From bd45531e0a3a100a1688ca64b616b7b927209073 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Bart=C5=82omiej=20Roma=C5=84ski?= Date: Tue, 12 May 2020 22:57:37 +0200 Subject: [PATCH] Add 'One-To-One Personalization' use case --- personalization.md | 35 ++++++++++++++++++++++++++++ support_for_advertising_use_cases.md | 5 ++++ 2 files changed, 40 insertions(+) create mode 100644 personalization.md diff --git a/personalization.md b/personalization.md new file mode 100644 index 0000000..c161f68 --- /dev/null +++ b/personalization.md @@ -0,0 +1,35 @@ +# One-to-One Personalization + +In the area of marketing, personalization typically means tailoring the ad content in order to meet the individual needs of each user. + +As shown by many well-known examples in various sectors (web search, social/video platforms, online retail marketplaces, mobile app stores…), it can be a game-changer for both: user experience and effectiveness of the marketing activities. + +While user lists are not enough to satisfy this use case, there are other technical solutions that allow 1:1 personalization without affecting the users’ privacy. + +## Personalization in RTB + +Modern process of ad delivery is typically led by three decisions: + +* **Targeting** – shall we consider this ad/campaign? +* **Bid valuation** – how much are we willing to pay for this? +* **Personalization** – what exact content shall we present in the ad? + +Current technical solutions allow us to answer all three questions with a perfect granularity – a single impression and a single user. + +However, the technology behind raises privacy concerns. To minimize them, one possible solution is to somehow group users instead of identifying them 1:1 – a concept called "users lists". Doing so for targeting and bid valuation wouldn’t necessarily hit hard effectiveness. Nor the user experience. Thus, this probably is one of the good, future directions. + +But, the situation with personalization is very much different – preserving the possibility of one-to-one personalization is crucial for delivering to each individual user the ad content that will be useful and valuable for him or her. + +## User’s perspective + +If I’m really into fantasy books, I may end up on a user list 'fantasy-lovers'. And that’s fine for targeting, and that’s OK for bid valuation. But that’s not enough for personalization. A broadly targeted static ad showing me best selling books in the category is useless for me and just annoying. Only the ad showing me recommendations carefully tailored towards my individual taste, my reading history, my past reviews etc... makes sense to me. + +The problem scales inversely with the size of the advertiser. Largest advertisers might have a big enough user base to be able to somewhat accurately personalize based on user lists (assuming reasonable size is allowed). However, the long tail of not-the-largest advertisers will be in a lost position – a campaign for local bookshop will be performing significantly worse than similar one from a large retailer. The only fair solution is 1:1 personalization. + +The imprecise personalization negatively affects user satisfaction, ad effectiveness from the advertiser's perspective, and as a consequence publishers' revenue stream. The whole ecosystem is hurt. + +## Privacy + +One-to-one personalization within marketing is widely accepted by users. The main concern is that current technical solutions allow to leak information to many unrelated parties and this can be abused. This is not an issue in case of big platforms where a single domain serves as publisher, advertiser and a data source – all at once. So, the only question left is how this can be addressed in a more scattered ecosystem of a, so called, open internet. + +Recent ideas can help with that – for instance, the concept of on-device auctions introduced in the [TURTLEDOVE](https://github.com/michaelkleber/turtledove). While the user lists cannot be used in this case, by slightly tweaking the proposal, it should be possible to guarantee user privacy by keeping the user data on the device, while still allowing 1:1 personalization. diff --git a/support_for_advertising_use_cases.md b/support_for_advertising_use_cases.md index 9509e19..c747081 100644 --- a/support_for_advertising_use_cases.md +++ b/support_for_advertising_use_cases.md @@ -72,6 +72,7 @@ These are use-cases that only apply to a (possibly large) subset of advertisers. - [Dynamic Ads](#dynamic-ads) - [Email Marketing](#email-marketing) - [Real time spend management](#real-time-spend-management) +- [One-to-One Personalization](personalization.md) | Use-case | Chrome | Safari | Community Proposals | @@ -85,6 +86,7 @@ These are use-cases that only apply to a (possibly large) subset of advertisers. | [Collaborative Ads](#collaborative-ads) / [Dynamic Ads](#dynamic-ads) | Some discussion in this [GitHub issue](https://github.com/WICG/conversion-measurement-api/issues/32). No solutions or even strong acknowledgement of the importance of this use-case yet. | Some discussion in this [GitHub issue](https://github.com/WICG/ad-click-attribution/issues/36). Good collaborative problem solving going on. No firm solution yet. | Facebook proposal for “[Conversion Filters](https://github.com/w3c/web-advertising/blob/master/conversion-filters.md)” | | [Email Marketing](#email-marketing) | unclear | unclear | | | [Real time spend management](#real-time-spend-management) | High latency in TURTLEDOVE as the javascript updates have unknown frequency, and reporting is delayed in the [Aggregate Reporting API](https://github.com/csharrison/aggregate-reporting-api#advanced-example-calibrating-a-frequency-capping-model) | Supported as ads are not done in opaque iframe and bidding is not done remotely | | +| [One-to-One Personalization](personalization.md) | Unclear. For instance, the [TURTLEDOVE](https://github.com/michaelkleber/turtledove) proposal offers the concepts of "interest groups" with unspecified granularity. | No support | | @@ -426,3 +428,6 @@ Some actors on the internet are able to build custom audiences, revealing strong Relationship/CRM Marketing: a marketing strategy designed to foster customer loyalty, interaction, and long-term relationships. It is designed to develop or reinforce a strong bond with existing customers by continuing a dialogue with them that provides the customer with information that matches their needs, problems, or interests +# One-to-One Personalization + +In marketing, personalization typically means tailoring the ad content in order to meet the individual needs of each user. It can be a game-changer for both: user experience and effectiveness of the marketing activities. The full page dedicated to one-to-one personalization can be found [here](personalization.md).