SEO would buy her own site a link.

But, I believe that so-called negative SEO, a term that describes trying to harm a competitor’s site’s rank rather than help one’s own, would be the preferred strategy for a rational SEO in destructive competition. That’s because in a destructive competition, a competitor can advance in one of two ways: by directly increasing her own utility or by indirectly gaining from her competitor’s loss of utility. Indeed, we can easily envision (and many of us are probably involved in) scenarios in which an SEO would benefit from having a competitor SEO’s site drop in ranking. It doesn’t matter if an SEO’s site holds a superior rank; if she’s in destructive competition, then she stands to gain from her competitor’s loss no matter where either’s site shows up in the serps. This sort of situation arises when, for example, two websites are selling the same limited-quantity item, such that even if one site owns the first place rank, it would still increase its sales by wiping the only other etailer, the second site, out of the serps.

As negative SEO is a case of sabotage, let’s suppose that each SEO believes that AcmeLinks.com sells links that are suspected purchased links and that the ill effect of these links upon the sabotaged SEO is -2, causing the saboteur’s utility to rise by 2. The utility values in the payoff matrix for SEOs competing under Assumption 2 (where sabotage is defined as buying a link for a competitor’s site and then making Google aware of that purchase, such that Google suspects the competitor of buying the link for her own site) are identical to those in Assumption 1.


Payoff Matrix in case Google punishes

The reason that the payoff matrix is the same as it is in PM1 is that each SEO will gain exactly as much utility as is equal to the loss from her competitor for being punished in the serps. SEO1 reasons similarly to how she would under Assumption 1: if I buy a link to sabotage SEO2, my utility could go up by 1 (in case SEO2 doesn’t also sabotage me) or it could remain unchanged (in case SEO2 also sabotages me). On the other hand if I don’t try to sabotage, either my utility is unchanged or it goes down by -1 (in the case that SEO2 sabotages me). Since I think it’s just as likely that SEO2 will sabotage as that she won’t, I’ll buy a link to sabotage her. SEO2 reasons the same way, and so both SEOs buy links under Assumption 2. Thus, SEOs in destructive competition who believe Google punishes will buy links in order to sabotage their competitors.

The caveats regarding the epistemic state of the SEOs hold for Assumption 2 as well.

ASSUMPTION 3: All SEOs are perfectly undecided about whether and by how much Google will punish suspected link buying by any SEO.
I’ve endeavored to show that if an SEO believes that Google will not punish suspected link buying, then that SEO will buy her own site a link. Indeed, if Google were to declare link buying a white hat tactic, the floodgates would open and TLA might just go public. On the other hand, I’ve also argued that if an SEO believes that Google will punish suspected link buyers, then that SEO will buy her competitor’s site a link and rat her out. Hence, if an SEO believes that Google will punish or if an SEO believes that Google will not punish suspected link buying, then that SEO will buy links.

Now let’s analyze link buying under a final assumption: suppose each competitive SEO is perfectly undecided regarding link buying; she neither believes that Google will punish nor that Google will not punish a site for a suspected purchased link. Rather, she thinks the probability that Google punishes or not are equal; fifty-fifty. Her dilemma is similar in kind (though perhaps not in probability assignments) to that faced by many SEOs today. What’s more, given what she knows about the benefits of external links and the risk from Google discovering that an SEO has purchased links, she thinks that the potential utility gain an SEO realizes from a non-punished bought link is about the same as the potential utility loss that SEO realizes from her site being punished for a suspected bought link.

Again, consider AcmeLinks.com. Both SEO1 and SEO2 believe that the links sold by AcmeLinks.com will cause any linked-to site a utility gain of 2 just in case Google doesn’t punish and a utility loss of 2 just in case Google does punish. As I see it, there are four possible moves each perfectly undecided SEO can make: buy a link for one’s own site, buy a link for one’s competitor’s site, buy a link for both (ie hedge one’s bets), or buy a link for neither. The payoff matrix is more complex for this sort of game. That’s because for each move, there is a 0.5 probability that Google punishes and a 0.5 probability that Google doesn’t punish. Too, there is a 0.25 probability that the competitor moves any one of the four ways under each punish/non punish outcome. Hence, for any one SEO, each of her moves could result in eight possible outcomes, where there is a probability of 0.125 (0.5*0.25) that each outcome occurs. By multiplying each outcome’s probability of occurrence by its associated utility and summing all eight products, the weighted total expected utility for each move can be calculated.

The top matrix in PM3 shows expected utility if Google doesn’t punish, the second shows expected utility if Google does punish. The third shows the sum of the products of each outcome’s utility multiplied by the probability that it will occur. Hence, the bottom matrix represents the final weighted expected payoffs.


Payoff Matrix for a perfectly undecided SEO

It turns out that a perfectly undecided SEO destructively competing against another believed perfectly undecided SEO will refrain from purchasing any links, because that strategy is the one where she stands to realize the greatest possible gain and least possible loss (0.5 and 0).

I’ve argued that if destructive competitor SEOs believe that Google punishes or believe that Google doesn’t punish, these SEOs buy links. Additionally, I’ve argued that if these SEOs are, instead, perfectly uncertain regarding Google’s propensity to punish, they will not buy links.

Hence, if Google wants SEOs not to buy links, Google would do best to keep SEOs uncertain regarding whether they will be punished for buying links. Further, Google would also want to make the uncertainty ubiquitous, such that each SEO thinks other SEOs are also uncertain. And, such uncertainty can be accomplished quite easily: by sometimes punishing, sometimes not punishing, in a seemingly arbitrary way. Of course, as described in this post’s opening, this is precisely the state of affairs at present. SEOs are confused and uncertain – seeing evidence both that Google punishes as well as evidence that Google doesn’t punish. This, I contend, is exactly as Google wants things – if stopping link buying is Google’s end goal, keeping SEOs in the dark is its proximate goal.

I anticipate several objections to my argument, the most significant of which are these:

1) First, in order for the game theoretical analyses under the assumptions above to work, the competitors would have to have complete knowledge about how purchasing each link will benefit or harm each SEO. But it’s not realistic to think that SEOs will ever have complete knowledge about how links will benefit themselves or their competition. SEO isn’t a chess game where each move has a requisite and known utility gain and loss for each player.

Although typical analyses like the ones above involve players who know and share beliefs regarding the utility values of any possible strategy of both themselves and their competitor, so long each competitor believes that whatever good comes from a purchased link outweighs the cost to the SEO who is doing the purchasing (and believes that her competitor believes the same), the net payoffs will always be higher for purchasing a link under Assumptions 1 and 2.

Additionally, so long as each SEO believes that the expected gain from a non-punished backlink for each site equals the negative of the expected loss from a punished backlink for the same site, we get the same dominant strategies under Assumption 3 no matter which utility values are assigned. Hence, not buying is always the dominant strategy under Assumption 3. But this leads me to the second objection.

2)  The argument assumes (Assumption 3) that a hypothetical SEO believes that, in the case of Google punishing, each SEO stands to lose exactly the negative of what she stands to gain were it the case that Google doesn’t punish. That is, SEO1 must believe that if Google punishes, her site stands to lose 2 from a purchased link pointing to it, and that her site stands to gain 2 in the case that Google doesn’t punish. What’s more she must also believe that SEO2’s potential loss is exactly the negative of SEO2’s potential gain.  If an SEO believes that her own site stands to gain more than it stands to lose or vice versa, or believes that her competitor’s site stands to gain more than it stands to lose or vice versa, the better choice might still be buying even for a an SEO who believes the probability that Google punishes is equal to the probability that it fails to punish. What’s more, in some such cases, it’s not clear which strategy is the best one, because choosing the best strategy depends on how risk averse an SEO is. But, is it so strange that an undecided SEO would believe that the potential risk is the additive inverse of the potential reward?

Suppose that an undecided SEO believed that all SEOs only risked a net -0.0000001 in buying a link that was ultimately punished and a net 1,000,000 in purchasing a link that was ultimately unpunished. Then, it’s not clear that she would even care whether Google punished or not. Unless she were inordinately risk averse, she’d buy every time; and in that case, I think she is very different from the undecided SEOs at work today.

Of course, the weighing and weighting of reasons and probabilities is much more complex than the simple scenarios I’ve laid out. In the real world of indecision, probabilities and expected utilities vary wildly. However, so long as a perfectly undecided SEO is less likely to buy than one who believes one way or the other, I’m happy with the argument. And, if we believe the conclusions in Assumptions 1 and 2, then so long as there are some perfectly undecided SEOs of the type I describe in Assumption 3, I think that the argument is in no real danger from this objection.

3)  Finally, it’s not clear that the hypothetical SEOs represent all that many SEOs in the real world. For starters, not all SEOs are or believe they are (or have second order beliefs about their competitors or themselves that either is) locked in zero-sum conflicts with their competitors. Some may gain a little from a competitor’s loss, some may gain nothing, and others may gain disproportionately more from a competitor’s loss. Others may believe differently about utilities relative to outcomes on their own site versus their competitor’s site. Additionally, even if we assume that there are many non-zero-sum destructive competitor SEOs, the references to “first-pagers” alienate the vast majority of SEOs whose sites are in the serps. Does my argument still apply to those SEOs whose sites lie in the netherworld?

Although it’s true that there is plenty of non-destructive competition in the serps, it’s just a matter of fact that there are only ten listings per page and that a newcomer’s ascension to page one entails the displacement of an erstwhile first page listing. And, while Google used to not let a very weighty domain take over the upper ranks, now that more than two results from the same domain can appear in the serps for some query, ranking for competitive terms is more cut throat than ever. The point is that even if an SEO doesn’t want to be in destructive competition, if she’s to achieve prominent placement for a competitive term, she has no choice but to have her gain be another’s loss; I think that most SEOs who rank well for certain terms know this. Furthermore, so long as each SEO believes each is negatively affected by the other’s gain by some amount that is the same under each assumption, competitors need not believe that they are in zero-sum competition (although all other assumptions in the argument must remain).

But what of this narrow focus on “prominent placement”, and even page one listings? Surely SEOs whose sites are in the top twenty to thirty positions don’t make up the majority of SEOs. Are those SEOs whose sites are listed in the lower pages of the serps also in destructive competition? And if not, then doesn’t my argument only hold for a very small minority of SEOs? My answer to these is that although it is true that SEOs working on top-ranking sites for competitive terms may be in the minority, their sites hold a disproportionately high value (relative to number of clicks) on account of their position. Indeed, unless a site is listed in the first three pages of search results, chances are, no one’s going to click on its listing. And that means that while the majority of sites competing for a term may not be among the first several pages in the serps, the overwhelming majority of clicks are on listings in the first several pages.  I submit that if Google is going to focus its attention anywhere it’s on those sites that capture over ninety percent of the clicks it refers.  Hence, it need not be the case that my hypothetical SEOs are representative of most SEOs; that they are representative of those SEOs that Google is most interested in influencing is the important bit.

Given the relative importance of external links in achieving good rank, it’s no surprise that an economy has developed around them. Plus, given Google’s genuine desire to create a search experience that is superior to its competitor’s, it’s no wonder that Google wants to prevent sites from diminishing serp listing relevance by buying their way to the top. Yet, I’ve endeavored to show that if Google were to have a consistent record either of punishing or of not punishing suspected link buying, SEOs would game the system all day long, buying links to benefit their sites. On the other hand, if Google can keep SEOs uncertain of the outcome of a purchased link, Google will actually prevent more link purchasing than would otherwise occur. For, an SEO who is uncertain about the outcome of his purchase is just the kind of SEO that is less likely to purchase.


Leave a Comment

Previous post:

Next post: