Alex Chisholm speaks about data and trust in digital markets
Speech given by CMA Chief Executive, Alex Chisholm, at the UEA Centre for Competition Policy Annual Conference in Norwich.
Data and trust in digital markets: what are the concerns for competition and for consumers?
I’m delighted to be here at UEA [the University of East Anglia] again for the Annual Conference of the Centre for Competition Policy, and especially so given your well-chosen theme of ‘Competition in the Digital Age’ (1 - see footnotes at the end). Every market is experiencing change from the challenge of digital. New business models and industry structures have emerged, disrupting established players and practices, and user behaviour continues to evolve, not always in expected directions.
Digital markets have brought significant benefits for consumers: improved services that offer more flexibility like online grocery shopping; greater choice between many more providers than you would be able to access in the real world; better price discovery because more information is available; better quality discovery through online review sites where you can read what experience other consumers have had of a hotel or restaurant. (Of course not all consumers can access these benefits because they are digitally excluded, but that is outside the scope of my talk today.)
In the UK we have embraced the opportunity of digital with remarkable enthusiasm. The UK is among the most advanced ecommerce users of the large western economies, with 13.5% of our retailing online (vs 11.5% in US and 9.7% in Germany, European average of 7.2%) (2). The average UK household will spend over £2,000 online this year (3) reflecting a growth in online sales of over 13% (4).
The CMA’s mission is to make markets work well, and this is the perspective we bring to digital markets. So the questions we are asking ourselves are what is happening that is in the interests of consumers, firms and the economy? What are the potential costs and risks? And where is attention justified from authorities like ourselves to tilt the balance towards more favourable outcomes?
A key focus to date has been on the question of inter-platform competition. The International Competition Network recently published work on this to bring together international perspectives. It covers problems such as vertical restraints, resale price maintenance and online sales bans. In September last year my colleague Philip Marsden spoke at the 11th Baltic Competition Conference about the range of competition cases in this area that the UK had dealt with.
Yesterday you heard from my colleague Tony Curzon-Price, who discussed ways in which price comparison websites can act as both friends and enemies of the competitive process. Tony and I worked together on a speech I gave last December at the CRA conference in Brussels which explored the challenges of inter-platform competition more widely. So I don’t propose to repeat this ground today.
Instead, I want to talk about 2 aspects of the epochal change we’re going through in business models brought on by tech. Many of the transformations we’re seeing revolve around matching goods and their buyers in new and better ways. If we think of a pre-web age, many search and trust problems were solved by shops and by brands. Shops presented goods to consumers who browsed their shelves; consumers trusted that the match made in the shop would be good because brands – of the retailers and of the goods – provided assurance. Consumer protection laws played a part too, of course, in providing a basic level of assurance.
In the world of ecommerce, the choice set a consumer is presented with is more often than not tailored to that individual through use of consumer data, and the choices themselves become data that influence future choice sets. And the ease of tailoring information means that many products and services that were never branded in a pre-web age become the object of possible choices. Think of the Airbnb I might have used here in Norwich if you hadn’t so generously put me up. That B&B is way out of the branding economy. But I would have had my trust levels raised by reading the reviews and feedback on Airbnb. And this story is being repeated across many, many new ways of buying: user feedback is functioning in the way that brands – and even possibly consumer protection – used to function.
In other words, the disruptions being wrought by ecommerce are raising 2 linked sets of fundamental questions:
- How are the markets for data working?
- Can reviews and ratings be trusted to perform the important roles they’re coming to play in the digital economy?
The CMA has carried out 2 calls for information (on the commercial use of consumer data and on online reviews and endorsements) covering this complex nexus of issues these last few months and this week we have published the results.
In both cases, we find exciting and powerful new solutions to the age-old problems of search and trust. But we shouldn’t be Panglossian about this: new solutions bring their own challenges.
Caricaturing the nay-sayers just a bit, they think that we’ve signed a ‘Faustian pact’ (5):
-
that consumers are helpless to control the use to which their data is put
-
that data – often harvested by apparently ‘free’ services – is being transformed into alarming market power by just a few digital giants
-
that review sites now hold great arbitrary power over businesses actually providing goods and services
Harm to victims is alleged – often by those who used to have good businesses supplying the alleged victims rather than the victims themselves – and the self-styled Cassandras predict ruin and catastrophe. They call for regulation to halt the changes.
We wanted to take our own look at both areas to assess how the markets for data and reputation are working and evolving.
Let me sketch out some of the key benefits we have seen. First, what are the benefits for me as a consumer?
-
Search and comparison has become orders of magnitude easier for many products and services.
-
The information on which I can make informed product choices has become much more accessible – my plumber is rated by hundreds of people rather than someone who once did an OK job for my neighbour years ago.
-
Product discovery – the things that add to my welfare that I didn’t know would do so – has improved by leaps and bounds: how in the old world would I have ever come across the music of Northern Mali that Spotify (via my colleague Tony) has allowed me to discover? Or Nicolas Bouvier’s wonderful book ‘The Way of the World’ (published in French in 1963)?
-
Product improvement – firms now have access to rapid and fine-grained feedback to iterate a product towards what works best – game developers, for example, track playing to improve the experience on the fly.
The markets for data have provided consumers with many ‘free’ services like search and social media that we pay for through the data we provide and the advertising that we click through. The evidence suggests that consumers in general don’t particularly like such advertising but equally they are unwilling to pay directly for the service.
And then there are the benefits arising from the supplier response:
-
Increased competition means better quality and lower prices. Think of hotels: how would Fawlty Towers fare on TripAdvisor?
-
Entry has become much easier in many markets because access to a large and targeted audience has become much cheaper.
-
Technical efficiency has increased as the cost of trial and error has fallen (for example the clothes industry test new styles and colours on social media to gauge interest) and data has improved operations and logistics.
These benefits are undeniable. So what’s not to like?
First, I will talk about consumer data. Well-functioning markets need informed and active consumers. On the face of it, at the moment, a lot of data is flowing without hindrance from the consumer to the firm. This might suggest “there is nothing to see here”.
But there are some danger signals, and consumer sentiment towards data sharing may be fragile.
-
There is low awareness – beyond the most visible use for advertising – of the scale and extent of data-use. The ICO reported recently that nearly two-thirds of the public consider they have lost control over how their information is collected and processed (6).
-
There is underlying mistrust of the organisations that collect data. The Royal Statistical Society reported that trust in data use was ‘low for all institutions especially those that rely heavily on data’ (7).
-
A significant minority of consumers – around 30% – have major concerns about privacy and say they do not want to share their data at all (8).
These danger signals are exacerbated by the fact that the mechanisms that allow consumers to make choices about data use are weak. Privacy policies and terms and conditions are long and complex. It can be hard to opt out of data collection, or to be aware what firms are collecting data. And the pressure on consumers is only set to increase. Developments such as the Internet of Things – like online devices we wear or carry and devices in the home or in our cars – will mean that data is collected and shared on a regular basis without the consumer having to make a conscious decision. This may test how far consumers will remain willing to provide data. Jonathan Porter from Ofcom will say more about this in the next talk this morning.
The risk is that significant numbers of consumers don’t maintain trust in firms and decrease or stop their data-sharing, and so online firms may cease to provide a range of valuable services.
The ideal solution to this problem is consumer engagement, and our report sets out how firms could help consumers become more active and informed, including suggesting simpler information about how data is used and how consumers benefit from use; and better mechanisms that enable consumers to choose how much, if any, data to share.
Data may also raise classic competition issues. These have been an area of active debate among governments in the European Union (EU), competition authorities, academics and commentators. One argument made is that data offers firms opportunities to create sustainable competitive advantage. And once a firm has a lot of data, network effects mean that it both gets even more data and can extract even more value from it, crowding out competitors. There is a sense of a land-grab going on.
But there are respectable counter arguments: that data is actually pretty freely available; that it is non-rivalrous (making a mockery of the notion of a land-grab); and that its value tends to degrade rapidly.
We found that the role of consumer data varies markedly across different sectors. We commissioned research from the economic consultancies DotEcon and Analysys Mason, which is published on GOV.UK. For example, in insurance it aids risk assessment, whereas in fashion retailing it can help drive sales through recommendations and cutting costs (for example reducing returns by providing information on size and how garments fit). Some firms choose to hold onto the information they collect – eg loyalty card data – so that they can exploit the commercial opportunities in it. Other firms become data sellers and offer opportunities for others to create targeted leads.
This variety argues against over-generalising about the market impact of consumer data collection and use, in particular the competition aspects, where wider conditions in the specific market may be as important as the data elements. We have, however, developed a number of indicators of where a specific market that collects and uses data may be more likely to generate competition concerns:
-
Simply the degree to which data is an important element in the competitive proposition of a firm is a first filter.
-
Competition concerns may be more likely to arise where firms use exclusive access to certain data to raise rivals’ costs or otherwise leave rivals at a disadvantage. Where there are few or no substitutes for this data, these actions may for example prevent entry and expansion.
-
Firms with existing power in a market may come to control the collection of consumer data, for example customers’ purchase history. The value of this data may give them an additional incentive to exploit that power.
-
The existence or not of competition over privacy (for example subscription options that make less use of data and harvest less data) may indicate whether consumers are driving effective competition. The absence of such markets may also indicate that consumers do not have a clear understanding of the collection, value and final uses of consumer data.
For each of these characteristics, a competition assessment would need to differentiate between the use of consumer data to generate efficiencies for firms and consumers, and the collection and use which might lead to exclusionary or exploitative conduct. When looking at any specific market, or case, it would also be important to think about the value of data. Some data loses value over time, so it is hard to see how persistent, unmatchable competitive advantage could be maintained. However some data has persistent value – for example in relation to customer transaction history on auction sites – and it is easier to see how the control of this data could become a barrier to entry.
We carried out this work because we recognised that data will become a bigger issue for us across the full range of work we do. We also recognise that there are a broader range of concerns relating to digital markets that are not the responsibility of a competition authority, in particular privacy, security and data protection.
One role for us in digital markets is to think about how best to incentivise competition to promote these non-economic aims. Where a market is working well, and there is healthy competition to win the custom of consumers, and where the consumer understands and values non-price aspects of a product like privacy, you would expect to see firms competing to provide a product that reflects that demand. We want to work with other regulators, in particular the Information Commissioner’s Office in the UK, and the EU institutions, to try and achieve that while minimising the need for regulation.
Data is transforming the efficiency with which goods and consumers are matched. But what about consumer assessments of the likely quality of those matches?
This brings me to our conclusions on online reviews and endorsements. Overall, we found that they provide information that is very valuable for UK consumers. We estimate that more than half of UK adults (54%) use online reviews, and that 6% use blogs or vlogs before making purchases. Most buyers find that the product or service they have bought after researching it online matches their expectations.
You would expect that bloggers and review sites have a strong incentive to ensure users get good outcomes – and we have seen some good practice that reflects this. For example, some reviews sites have developed systems to: detect and verify fake reviews; identify and give greater prominence to reviews that are likely to be more helpful; and enable users to flag suspicious-looking reviews themselves. Another interesting development is that a number of providers have entered the sector offering to manage the collection, verification and publication of reviews on behalf of retailers and others.
All of this is positive – but our findings suggest that there is no room for complacency. We have received information about a number of business practices that may breach consumer law and that have the potential to be damaging for affected businesses and consumers. These include: fake reviews being posted onto review sites; genuine negative reviews being suppressed; and businesses paying for endorsements in blogs and other online articles without ensuring that this has been made obvious to consumers.
We are also concerned that these practices may negate some of the pro-competitive benefits that I set out earlier. They may have led to some good businesses that play by the rules losing custom because of the behaviour of other non-compliant businesses. They may also distort the information that is being presented to consumers in a way that might prevent them choosing – or even becoming aware of – products and services that best suit their needs.
In deciding what to do next, it is important to consider whether the problems that we have seen are sustainable in anything but the short term.
To use some economics jargon for a moment, we would not expect a pooling equilibrium to be sustainable, so we should expect to see a separating equilibrium. A separating equilibrium in this case is where reviews correctly characterise the value of the relevant good (ie good products get good reviews and bad ones get bad reviews). The alternative would be a pooling equilibrium in which reviews do not convey accurate information. This would not be sustainable as such reviews have no value. Given that we see ubiquitous online reviews and that consumers say they find them useful, this implies we are in a separating equilibrium. It is interesting to understand why such an equilibrium is sustainable. In particular, such an equilibrium is only sustainable if the cost for bad product suppliers to pretend to be good product suppliers (ie fake reviews) is too high to be worthwhile. This may be because of the willingness of consumers who feel duped to share their disappointment rapidly and effectively with other online consumers. When talking about online reviews, I think we could usefully frame it around a discussion of this issue.
We can also reflect that markets tend to be pretty good at addressing information asymmetries. If review sites were of little value to consumers, we would all likely stop using them, unless those that published reviews took action to address this. As we have seen in many online markets, businesses that have strong positions can find that those positions can be eroded very quickly if customers lose trust in them. We would therefore expect to see a separating equilibrium, where reviews correctly characterise the value of the relevant product or service.
However, looking at the specific example of genuine negative reviews being suppressed by some sites, there seem to be a number of explanations why a ‘correction’ in the sector may be slow to occur:
-
First, the practices are clandestine. We might therefore expect consumers to have to learn from their experiences of the products and services that they buy.
-
A second related point is that consumers often use these tools to make one-off decisions. Three of the sectors where we have seen them used most are travel, electronics and home repairs. It is therefore likely that it will take longer for consumers in these sectors to learn lessons and adapt their behaviour.
-
Nevertheless, if hundreds of people were having a horrendous experience as a result of reading online reviews we might expect that some would find other ways of sharing their experiences – including social media. We have, however, identified a third reason why a ‘correction’ in the sector may be slow to occur: information provided may still be of some limited use to consumers even where it presents a distorted picture. Our qualitative research suggests that review sites are sometimes used as a ‘last resort’, when consumers have exhausted other options. Reading a selection of positive reviews about a number of relevant products or services may help a consumer to choose a product that meets their needs, even if the distorted set of information presented makes it very difficult to select the best option for them.
The blurring of the boundaries between content and commerce, so that online copy looks like editorial but is in reality advertising, might appear less obviously wrong, than posting a fake review – but still has the potential to be harmful. We expect newspapers and magazines to clearly label promotional material to alert the reader as to its nature eg ‘Advertising Supplement’. However, it appears that standards on some blogs and other online publications may not always be so pristine. In the course of our recent work we have received information about many potential instances of businesses buying undisclosed endorsements.
This matters because as consumers we are not passive: we behave differently when we know we are experiencing advertising from when we believe we are receiving unbiased, impartial advice. In the face of advertising we have learnt to take on a protective shield of scepticism, and aim off or ‘add salt’ when assessing some of the bigger claims made. The commercial temptation to find ways to sell to the ‘unshielded consumer’ is plain. An example of one issue where we believe an intervention may help to encourage compliance with consumer law in these sectors is the non-disclosure of paid endorsements. We have announced today that we have opened an investigation into the practices of a number of parties in connection with this practice.
We have learned a lot from doing both reviews. It has given the CMA a strong foundation of knowledge about these activities and markets. It has helped us clarify where problems are arising or may arise in future. That will help us decide whether and how to act when specific problems are brought to our attention.
We also learned about the rapid change that digital markets are bringing about. And that makes us believe that care is needed when considering additional regulation. There may be problems in markets today, but these may be transient, and over time, the competitive process may find ways to resolve these without our input. There are strong downside risks of regulating fast moving markets, where this is based on existing businesses and business models, as this may lock in incumbents, raise costs for new entrants and deter potentially valuable innovation.
The most significant regulatory activity under way at the moment is in the EU. We will be feeding in our knowledge on digital markets, particularly from these 2 reports: to the European Commission’s recently opened sector inquiry into ecommerce, and the wider Digital Single Market programme of work. In doing so we will be stressing the need to look at online markets, whether they be data markets, online platforms or review sites on a case-by-case and market-by-market basis. Our analysis has shown that important factors on the demand side include: how frequently consumers purchase the relevant product, how quickly the underlying product quality is revealed and how important information is to the purchase decision. On the supply side important factors include: the scope for data advantages to lead to foreclosure (how long does the data remain an important competitive advantage) and what are the incentives for suppliers to share or withhold data. We will also stress the need to understand the dynamic nature of these markets, and that many of them are evolving rapidly to reach a new equilibrium between the consumer and the firm.
Finally, I want to say that I am pleased that my first opportunity to talk about these reports is at the CCP conference. UEA Centre for Competition Policy has rightly identified this topic as one of great importance, and one where there needs to be well-informed and well-evidenced debate if society is going to make the right decisions on how these markets develop.
Our reports are not ‘The Answer’. However, they draw evidence together in a way that has not been done before in these areas and – I hope – assess it dispassionately to set out our views on how we should consider these markets in future. We need more research and debate from academics as well as practitioners, to help us think through the challenges ahead.
If we can help consumers derive the undoubted benefits from the use of consumer data and online reviews, without suffering the potential downsides we have identified, then the economic rewards will amply justify our collective effort.
Footnotes
- My thanks to CMA colleagues who collaborated with me on this speech, and in particular Helen Fleming, Jon Riley, Tony Curzon-Price and Mike Walker.
- Centre for Retail Research, 2013.
- International Communications Market Report (ICMR), Ofcom, 2014.
- Office of National Statistics, 2015.
- ‘Big Data’s infinite harvest’, Edward Luce, Financial Times, 24 May 2015.
- Information Commissioner’s Office (ICO), Annual Track, September 2014.
- The Data Dialogue, Demos, September 2012.