GDPR consent design: how granular must adtech opt-ins be?

Dr Johnny Ryan GDPR Leave a Comment

This note examines the range of distinct adtech data processing purposes that will require opt-in under the GDPR.[1]

In late 2017 the Article 29 Working Party cautioned that “data subjects should be free to choose which purpose they accept, rather than having to consent to a bundle of processing purposes”.[2] Consent requests for multiple purposes should “allow users to give specific consent for specific purposes”.[3]  Rather than conflate several purposes for processing, Europe’s regulators caution that “the solution to comply with the conditions for valid consent lies in granularity, i.e. the separation of these purposes and obtaining consent for each purpose”.[4] This draws upon GDPR, Recital 32.[5]

In short, consent requests must be granular, showing opt-ins for each distinct purpose.

How granular must consent opt-ins be?

In its 2013 opinion on “purpose limitation”, the Article 29 Working Party went some way toward defining the scope of a single purpose: a purpose must be “sufficiently defined to enable the implementation of any necessary data protection safeguards,” and must be “sufficiently unambiguous and clearly expressed.”[6]

The test is “If a purpose is sufficiently specific and clear, individuals will know what to expect: the way data are processed will be predictable.”[7] The objective is to prevent “unanticipated use of personal data by the controller or by third parties and in loss of data subject control [of these personal data]”.[8]

In short, a purpose must be specific, transparent and predictable.[9] It must be describable to the extent that the processing undertaken for it would not surprise the person who gave consent for it.

The process of showing an ad to a single person (in online behavioral advertising) involves the processing of personal data for several distinct purposes, by hundreds of different companies.

Therefore, a broad, all-encompassing sentence such as “to show you relevant advertising” does not make it possible for one to grasp how one’s data will be used by a large number of companies. It would not be possible to understand from this sentence, for example, that inferences about one’s characteristics would be inferred, or what types of consequences may result.

The following table shows an indicative list of ten purposes for which personal data are currently processed in the online behavioral advertising system. In practice, there may be more purposes at play. The table also generalizes the types of company involved in the selection and display of an ad.

Download high resolution PDF

(Refer to footnote 10 to for a discussion the challenges presented by these purposes for all businesses involved.[10])

Pre-consent naming of each controller, and granular post-consent controller consent withdrawal

Recital 42 of the GDPR notes that “For consent to be informed, the data subject should be aware at least of the identity of the controller and the purposes of the processing”.[11] All controllers (including “joint controllers” that “jointly determine the purposes and means of processing”[12]) must be named.[13]

Each purpose must be very clear, and each opt-in requires a “clear affirmative action” that is both “specific”, and “unambiguous”.[14] There can be no pre-ticked boxes,[15] and “consent based on silence” is not permitted.[16]

Therefore, a consent request should be made with granular options for each of these purposes, and the names each controller that processes personal data for each of these purposes. For example:  

Specific purpose 1 | controllers A, B, C | options: Accept / Refuse 

There are two different scenarios for how consent for these purposes will be presented: the best case, and the more likely worst case.

The best scenario

At a minimum, then, assuming that all websites, SSPs, Ad Exchanges, DSPs, DMPs, and advertisers could align to pursue only these purposes, a consent request for this would include granular opt-in controls for a wide range of diverse purposes, the categories of processor pursuing each, and a very long list of controller names pursuing each.

The language and presentation of the request must be simple and clear, ideally the result of user testing.[17]

A consent request for a single purpose, on behalf of many controllers, might look like this.

Specific processing purpose consent, for multiple controllers,
with “next” button for multiple processing purpose opt-ins

Download wireframes

What is presented when?

The Article 29 Working Party suggests that consent notices should have layers of information so that they do not overload viewers with information, but make necessary details easily available.[18] This is adopted in the design above using “View details”, “Learn about your data rights here”, and similar buttons and links.

When a user clicks “view details” to see the next layer of information about a controller

Download wireframes

While some details, such as contact details for a company’s data protection officer, can be placed in a secondary layer, the primary layer must include “all basic details of the controller and the data processing activities envisaged”.[19]

Elements presented in this layer

Download wireframes

The likely scenario:

The scenario above assumes that all businesses in online behavioral advertising can agree to pursue tightly defined purposes without deviation. However, it is more likely that controllers will need granular opt-ins, because their purposes are unique.

Any individual controllers who intend to process data for their own unique purposes will need further granular opt-ins for these purposes. Since adtech companies tend to deviate from the common purposes outlined above, it is likely that most or all of them would ultimately require granular purpose consent for each controller.

However, even if all controllers pursued an identical set of purposes so that they could all receive consent via a single consent dialogue that contained a series of opt-ins, there would need to be a granular set of consent withdrawal controls that covered every single controller once consent had been given. The GDPR says that “the data subject may exercise his or her rights under this Regulation in respect of and against each of the controllers”.[20]

A higher bar: “explicit consent”

Processing of personal data in online behavioral advertising (for example, purposes 2, 3, 5, 8, and 10 in the table above) is highly likely to produce special categories of data by inference.[21] Where this occurs, these purposes require “explicit” consent.[22]

Special categories of data reveal “racial or ethnic origin, political opinions, religious or philosophical beliefs, or trade union membership, … [and] data concerning health or data concerning a natural person’s sex life or sexual orientation”.[23] 
To make consent explicit requires more confirmation. For example, the Article 29 Working Party suggests that two-stage verification is a suitable means of obtaining explicit consent.[24] One possible approach to this is suggested in PageFair’s design below.

Suggested mechanism for “explicit consent” 

Download wireframes

One can confirm one’s opt-in in a second movement of the finger, or cursor and click. It is unlikely that a person could confirm using this interface unless it was their intention.  

Download wireframes

Note that even this high bar, however, may not be permitted in some Member States. The GDPR gives European Member States the latitude to enact national legislation that prohibits consent as a legal basis for processing of special categories of data.[25] Therefore, it may not be legal to process any special categories of personal data in some EU Member States.

Conclusion 

Consent for website and app publishers is certainly an important objective, but the personal data it provides must only be processed after data leakage has been stopped. Data leakage (through in RTB bid requests, cookie syncs, JavaScript ad units, and mobile SDKs) exposes publishers as the most obviously culpable parties that regulators and privacy NGOs can target. At the same time, it also exposes their adtech vendors, and advertisers, to large fines and legal actions too.[26]

Websites, apps, and adtech vendors, should switch from using personal data to monetize direct and RTB advertising to “non-personal data”.[27] Using non-personal, rather than personal, data neutralizes the risks of the GDPR for advertisers, publishers, and adtech vendors. And it enables them to address the majority (80%-97%) of the audience that will not give consent for 3rd party tracking across the web.[28]

We recently revealed PageFair Perimeter, a regulatory firewall that blocks party data leakage, and enables publishers and adtech partners to use non-personal data for direct and RTB monetization when consent is absent (and leverage personal data when adequate consent has been given). You can learn more about Perimeter here. Publishers using Perimeter do not need people’s personal data (nor the consent required to process it) to monetize websites and apps.

Learn about Perimeter

Postscript

A hiccup in the choreography of the European Commission’s legislative proposals means that non-tracking cookies will need storage consent, at least until the application of the forthcoming ePrivacy Regulation. These cookies, however, contain no personal data, and obtaining consent for their storage is significantly less burdensome than obtaining consent for to process personal data for multiple purposes and multiple controllers. Update: 16 January 2018: See PageFair Insider note on storage consent for non-tracking cookies.

Perimeter: the regulatory firewall for online media and adtech.

Feature-rich adtech, even without personal data. Control user data and 3rd parties in websites + apps. Get robust consent.

Learn more

Notes:

[1] See our discussion of why consent is the appropriate legal basis for online behavioral advertising in “Why the GDPR ‘legitimate interest’ provision will not save you” , PageFair Insider, 13 March 2017 (URL: https://pagefair.com/blog/2017/gdpr-legitimate-interest/).

[2] “Guidelines on consent under Regulation 2016/679”, Article 29 Working Party, 28 November 2017, p. 11.

[3] ibid., p. 13.

[4] ibid., p. 11.

[5] Regulation (EU) 2016/679 of The European Parliament and of The Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation), Recital 32. “…Consent should cover all processing activities carried out for the same purpose or purposes. When the processing has multiple purposes, consent should be given for all of them. …”

[6] “Opinion 03/2013 on purpose limitation”, Article 29 Working Party, 2 April 2013, p. 12.
Curiously, the Spanish Data Protection Authority has issued guidance that contains a sentence suggesting that continuing to browse a website might constitute consent, which is at odds with the Article 29 Working Party guidance on consent and appears to be entirely at odds with the text of the Regulation. See “Guía del Reglamento General de Protección de Datos para responsables de tratamiento”, Agencia Española de Protección de Datos, November 2017, p. 6.

[7] “Opinion 03/2013 on purpose limitation”, Article 29 Working Party, 2 April 2013, p. 13.

[8] “Guidelines on consent under Regulation 2016/679”, Article 29 Working Party, 28 November 2017, p. 12.

[9] “Opinion 03/2013 on purpose limitation”, Article 29 Working Party, 2 April 2013, p. 13.

[10] None of these purposes would be permissible unless data leakage were first addressed. See “Consent to use personal data has no value unless one prevents all data leakage”, PageFair Insider, October 2017 (URL: https://pagefair.com/blog/2017/understanding-data-leakage/). Furthermore,

  • Purpose 3 could not be permissible in any situation.
  • Purposes 2, 3, 5, 8, and 10 are highly likely to produce special categories of data by inference. See discussion of “explicit consent” in this note.
  • Regarding the purposes for which data have been sold, and to what category of customer, see “Data brokers: a call for transparency and accountability”, Federal Trade Commission, May 2014, pp 39-40, and B3-B

[11] The GDPR, Recital 42.

[12] The GDPR, Article 26, paragraph 1.

[13] “Guidelines on consent under Regulation 2016/679”, Article 29 Working Party, 28 November 2017, p. 14.

[14] The GDPR, Article 4, paragraph 11.

[15] ibid., Recital 32.

[16] “Guidelines on consent under Regulation 2016/679”, Article 29 Working Party, 28 November 2017, p. 16.

[17] “Guidelines on transparency under Regulation 216/679” Article 29 Working Party, November 2017, pp 8, 13.

[18] “Guidelines on consent under Regulation 2016/679”, Article 29 Working Party, 28 November 2017, p. 14.

[19] ibid., p. 15.

[20] The GDPR, Article 26, paragraph 3.

[21] “Informing data subjects is particularly important in the case of inferences about sensitive preferences and characteristics. The controller should make the data subject aware that not only do they process (non-special category) personal data collected from the data subject or other sources but also that they derive from such data other (and special) categories of personal data relating to them.” See “Guidelines on Automated individual decision-making and Profiling for the purposes of Regulation 2016/679”, Article 29 Working Party, 3 October 2017, p. 22.

[22] The GDPR, Article 9, paragraph 2, a.

[23] ibid., Article 9, paragraph 1.

[24] “Guidelines on consent under Regulation 2016/679”, Article 29 Working Party, 28 November 2017, p. 19.

[25] The GDPR, Article 9, paragraph 2, a.

[26] See “Consent to use personal data has no value unless one prevents all data leakage”, PageFair Insider, October 2017 (URL: https://pagefair.com/blog/2017/understanding-data-leakage/).

[27] Non-personal data are any data that can not be related to an identifiable person. As Recital 26 of the GDPR observes, “the principles of data protection should therefore not apply to anonymous information, namely information which does not relate to an identified or identifiable natural person or to personal data rendered anonymous in such a manner that the data subject is not or no longer identifiable”. This recital reflects the finding of the European Court of Justice in 2016 that data are not personal “if the identification of the data subject was prohibited by law or practically impossible on account of the fact that it requires a disproportionate effort in terms of time, cost and manpower, so that the risk of identification appears in reality to be insignificant”. Judgment of the Court (Second Chamber) Patrick Breyer v Bundesrepublik Deutschland, Case C-582/14, 19 October 2016.

Non-tracking cookies, which contain no personal data, are useful for privacy-friendly advertising, and for other functions where an individual does not need to be identified such as A/B testing.

[28] See “Research result: what percentage will consent to tracking for advertising?”, PageFair Insider, September 2017 (URL: https://pagefair.com/blog/2017/new-research-how-many-consent-to-tracking/).

The granularity of consent required for online behavioral advertising will make the consenting audience even smaller. Moreover, consent for adtech will not only be hard to get, it will also be easy to lose. Consent can be withdrawn with the same degree of ease as it was given, under The GDPR, Article 7, paragraph 3.

The Article 29 Working Party demonstrates what this means in practice: “When consent is obtained … through only one mouse-click, swipe, or keystroke, data subjects must … be able to withdraw that consent equally as easily”.

“Guidelines on consent under Regulation 2016/679”, Article 29 Working Party, 28 November 2017, p. 21.

The guidance also says that “Where consent is obtained through use of a service specific user interface (for example, via a website, an app, a log-on account, the interface of a IoT device or by e-mail), there is no doubt a data subject must be able to withdraw consent via the same electronic interface, as switching to another interface for the solve reason of withdrawing consent would require undue effort”.