Response to discussion Paper on innovative uses of consumer data by financial institutions
Go back
However, we do not believe that cost efficiency gains obtained through savings in marketing and other costs will be automatically passed on to consumers in the form of lower prices. Lower costs for financial institutions are not generally passed on to consumers. Greater availability of data may also lead to longer chains of intermediaries between the consumer and their end product. This can lead to increased costs and conflicts of interest, particularly if intermediaries are remunerated through commission-based business models.1 Although marketing costs may be reduced for financial institutions there may be an increase in marketing costs for the new intermediaries that will help consumers interpret data and compare products.2 Academic research has suggested that an increasing number of price comparison intermediaries can increase costs for consumers. 3
It is not clear whether the supposed benefits of firms being able to use consumers’ data (including social media) to assess creditworthiness outweigh the potential risks to the consumer. There is no accepted methodology for impartial assessment of data derived from social media. It is therefore important to monitor this development carefully so that vulnerable people are not exploited and firms lend responsibly. The Panel does not agree that social media data would necessarily increase product quality for consumers but it could increase access to some products.
1. http://www.dotecon.com/assets/images/crmain.pdf
2 . Snoop Dogg doesn't work for free - 4 big comparison sites spend £110 million on advertising http://www.thisismoney.co.uk/money/bills/article-2933401/Energy-price-comparison-sites-spend-110m-annoying-adverts.html
3. http://www2.warwick.ac.uk/fac/soc/economics/research/workingpapers/2015/twerp_1056b_ronayne.pdf
A consumer could be wrongfully accused of fraud and find themselves systematically ‘locked out’ and excluded from all financial products with little recourse to appeal.
Additionally, there may be concerns that such a centralised system may operate in a ‘big brother’ way. There may thus be arguments against the development of such a system because it puts too much power in the hands of a small group of organisations and could be abused.
Financial exclusion
The Panel accepts that digitalisation can offer some customers enhanced access, especially where the products or services involved are straightforward, non-complex and readily understood.
However, use of online data could lead to discrimination, exclusion, and overplay the impression of consumer empowerment. If Big Data is commonly used in risk assessment, consumers will be forced to create an “online CV” for themselves and actively share data, or risk exclusion from some financial services.
At the same time, the use of data does not always mean consumers are offered products that are suited to their needs at more competitive prices. They may be offered products they find hard to refuse (eg. a loan in a time of need) at a price which exploits them.
Using data to assess creditworthiness /credit scores
There is evidence that consumers in the UK do not understand how credit scoring works, so it seems very unlikely that they know how their data are obtained, used or stored by the credit industry.
Consumer misunderstandings about how credit scoring works include when it is used; the role of credit reference agencies; who makes lending decisions and consumer rights to check their credit files. People commonly believe that credit reference agencies make lending decisions and individuals only have one credit score. 4
Some firms bypass traditional credit scoring and build scorecards using Big Data. However, this raises various issues for consumers, which Cullerton5 categorises as follows:
• Transparency: most computer users are unclear about who is tracking them, what data is being gathered and how it is being used. In the UK, the strengthening of ‘cookie’ policies6 has given consumers greater awareness that data are being collected as they browse the internet. However, research suggests understanding of what happens to the data is poor7.
• Consent: since how data is used is unclear, it is debatable whether consumers have really given consent as required under the Data Protection Act 1998 (Schedule 2, paragraph 1).
• Discrimination: while traditional credit scoring methods may exclude people with thin credit files, Big Data may discriminate in other ways, for example, excluding those who do not shop in particular stores or who make limited use of social media. There is also a feedback loop because data collected is used to create targeted adverts that may drive a consumer’s behaviour and habits in a particular way, thus reinforcing this social discrimination. 8
• Context: information may be freely shared in one context, such as on a social media site, but its reuse in another, such as lending decisions, may seem inappropriate both to the consumer and society more generally, and regulators should regulate such practices.
A characteristic of Big Data is that very wide and varied types of data are used collectively. Kreditech, for example, is reported to use 15,000 data points in its credit scoring algorithm.9 This makes it even more difficult to give consumers an indication of the reason if they are declined for credit. There is also a question mark over how relevant all this data might be and whether it can be justified under data protection legislation.10
Given the above, the Panel strongly believes that the use of Big Data for calculating credit scores should be subject to the consumer’s explicit consent and consumers should be able to choose the types of data they are willing to have included in their credit assessment. This would help them know that data is collated from different sources and give them control over what they would be happy to share.
Using data to assess risk (insurance)
Insurers’ increased use of Big Data to inform risk and pricing strategies requires careful consideration as it may increase access for some groups of consumers while restricting it for others.
Use of such data offers firms opportunities for increasingly individualised risk assessments, which would have a significant impact on risk pooling and individual premiums. In the longer term, the use of Big Data could, for example, fundamentally alter the structure of the insurance industry, as the pooling of risk would reduce significantly. The use of individualised micro risk assessments means that some people are likely not to be served at all. Others will pay much higher premiums. Conversely of course, some consumers, for instance some young drivers or elderly travellers, should pay lower premiums because their individual risk, or the average risk in their pool, is lower.
The level of transparency in risk profiling is another issue of concern. At present it is unclear how firms assess risk and it is impossible for individuals to know if they are getting value for money as a consequence. Big Data makes this even more opaque as algorithms are used to identify potential risks. Consumers are not able to check the methodology, nor to correct their own behaviour or attributes to improve their ‘score’.
Data optimisation
Firms’ optimisation of data will increase the disadvantage felt by consumers in financial services and further balance power in the favour of the firm.
Big Data enables firms to use information about potential and existing customers that is not risk-related to “optimise” the price of their product, by estimating more accurately the price increase an individual consumer will put up with before they switch to a different provider. Similarly, data on individual customers’ propensity to shop around could be used to inflate prices for loyal customers. 11 Financial institutions already rely on consumer inertia to drive their business models – for instance, using low insurance premiums to gain customers but systematically increasing the premium year on year for loyal customers. Given there are already low levels of switching in many financial services markets, optimisation could further reduce value for the consumer and make the regulator’s job of promoting competition a lot harder.
Control of information
Ownership of data
The Discussion Paper mentions the exploitation and sale of data. However, ownership of data particularly in the payments sector, where there can be several parties in a transaction at one time, also needs to be clarified. For example, a payment online involves at least three parties: the payer, the payee and the bank (assuming they use the same bank). With the introduction of payment account services and payment initiation services the chain is likely to become longer. Which party in the chain owns the data? The payer, the payee, their respective banks, the Payment Initiation Service or the platform through which the payment was made online?
Ideally, consumers should own their data and be able to ‘plug it in’ to a provider of their choice and ‘unplug’ it at will. However, there are currently no tools available that allow consumers to control their data in this way, or their online ‘CVs’ that firms are developing without their permission, or even knowledge.
Ability to modify data
We strongly agree with the EBA that many consumers do not understand that their data is being captured, stored or used and that this creates risks both to the consumer and the reputation of the financial institution. People are not able to control the storage or use of their data, nor influence or change how it is used.
This means that ‘standard practices’ in data collection, storage and usage may develop with little consideration for the consumer.
Tools to enable the consumer to have control over their own data in a way that actually works for them should be developed as a priority and governed properly. Otherwise there is a risk that regulation will be needed ‘after the event’.
Lack of redress for inaccurate data
It is not clear whether consumers have a right of redress against firms making use of inaccurate or misleading data. The creator and submitter of the data may not be a financial services firm and may therefore be outside the jurisdiction of the regulator and the relevant ADR scheme. The EBA could explore making the use of the data liable for any inaccuracies. This would encourage firms to check the quality of the data they use.
In research by Which?12, 81 volunteers ordered statutory reports from all three UK credit reference agencies. A third found a problem on their file which they disputed. In addition, a third of participants found the information full of jargon and confusing to understand, despite a legal requirement that statutory reports should be given in plain English (Data Protection Act 1998, s158 (5)).
A study in the USA by the National Consumer Law Center found that 20 per cent of traditional credit reports contain errors, with around a quarter of these errors reducing credit scores. New-style Big Data credit scoring was also been shown to be inaccurate. Again drawing on US research, obtaining the data was challenging, the reports were not comprehensive and between 67 and 100 per cent of them contained inaccuracies which ‘ranged from the mundane—a wrong e-mail address or incorrect phone number—to seriously flawed’ (National Consumer Law Center, 2014: 4). 13
Sharing data
The Panel believes that the use of Big Data by providers should be subject to the consumer’s explicit consent. We also believe that this consent should not extend – nor be required to extend - beyond the immediate product sale. That is, the provision of a product or service should not be conditional on the consumer allowing their data to be retained and used by the firm for other purposes. The request for consent should be clear and explicit, and not buried within a long list of terms and conditions.
Re-identification of data
The risk of ‘re-identification’ is a particular concern. Anonymised data that is made available publicly may enable the identification of individual consumers with significant consequences. For instance, firms wishing to avoid consumer consent could buy anonymised data, process it to identify individuals, and use it without the consumer’s knowledge or permission.
Governance of algorithms
There needs to be clearer governance to regulate how Big Data analytics develop and how their quality can be assured. Data analytics is a developing science with a lot of room for experimentation and development. There is a risk of placing undue weight on the output of algorithms that are subsequently found to be faulty, for example, some of Google’s early attempts to predict the spread of the flu virus. 14 Algorithms are set up and ‘trained’ by humans with their own commercial objectives and natural human biases. This can affect the quality of algorithms or the use to which they are put.
Algorithms can also pose systemic risks. A recent paper on ‘robo-advice’ from a UK investment firm suggests that robo-advice firms are potentially storing up large liabilities if their algorithms fail further down the line. 15 Another industry paper highlights the practical issues that robo-advice firms need to consider and plan for, including processes for temporarily suspending algorithm-driven advice if, for example, there are surprise changes to legislation that require algorithms to be modified. 16
Asymmetries of data sharing
The benefits of Big Data appear to rely primarily on consumers sharing their data. However, there is little onus on banks and other financial institutions to offer commensurate exposure to data on the quality of their products and services. There is a vast amount of ‘hidden’ information that consumers may find helpful in making informed choices about financial products. Because this information is not currently available it makes it very difficult for consumers to effect market power or take more responsibility in the market. The balance of power is too firmly in favour of firms and further opening up access to consumer data exacerbates this. Firms should be required to provide much more detailed information about their products in order to re-balance the asymmetry of power and increase competition in the market.
This will be particularly important as technology allows services to develop which will in effect, assess the market for a consumer. Availability of data to enable good assessments will be necessary and technology will provide the capacity for all data to be considered in a way which is not currently possible with the limitations of a comparison table or human processing capability.
4. Experian (2015) “Common credit myths. Help and advice http://www.experian.co.uk/consumer/credit-education/common-credit-myths.html
5. Cullerton, N (2013). Behavioural credit scoring’ in Georgetown Law Journal, 2012-13, Issue 3, March http://georgetownlawjournal.org/files/2013/03/Cullerton.pdf
6. Among other measures, the Privacy and Electronic Communications Regulations (PECR) require that people give consent before information can be gathered from their computers. A cookie is a small file of information stored on a user’s computer to be sent to a website.
7. Information Commissioners Office (2012) Guidance on the rules on use of cookies and similar technologies https://ico.org.uk/media/for-organisations/documents/1545/cookies_guidance.pdf
8. Information Commissioners Office (2014) “Big data and data protection” https://ico.org.uk/media/for-organisations/documents/1541/big-data-and-data-protection.pdf
9. King, J. (2014). ‘IMF World Bank: Credit scoring – friends, followers and settling scores’ in The Banker, October
10. Information Commissioners Office (2014) “Big data and data protection” https://ico.org.uk/media/for-organisations/documents/1541/big-data-and-data-protection.pdf
11. The issue of price optimisation has already received significant attention in the United States. See for example, Consumer Federation of America & Center for Economic Justice, “Comments on CASTF’s Draft Price Optimization White Paper” (June 2015)
12. Which? (2014) “Unlocking your credit report” https://press.which.co.uk/whichpressreleases/unlocking-your-credit-report-2/
13. National Consumer Law Center (2014). Big data. A big disappointment in scoring consumer credit risk https://www.nclc.org/images/pdf/pr-reports/report-big-data.pdf
14. http://www.wired.com/2015/10/can-learn-epic-failure-google-flu-trends/
15. SCM Direct (2016). Fintech Folly: the sense and sensibilities of UK robo-advice https://scmdirect.com/press-and-videos#block-views-resources-scm-research-tab
16. Storey, A. (2016). How to monitor robo-advice https://www.linkedin.com/pulse/how-monitor-robo-advice-andrew-storey
6. Do you consider the potential benefits described in this chapter to be complete and accurate? If not, what other benefits do you consider should be included?
The Panel recognises the benefits that the use of information by financial services firms can deliver for consumers.However, we do not believe that cost efficiency gains obtained through savings in marketing and other costs will be automatically passed on to consumers in the form of lower prices. Lower costs for financial institutions are not generally passed on to consumers. Greater availability of data may also lead to longer chains of intermediaries between the consumer and their end product. This can lead to increased costs and conflicts of interest, particularly if intermediaries are remunerated through commission-based business models.1 Although marketing costs may be reduced for financial institutions there may be an increase in marketing costs for the new intermediaries that will help consumers interpret data and compare products.2 Academic research has suggested that an increasing number of price comparison intermediaries can increase costs for consumers. 3
It is not clear whether the supposed benefits of firms being able to use consumers’ data (including social media) to assess creditworthiness outweigh the potential risks to the consumer. There is no accepted methodology for impartial assessment of data derived from social media. It is therefore important to monitor this development carefully so that vulnerable people are not exploited and firms lend responsibly. The Panel does not agree that social media data would necessarily increase product quality for consumers but it could increase access to some products.
1. http://www.dotecon.com/assets/images/crmain.pdf
2 . Snoop Dogg doesn't work for free - 4 big comparison sites spend £110 million on advertising http://www.thisismoney.co.uk/money/bills/article-2933401/Energy-price-comparison-sites-spend-110m-annoying-adverts.html
3. http://www2.warwick.ac.uk/fac/soc/economics/research/workingpapers/2015/twerp_1056b_ronayne.pdf
7. Are you aware of any barriers that prevent financial institutions from using consumer data in a beneficial way? If so, what are these barriers?
Data could be used more effectively to reduce fraud and financial crime and improve the identification and authentication process for consumers. For instance, it should be possible to move a bank account without being required to provide all your details again. However, this would require new ways of storing and sharing data securely. This poses questions about data protection and legislation, which may create a barrier. Storing data centrally poses risks to consumers as the prize for hacking such a system would be high. Storing such data on a distributed system may help get over this barrier.A consumer could be wrongfully accused of fraud and find themselves systematically ‘locked out’ and excluded from all financial products with little recourse to appeal.
Additionally, there may be concerns that such a centralised system may operate in a ‘big brother’ way. There may thus be arguments against the development of such a system because it puts too much power in the hands of a small group of organisations and could be abused.
8. Do you consider the potential risks described in this chapter to be complete and accurate? If not, what other risks do you consider should be included?
The Panel agrees with the risks identified. In addition, the Panel would like to focus on additional areas of concern which highlight the increased asymmetries of power created by the use of Big Data.Financial exclusion
The Panel accepts that digitalisation can offer some customers enhanced access, especially where the products or services involved are straightforward, non-complex and readily understood.
However, use of online data could lead to discrimination, exclusion, and overplay the impression of consumer empowerment. If Big Data is commonly used in risk assessment, consumers will be forced to create an “online CV” for themselves and actively share data, or risk exclusion from some financial services.
At the same time, the use of data does not always mean consumers are offered products that are suited to their needs at more competitive prices. They may be offered products they find hard to refuse (eg. a loan in a time of need) at a price which exploits them.
Using data to assess creditworthiness /credit scores
There is evidence that consumers in the UK do not understand how credit scoring works, so it seems very unlikely that they know how their data are obtained, used or stored by the credit industry.
Consumer misunderstandings about how credit scoring works include when it is used; the role of credit reference agencies; who makes lending decisions and consumer rights to check their credit files. People commonly believe that credit reference agencies make lending decisions and individuals only have one credit score. 4
Some firms bypass traditional credit scoring and build scorecards using Big Data. However, this raises various issues for consumers, which Cullerton5 categorises as follows:
• Transparency: most computer users are unclear about who is tracking them, what data is being gathered and how it is being used. In the UK, the strengthening of ‘cookie’ policies6 has given consumers greater awareness that data are being collected as they browse the internet. However, research suggests understanding of what happens to the data is poor7.
• Consent: since how data is used is unclear, it is debatable whether consumers have really given consent as required under the Data Protection Act 1998 (Schedule 2, paragraph 1).
• Discrimination: while traditional credit scoring methods may exclude people with thin credit files, Big Data may discriminate in other ways, for example, excluding those who do not shop in particular stores or who make limited use of social media. There is also a feedback loop because data collected is used to create targeted adverts that may drive a consumer’s behaviour and habits in a particular way, thus reinforcing this social discrimination. 8
• Context: information may be freely shared in one context, such as on a social media site, but its reuse in another, such as lending decisions, may seem inappropriate both to the consumer and society more generally, and regulators should regulate such practices.
A characteristic of Big Data is that very wide and varied types of data are used collectively. Kreditech, for example, is reported to use 15,000 data points in its credit scoring algorithm.9 This makes it even more difficult to give consumers an indication of the reason if they are declined for credit. There is also a question mark over how relevant all this data might be and whether it can be justified under data protection legislation.10
Given the above, the Panel strongly believes that the use of Big Data for calculating credit scores should be subject to the consumer’s explicit consent and consumers should be able to choose the types of data they are willing to have included in their credit assessment. This would help them know that data is collated from different sources and give them control over what they would be happy to share.
Using data to assess risk (insurance)
Insurers’ increased use of Big Data to inform risk and pricing strategies requires careful consideration as it may increase access for some groups of consumers while restricting it for others.
Use of such data offers firms opportunities for increasingly individualised risk assessments, which would have a significant impact on risk pooling and individual premiums. In the longer term, the use of Big Data could, for example, fundamentally alter the structure of the insurance industry, as the pooling of risk would reduce significantly. The use of individualised micro risk assessments means that some people are likely not to be served at all. Others will pay much higher premiums. Conversely of course, some consumers, for instance some young drivers or elderly travellers, should pay lower premiums because their individual risk, or the average risk in their pool, is lower.
The level of transparency in risk profiling is another issue of concern. At present it is unclear how firms assess risk and it is impossible for individuals to know if they are getting value for money as a consequence. Big Data makes this even more opaque as algorithms are used to identify potential risks. Consumers are not able to check the methodology, nor to correct their own behaviour or attributes to improve their ‘score’.
Data optimisation
Firms’ optimisation of data will increase the disadvantage felt by consumers in financial services and further balance power in the favour of the firm.
Big Data enables firms to use information about potential and existing customers that is not risk-related to “optimise” the price of their product, by estimating more accurately the price increase an individual consumer will put up with before they switch to a different provider. Similarly, data on individual customers’ propensity to shop around could be used to inflate prices for loyal customers. 11 Financial institutions already rely on consumer inertia to drive their business models – for instance, using low insurance premiums to gain customers but systematically increasing the premium year on year for loyal customers. Given there are already low levels of switching in many financial services markets, optimisation could further reduce value for the consumer and make the regulator’s job of promoting competition a lot harder.
Control of information
Ownership of data
The Discussion Paper mentions the exploitation and sale of data. However, ownership of data particularly in the payments sector, where there can be several parties in a transaction at one time, also needs to be clarified. For example, a payment online involves at least three parties: the payer, the payee and the bank (assuming they use the same bank). With the introduction of payment account services and payment initiation services the chain is likely to become longer. Which party in the chain owns the data? The payer, the payee, their respective banks, the Payment Initiation Service or the platform through which the payment was made online?
Ideally, consumers should own their data and be able to ‘plug it in’ to a provider of their choice and ‘unplug’ it at will. However, there are currently no tools available that allow consumers to control their data in this way, or their online ‘CVs’ that firms are developing without their permission, or even knowledge.
Ability to modify data
We strongly agree with the EBA that many consumers do not understand that their data is being captured, stored or used and that this creates risks both to the consumer and the reputation of the financial institution. People are not able to control the storage or use of their data, nor influence or change how it is used.
This means that ‘standard practices’ in data collection, storage and usage may develop with little consideration for the consumer.
Tools to enable the consumer to have control over their own data in a way that actually works for them should be developed as a priority and governed properly. Otherwise there is a risk that regulation will be needed ‘after the event’.
Lack of redress for inaccurate data
It is not clear whether consumers have a right of redress against firms making use of inaccurate or misleading data. The creator and submitter of the data may not be a financial services firm and may therefore be outside the jurisdiction of the regulator and the relevant ADR scheme. The EBA could explore making the use of the data liable for any inaccuracies. This would encourage firms to check the quality of the data they use.
In research by Which?12, 81 volunteers ordered statutory reports from all three UK credit reference agencies. A third found a problem on their file which they disputed. In addition, a third of participants found the information full of jargon and confusing to understand, despite a legal requirement that statutory reports should be given in plain English (Data Protection Act 1998, s158 (5)).
A study in the USA by the National Consumer Law Center found that 20 per cent of traditional credit reports contain errors, with around a quarter of these errors reducing credit scores. New-style Big Data credit scoring was also been shown to be inaccurate. Again drawing on US research, obtaining the data was challenging, the reports were not comprehensive and between 67 and 100 per cent of them contained inaccuracies which ‘ranged from the mundane—a wrong e-mail address or incorrect phone number—to seriously flawed’ (National Consumer Law Center, 2014: 4). 13
Sharing data
The Panel believes that the use of Big Data by providers should be subject to the consumer’s explicit consent. We also believe that this consent should not extend – nor be required to extend - beyond the immediate product sale. That is, the provision of a product or service should not be conditional on the consumer allowing their data to be retained and used by the firm for other purposes. The request for consent should be clear and explicit, and not buried within a long list of terms and conditions.
Re-identification of data
The risk of ‘re-identification’ is a particular concern. Anonymised data that is made available publicly may enable the identification of individual consumers with significant consequences. For instance, firms wishing to avoid consumer consent could buy anonymised data, process it to identify individuals, and use it without the consumer’s knowledge or permission.
Governance of algorithms
There needs to be clearer governance to regulate how Big Data analytics develop and how their quality can be assured. Data analytics is a developing science with a lot of room for experimentation and development. There is a risk of placing undue weight on the output of algorithms that are subsequently found to be faulty, for example, some of Google’s early attempts to predict the spread of the flu virus. 14 Algorithms are set up and ‘trained’ by humans with their own commercial objectives and natural human biases. This can affect the quality of algorithms or the use to which they are put.
Algorithms can also pose systemic risks. A recent paper on ‘robo-advice’ from a UK investment firm suggests that robo-advice firms are potentially storing up large liabilities if their algorithms fail further down the line. 15 Another industry paper highlights the practical issues that robo-advice firms need to consider and plan for, including processes for temporarily suspending algorithm-driven advice if, for example, there are surprise changes to legislation that require algorithms to be modified. 16
Asymmetries of data sharing
The benefits of Big Data appear to rely primarily on consumers sharing their data. However, there is little onus on banks and other financial institutions to offer commensurate exposure to data on the quality of their products and services. There is a vast amount of ‘hidden’ information that consumers may find helpful in making informed choices about financial products. Because this information is not currently available it makes it very difficult for consumers to effect market power or take more responsibility in the market. The balance of power is too firmly in favour of firms and further opening up access to consumer data exacerbates this. Firms should be required to provide much more detailed information about their products in order to re-balance the asymmetry of power and increase competition in the market.
This will be particularly important as technology allows services to develop which will in effect, assess the market for a consumer. Availability of data to enable good assessments will be necessary and technology will provide the capacity for all data to be considered in a way which is not currently possible with the limitations of a comparison table or human processing capability.
4. Experian (2015) “Common credit myths. Help and advice http://www.experian.co.uk/consumer/credit-education/common-credit-myths.html
5. Cullerton, N (2013). Behavioural credit scoring’ in Georgetown Law Journal, 2012-13, Issue 3, March http://georgetownlawjournal.org/files/2013/03/Cullerton.pdf
6. Among other measures, the Privacy and Electronic Communications Regulations (PECR) require that people give consent before information can be gathered from their computers. A cookie is a small file of information stored on a user’s computer to be sent to a website.
7. Information Commissioners Office (2012) Guidance on the rules on use of cookies and similar technologies https://ico.org.uk/media/for-organisations/documents/1545/cookies_guidance.pdf
8. Information Commissioners Office (2014) “Big data and data protection” https://ico.org.uk/media/for-organisations/documents/1541/big-data-and-data-protection.pdf
9. King, J. (2014). ‘IMF World Bank: Credit scoring – friends, followers and settling scores’ in The Banker, October
10. Information Commissioners Office (2014) “Big data and data protection” https://ico.org.uk/media/for-organisations/documents/1541/big-data-and-data-protection.pdf
11. The issue of price optimisation has already received significant attention in the United States. See for example, Consumer Federation of America & Center for Economic Justice, “Comments on CASTF’s Draft Price Optimization White Paper” (June 2015)
12. Which? (2014) “Unlocking your credit report” https://press.which.co.uk/whichpressreleases/unlocking-your-credit-report-2/
13. National Consumer Law Center (2014). Big data. A big disappointment in scoring consumer credit risk https://www.nclc.org/images/pdf/pr-reports/report-big-data.pdf
14. http://www.wired.com/2015/10/can-learn-epic-failure-google-flu-trends/
15. SCM Direct (2016). Fintech Folly: the sense and sensibilities of UK robo-advice https://scmdirect.com/press-and-videos#block-views-resources-scm-research-tab
16. Storey, A. (2016). How to monitor robo-advice https://www.linkedin.com/pulse/how-monitor-robo-advice-andrew-storey