Threats to Google’s Model?

Google’s business model has become increasingly diversified over the years with significant revenue streams coming from its Ad Network, its Android OS, its publishing platforms (YouTube), etc. But let’s not kid ourselves: AdWords remains the cash cow in terms of revenue and most certainly in terms of profits.

Could collusion threaten that cash cow?

Suppose for a minute that all participants in a particular auction space were to agree to lower their bids in unison. This could happen by mutual agreement of advertisers, or conceivably through intermediaries like an agency that works with all the key players in a particular vertical, or a software platform that might have visibility into a big chunk of the bid landscape.

An agreement honored by all parties would have tremendous positive impact on the advertisers’ bottom lines, and potentially a fairly devastating impact on Google’s bottom line if that practice spread far enough.

It will never happen

Prisoner’s Dilemma and the Tragedy of the Commons tell us that while cooperation with others produces the greatest good for the greatest number, self-interest drives individual decision makers to not cooperate, even though the result of not cooperating is the worst outcome for the participants collectively.

A classic formulation of prisoner’s dilemma has two people, Fred and Sue, arrested on suspicion of a serious crime, let’s say grand theft. If they each keep their mouth shut (cooperate with each other, not the police) they’ll get 1 year in prison each for breaking and entering. If they both confess they’ll get 8 years each. If one rats on the other but the other doesn’t confess or implicate the first, then the confessor goes free and the one who stuck to the agreement gets 10 years.

The best outcome for the prisoners collectively is to not rat each other out (two years total). The worst outcome collectively is to rat each other out (16 years collectively), yet that’s what will happen according to the pursuit of rational self-interest.

From Fred’s perspective, Sue will either rat him out or not. He has no way to know which she will do, though they agreed to maintain their mutual innocence. If she rats him out, he’s better off having confessed that they both did it because he’ll get 8 years instead of 10. If she doesn’t rat him out, he’s still better off confessing/ratting her out because he’ll get no time instead of a year. Easy choice for him, and Sue has exactly the same choice and will reach the same conclusion.

If this logic applies to two prisoners who presumably had some sort of friendship between them, imagine how well it applies to competitors in the Google auction. How much would you sacrifice to help out your fiercest business rivals?

Trust is not hard, it's foolish

Someone might object that this is a plan for mutual benefit: everyone wins, so why not play along? Simple: everyone doesn’t win equally, and you’d win much more by being the only cheater in the group!

For want of a better construct, let’s talk about ad positions though the concept is all but meaningless these days. The advertiser at the top of the page gets 10 times the traffic of the advertiser in position 6 (making up numbers that are often in the right ballpark depending on context). If everyone agrees to cut their bids by 75%, the advertiser in position 6 enjoys a much better ROI than they had, but still gets the same volume of traffic. However, by NOT dropping their bid when other participants do, or by dropping it less than 75% they could get the same favorable ROI they’ve been getting, but be positioned much much higher on the page getting vastly much more traffic.

A huge potential win, but only if the position 6 bidder “cheats” and others don’t. The position 3,4,5,7,8,9… advertisers are all thinking the exact same thing. On the flip side, the position 6 bidder could lose big if s/he is the only one who does cooperate with the scheme. If everyone else maintains their bid and position 6 cuts theirs by 75%, they’ll disappear from the auction.

When the benefits of cheating are so large, and the consequences of being the only rube who cooperates are significant, the odds of competitors coordinating/agreeing to take Google to the cleaners is zero by my calculation.

Not only would it require collusion of all participants in the market — some who are currently priced out would be brought back in the game — which is almost inconceivable, it would also require that they stick to the agreement, which to me is thoroughly inconceivable.

The auction model gives advertisers the power to set the price charged for traffic which on the surface seems threatening to Google. However, as long as the value of that traffic remains high, advertisers will gladly pay for it. By handing over control of the pricing, the auction model turns out to be great protection for Google’s revenue stream.

5 Responses to “Threats to Google’s Model?”
  1. Hugo says:

    Always going to dig any marketing article that references game theory! I’d argue that the concept of ESS (evolution stable strategy) is also at play here.

  2. Hugo, thanks for your comment, and excellent point re: Evolutionarily Stable Strategy. I agree, this is a better metaphor as the decision in this case is not one-off, but iterative. You have to agree to cooperate over and over (auction by auction) hence defection propensities and reactions to those events will dominate. Same conclusion, of course: it’s hard to see advertiser collaboration hurting Google.

  3. Brian Bien says:

    This is why we will not see Google disclosing or encouraging the display of advertiser’s bids. What if there were a way to systematically and accurately track what others are bidding? How would this change the game?

  4. Brian, thanks for your comment. I’m not sure it would change the game, even if done reasonably well. Clarity would show which advertisers were breaking the bargain, but why wouldn’t the advertiser who’s in position 10 (not to mention those off the page) crank up their bid to get more traffic if they could do so profitably? The incentives for the “losers” to decide that the grand bargain is in the way of riches will be too great, I think…(I’ll just cheat on cyber Monday…what are my co-conspirators going to do to punish me?)

    Moreover, the only accurate way to talk about what other advertisers are bidding is one auction at a time. Vagaries of match types, regional targeting, search retargeting, dayparting, etc mean that there is no single answer to the question “what are my competitors bidding on “foo bar”? You’ll have local competition in some areas that mean who your competitors are will vary; you’ll have matchtype games that mean different ads from the same competitor might be served depending on QS and bid of each ad in their campaign; and increasingly RTB functionality (targeting for higher bids individual browsers that have demonstrated certain behaviors means the question has no meaningful answer.

    Ultimately the best strategy in paid search is to focus on what works for your business and let others play games :-)

  5. Brian Bien says:

    Thanks for the RE, George.

    I believe that if for some given market, there were a sufficiently large PPC account manager with a practical monopoly on data flowing through a system of his, that then maybe the conditions would approach being met.