top of page
Writer's pictureAdam Timlett

Data Quality Management Using Game Theory & Organisation Games


These slides are part of a presentation on utilising game theory and ‘organisation games’, an extension of game theory developed to study biological systems, in order to better understand data quality management strategies and solutions.


Training your teams to utilise these analytical methods will improve the ROI of any data quality initiatives in your organisation. You will understand when to use data contracts, the risks and downsides, and also understand when powerful alternative analysis approaches can be used instead, namely organisation games. This new approach developed by Turing Meta has been developed to jointly capture problems and solutions common to both biological systems and human organisations. As a result, it leverages high quality biological science research to give businesses confidence in the value of the approach.




The slides below are my annotations of a presentation given by David Bartolini to the OECD, the Organisation for Economic Cooperation and Development, on the subject of achieving cooperation between different agencies. They show a taxonomy of situations called ‘games’ that are increasingly more difficult for the ‘players’ to achieve cooperation (i.e to ‘coordinate’), culminating in the need for ‘contracts’ to solve coordination problems.




In the slide below we start with a pure coordination game, in which it is in both player’s interests to choose the same option, option A. An example of this is where Adam and Joe both want to play a recreational sport one weekend, either Cricket or Football. Both prefer to play Football (option A), so it is relatively easy to coordinate, Adam brings a football and Joe the goalposts. However, let’s say that Adam wrongly thinks Joe prefers Cricket (option B), and Joe also wrongly thinks Adam prefers Cricket, too. Then Adam might bring a cricket bat and Joe a cricket ball, and so they both choose option B and end up getting a lower reward or ‘payoff’ simply because they didn’t communicate their preferences to each other effectively. If Adam brings a cricket bat and Joe goalposts they fail to coordinate and get no payoff (B,A) or (A,B).



In the next game below, the ‘Battle of the Sexes’, the level of alignment of objectives is lower. An example of this is if Adam and Joe want to meet, and Adam lives at home A and wants to meet at home A, and Joe wants to meet at place, B, Joe’s Home.

Adam and Joe both want to meet, but different payoffs depending on whose home we meet at, A or B. No payoff for not meeting.


It is natural in that situation that Adam ‘bribes’ or ‘incentivises’ Joe to come to Adam’s house, e.g. ‘I’ll buy you Pizza and Beer’. So this type of coordination problem is naturally solved by some negotiation.



In the final game discussed, the options are marked ‘C’ and ‘NC’ which stands for ‘Cooperate’ and ‘Non-Cooperate’ (or cheat). This game, the 'Prisoner’s Dilemma' is the most difficult to achieve coordination because both players have an incentive to agree to cooperation, and then not actually cooperate, so that the other player who does cooperate does the work, while the cheating of ‘free-riding’ player gets the benefits without the cost of doing the work.


The solution to this sort of alignment problem is usually seen as contracts, but this only works if the penalties for not cooperating are real, otherwise the contract is not worth the paper it is written on.


‘Data Contracts’ are an example of applying these ideas of using contracts to create cooperation where there is otherwise no effecive cooperation on data quality. The difficulty of agreeing and using contracts however extends beyond the mere design of them, but also to the ‘levers’ available in terms of penalties for teams not following the contracts other teams want them to agree to. Different teams may not be incentivised to agree to the contracts, or water down the level of agreement. Further, if data sources are external it is unclear why they would agree to binding contracts with real penalties.




The next slides focus on the theory of ‘organisation games’ which Turing Meta has developed to jointly study both coordination problems in biology and human organisations.


It is an extension of game theory based on the game type discussed first, ‘pure coordination games’. It is the method of analysis that leverages biological science to provide a deeper view of data quality issues and how to rationally deal with them.




The idea of 'organisation games' is to capture the idea, such as in a natural conversation, where the payoffs for everyone participating are changing as the flow of value of information changes. To begin with it might make sense for Alice to do all the talking. But at some stage it makes sense for everyone for Bob to talk more or to do all the talking. The natural flow of the conversation continuing depends on agents reacting rationally to the changing payoffs, as the game changes and adjusting their individual strategy, to talk or not to talk, accordingly.


Failures to react rationally lead to conversations in which no-one is benefitting, such as getting stuck at a party talking to someone who doesn't respond to your signals that you are bored and would like to move on to another person or group.



Organisation Games are a type of ‘pure coordination game’ a type discussed earlier, where for simplicity we assume the payoff is for the whole organisation, e.g.  for both Alice and Bob benefit the same way from the organisation. As the game changes, they payoffs they receive changes but is for the whole group, not different between members of the group. 




This next slide shows different sub-games of the conversation game discussed. Each game is a different set of group payoffs for different conversation strategies by the two players.



An ‘organisation game’ is really a set of sub-games and transitions occur as we move from one sub-game to another. Some transitions require only one player to change strategy for the payoffs for the group to be maintained, while other transitions require both players to change strategy, so are more difficult to achieve.



Now we relate ‘organisation games’ back to the problem of data quality, giving an alternative approach to the problem than classical game theory which as we’ve already seen, tends to think in terms of 'data contracts' and 'incentives'.




In biology, there is the concept of ‘genetic hitchhiking’ this is where bad genes get promoted in a population (become more frequent in that population) despite not being good for the organism in that environment, simply because they are very close to ‘good genes’ that are being selected for in that environment. Hence, in biology this problem of the ‘data quality’ of the genome is similar, and in fact, directly analogous, to data quality issues in human organisations. Organisation Games make this direct analogy even clearer. We want to play ‘organisation games’ ‘rationally’ in order to ‘de-link’ bad quality data from high value, high quality data, as far as possible.





Using both lenses of classical game theory and organisation games, will allow your organisation to take a holistic and ‘joined up’ view of data quality management and analysis.

Training your teams to see the organisation games to analyse data quality problems, will offer a higher ROI and allow:


·        More effective communication between teams,


·        Link data quality, ELT, reporting analytics, and business processes together and motivate more imaginative, flexible solutions.


·        Focus on the data quality issues that matter in your organisation and drive long term improvements and strategy.


Get in touch today to book an appointment to review your data quality strategy and training options.


Adam Timlett,



Recent Posts

See All
bottom of page