Meta ads in Germany and Austria: a smaller signal, stricter rules, and how an Irish shop plans for both

The point: same ad account, different rules, smaller signal

An Irish shop can run Meta ads into Germany and Austria from the same ad account it uses at home. The interface is identical. Pick a country, write an ad, set a budget. The prices are even in the same currency. That sameness is the trap.

Three things shape the result, and none of them is in the interface. Fewer people use social media in Germany than almost anywhere else in the EU. The law in Germany and Austria says the Meta pixel waits for a yes, so Meta sees only part of what your ads sell. And since the EU's Digital Markets Act, Meta itself has to offer people a version of Facebook and Instagram that uses less of their data, which changes what targeting can do. The last two will sound familiar to an Irish shop, because Ireland is in the EU and both apply at home too. The first one does not, and together the three mean less for Meta to work with than the dashboard suggests.

Put together, they mean a smaller signal for Meta's machine to learn from. Most of what goes wrong with Irish campaigns in the German speaking markets follows from that, not from the colour of the button. The old version of this article blamed tone first. I still think tone matters, and I come to it, but the plumbing and the arithmetic come first, because they decide whether a good ad ever gets a fair test.

I work through all of it with one invented shop. Marble City Tea is a loose leaf tea shop in Kilkenny, run by Clodagh, selling at marblecitytea.example. It ships to Germany and Austria and pays for its ads in euro, the same currency its German and Austrian customers pay in, so there is no exchange rate anywhere in the sums below. Every business number about Marble City is made up. The law, the platform rules and the statistics are not.

How many people you can reach, from official numbers

Start with the ceiling. In 2025, 59% of people aged 16 to 74 in Germany used social media for private purposes, according to the Federal Statistical Office. Only Italy, at 56%, was lower in the EU. The EU average was 67%.¹ Use falls steeply with age: 85% of 16 to 24 year olds, 25% of 65 to 74 year olds.¹

Austria sits higher. Statistics Austria's 2026 household survey counts 6,813,300 people aged 16 to 74, of whom 68.6% used social networks in the three months before they were asked.² Eurostat's table, which also carries Switzerland, puts 2025 at 59% for Germany, 68% for Austria and 74% for Switzerland.³

Each bar is the share of people aged 16 to 74 who used social networks in 2025, from Eurostat's internet activities table.³ Germany is the biggest German speaking market by a distance, and it is also the one where the smallest share of people is on social media at all. A plan that assumes Germany behaves like the EU average overestimates it before a single ad runs.

The way to turn a share into a ceiling is to multiply it by the population it describes:

$$A = P \times s$$

Here $A$ is the number of people who use social networks, $P$ is the population the survey covers and $s$ is the share of them who use social networks, written as a decimal. The groundwork for turning a percentage into a decimal and back is in Foundations: percentages.

Worked example: Marble City's ceiling in Austria

Clodagh wants to know how many people in Austria she could reach at most. The survey population is $P = 6{,}813{,}300$ and the share is $s = 0.686$,² so $A = 6{,}813{,}300 \times 0.686 = 4{,}673{,}924$.

So about 4.67 million Austrians aged 16 to 74 use social networks. That is every platform together, not Facebook and Instagram alone, so the real Meta ceiling is lower. Clodagh's actual audience, loose leaf tea drinkers who order online from an Irish shop, is a small slice of it.

Now compare that with the number Ads Manager shows her. Meta says its estimated audience size counts Meta accounts, not people, is not designed to match population or census figures, and should not be read as the number of accounts who will see the ads.⁴ So when Ads Manager's estimate and the official statistics disagree, they are counting different things. I use the official number to sanity check the plan, and Meta's estimate only to compare one targeting choice with another inside the account.

Consent decides what your pixel sees

The Meta pixel is a piece of JavaScript from Meta. Meta's own setup code loads it from Meta's servers, from a file called fbevents.js.⁵ Once loaded, it reads and sets identifiers in the browser and sends events back to Meta.

In Germany that is governed by section 25 of the TDDDG. Storing information on the visitor's device, or reading information already there, is only allowed if the visitor has consented on the basis of clear and comprehensive information. The only exceptions are transmitting a message and what is strictly necessary for a service the visitor explicitly asked for.⁶ In Austria, section 165(3) of the TKG 2021 says the same: data may only be collected once the user has actively consented, with the same two exceptions.⁷ Advertising is neither transmission nor strictly necessary. So the pixel waits for a yes in both countries. The full legal picture, with what a valid banner looks like, is in cookies and consent in Germany and Austria.

None of this should be news to an Irish shop. At home, regulation 5 of the ePrivacy Regulations 2011 asks for the same yes, and the Data Protection Commission (DPC) says cookies related to advertising are not strictly necessary.⁸⁹ All three countries carry the same EU rule into their own law, so a banner built properly for Ireland is built on the right logic for Germany and Austria. The work is in checking that it really runs on the German and Austrian pages, in German, before the pixel does.

There is a second reason to care. In Fashion ID, a German online clothing retailer had embedded a Facebook button that sent visitors' data to Facebook. The Court of Justice held that the shop could be a controller jointly with Facebook for collecting and transmitting that data, and that where consent is the basis, the shop must obtain it for that collection and transmission.¹⁰ The pixel on your checkout is your responsibility, not Meta's.

Meta gives you a switch for this. Its documentation describes two calls, fbq('consent', 'revoke') to pause sending pixel events and fbq('consent', 'grant') to send them once consent is given, and says revoke must be called on every page.¹¹ That pauses the events. It does not stop the script loading. For German and Austrian visitors I go one step further and do not load the pixel at all until the visitor presses accept. That is the same line I take with Google's tags, where basic consent mode keeps the tags from loading until the visitor interacts with the banner and sends nothing at all on a refusal.¹² Basic mode is my default in Germany and Austria, and the reasoning is in event tracking and consent mode.

Worked example: what Meta sees of Marble City's October

In October, Marble City's Meta ads send 9,000 sessions to its German pages. Clodagh's shop database matches orders to the landing session and counts 180 orders from those sessions. For the example I assume 45% of German visitors press accept on the banner.

Meta can only see an order if the pixel was allowed to load:

$$C_{\text{obs}} = C \times r$$

$C_{\text{obs}}$ is the number of purchases Meta observes, $C$ is the true number of purchases from the ad clicks and $r$ is the consent rate, the share of visitors who accepted. It is a rate applied to a count, so the groundwork is Foundations: ratios and rates.

For Germany in October that gives $C_{\text{obs}} = 180 \times 0.45 = 81$.

Clodagh spent €2,400 on German ads in October. Her true cost per purchase is €2,400 divided by 180, which is €13.33. The cost per purchase Ads Manager reports is €2,400 divided by 81, which is €29.63. Same ads, same customers, more than twice the price on the dashboard. This assumes refusers buy at the same rate as accepters, which is generous to Meta. How to correct for the difference is worked through in the consent article linked above.

The Conversions API does not get around the banner

Someone regularly tells shop owners that the Conversions API fixes the consent gap. It does not.

The Conversions API sends events from your server to Meta instead of from the visitor's browser. Meta's documentation says server events are processed like pixel events and may be used in measurement, reporting and optimisation in the same way.¹³ Meta recommends running both, the pixel and the server, side by side, and asks you to give each event a shared event_id so it can throw away the duplicate. It discards a matching pair received within 48 hours.¹⁴

That is plumbing, and good plumbing. A server event is more reliable than a browser event that an ad blocker or a closed tab can kill. But the law asks whether the person agreed to their data going to an advertising platform. Moving the sending from the browser to your server does not change the answer. So in my setups the server forwards a purchase to Meta only when the order carries a recorded consent for advertising. An order from a visitor who refused is counted in Clodagh's database and in her own reports, renamed out of anything that feeds an audience or an optimisation signal, and never sent to Meta with an identifier.

Read it top to bottom. The first two boxes happen before the visitor has done anything, and they are where most setups I see fail in Germany: the tag manager loads the pixel with the page, and the banner is decoration. The accept branch is the normal, redundant setup Meta recommends. The reject branch is where the discipline lives. The plumbing for the server side is in my technical guide to server side tracking.

The EU changed Meta itself, and that changes targeting

Even with a perfect banner, a slice of German and Austrian Facebook and Instagram users are now out of reach of personalised targeting. That is the Digital Markets Act at work, and it is at work in Ireland too.

In November 2023, Meta started offering people in the EU, the EEA and Switzerland a choice: use Facebook and Instagram free with personalised ads, or pay a monthly subscription for no ads, at €9.99 a month on the web or €12.99 on iOS and Android. Meta says subscribers' information is not used for ads.¹⁵ In November 2024 it cut the price to €5.99 on the web and €7.99 on mobile, and added a third option for people who keep the free version: less personalised ads. Meta says those ads are based only on context, meaning what a person sees in that session, and a minimal set of data points: age, location, gender and how the person engages with ads. Meta also said some ads in that experience would be unskippable ad breaks of a few seconds, and that the November 2024 changes apply in the EU, the EEA and Switzerland.¹⁵

So your customers in Cork and Galway have been shown the same choice as your customers in Munich and Graz. Whatever it has done to your targeting at home, it does in Germany and Austria as well. The difference is only that you may not have looked at your Irish campaigns with this in mind yet.

The European Commission looked at the original choice and did not accept it. On 23 April 2025 it found that Meta's "consent or pay" model breached the DMA, because users did not get a less personalised but otherwise equivalent service, and fined Meta €200 million. The decision covered March to November 2024, and the Commission said it was still assessing the November 2024 version.¹⁶ On 8 December 2025 the Commission acknowledged Meta's undertaking to give EU users an effective choice between sharing all their data for fully personalised ads and sharing less data for more limited personalisation, presented to users from January 2026. The Commission said it would then seek evidence on the uptake and impact.¹⁷

What it means for an advertiser, in plain terms:

1. People who pay for no ads see none of yours. They are not an audience, at any price.
2. People on less personalised ads still see ads, but by Meta's description those ads are chosen from context, age, location, gender and engagement with ads.¹⁵ My reading is that interest targeting and your website audiences have little or nothing to work with for those people, because Meta has said what data the choice uses, and your pixel data is not on the list.
3. Location still works for everyone who sees ads, because it is on that list. So does a good ad, because context is on it too.
4. I found no official figure for how many EU users have chosen each option. The Commission has said it will gather that evidence.¹⁷ Until it is public, anyone quoting a precise share is guessing.

Worked example: Marble City plans for a smaller personalised pool

Clodagh has been running an interest audience in Germany built on tea, organic food and cooking. She cannot see how many people in it chose less personalised ads, and neither can I. So she plans for it rather than measuring it.

She keeps one broad ad set, targeted only by country and language, with the ad doing the selecting: a clear picture of the tea, a German headline that says what it is, and the price. Next to it she runs her interest ad set for four weeks. At the end she compares the two on cost per purchase from her own database, not on Meta's figure, because the consent gap hits both equally and the database counts what really happened. If the broad ad set is within 10% of the interest one, she keeps the broad one, because it does not depend on data that a growing share of users may withhold. The 10% is my rule of thumb, not a law of nature.

Budget for the learning phase in a smaller signal

Meta's delivery system has a learning phase. Meta says ad sets usually exit it after about 50 results in the week after the last significant edit, and that during learning, performance is less stable and the cost per result is usually higher.¹⁸ A result is whatever the ad set optimises for: a purchase, an add to cart, a click.

Fifty purchases a week per ad set is a budget question, and anywhere the pixel waits for consent it is a consent question as well, because Meta counts only the results it sees. That holds in Ireland as much as in Germany and Austria, so the arithmetic will not be new to you. What changes abroad is the consent rate you put into it: use the rate you measure on your German pages, not the one from your Irish ones. The weekly budget an ad set needs to see 50 results is roughly:

$$B_{w} = n \times \frac{\text{CPA}}{r}$$

$B_{w}$ is the weekly budget, $n$ is the number of results Meta wants, about 50, $\text{CPA}$ is the true cost per result from your own database and $r$ is the consent rate. Dividing by $r$ turns the true cost into the cost Meta observes. It is the same rate arithmetic as before, so Foundations: ratios and rates covers it, and why the average cost can mislead at the margin is in Foundations: marginal versus average.

Worked example: can Marble City afford to optimise for purchases?

Clodagh's budget for Germany and Austria together is €3,000 a month, about €692 a week. Her true cost per purchase in Germany was €13.33 and the consent rate is 0.45, so $B_{w} = 50 \times \frac{13.33}{0.45} = 1{,}481$.

Optimising for purchases needs about €1,481 a week per ad set, more than twice what she has. The ad set would sit in "learning limited" and never settle.

Two changes fix it. First, Germany and Austria go into one ad set, because both are German speaking and Meta lets one ad set target several countries.¹⁹ Splitting a small budget by country halves the signal for nothing. Second, she optimises for add to cart, which happens far more often. Her database says a German add to cart costs her €3.00 in true terms, so $B_{w} = 50 \times \frac{3.00}{0.45} = 333$.

That is €333 a week, inside her €692. The ad set can learn. Purchases still happen and still count in her database. Meta simply learns from a more frequent signal.

The bars are weekly budgets in euro. The first two are what one ad set needs for Meta to observe about 50 results at a 45% consent rate, the third is what Clodagh has. Only the add to cart bar sits under the budget. The takeaway: in a market where Meta sees less than half your sales, pick an optimisation event your budget can feed, and keep the true sales count in your own system.

One language per ad set, and know where German stops

Location and language are separate settings in Meta. Location is the country. Language is a further filter: Meta's documentation describes locale targeting as reaching accounts with a language other than the common language for a location, and gives German as an example.²⁰ Meta also offers multi language ads, where one ad carries versions in several languages and Meta shows the matching one to each viewer.²¹

For Germany and Austria, one German ad set covers both. Switzerland is different. The Federal Statistical Office counts German as a main language for 61% of the population in 2024, French for 23% and Italian for 8%.²² A German ad aimed at all of Switzerland misses about a third of it. If Clodagh ever adds Switzerland, it gets a German version and a French version at the least, and its own banner rules: Swiss law asks for information and a way to refuse, with explicit consent required only where the processing is unexpected, high risk or involves sensitive data, according to the Swiss data protection commissioner.²³ Switzerland is also outside the EU and the euro area, with prices in francs and its own customs and VAT questions, which belong in the cross border setup for Germany.

Worked example: Marble City's ad set map

Clodagh writes down her structure before building it:

1. Ad set one: Germany and Austria, German language, optimised for add to cart, broad targeting. Ads in German, landing on the German shop pages.
2. Ad set two: Germany and Austria, German language, the interest audience, for the four week comparison only.
3. Ad set three: retargeting, consented website visitors from Germany and Austria in the last 30 days. More on that below.
4. Switzerland: not yet. Written on the list so it is a decision, not an oversight.

Three ad sets, one language, one budget. The structure I most often find in accounts that come to me from Ireland and the UK is the opposite: an ad set per country, the same English ad in each, none of them past the learning phase.

Write the ad in German, and leave the fake urgency at home

Now the tone, which the old version of this article put first. I still stand by its core: in my experience German and Austrian buyers respond better to an ad that says plainly what the product is, what it costs and why it is good, than to an ad that only performs energy. I have no study to quote on that, and the figures the old version quoted could not be traced to a source, so I have taken them out. What follows is my practice, and the law where the law has something to say. The longer argument on tone is in why your home brand tone can undermine trust in DACH.

Write in German from scratch, with a native speaker, rather than translate. A translated ad reads as translated, and readers notice.

Say the concrete thing. The tea, the origin, the grams, the price in euro with delivery, how long delivery to Germany takes. A German buyer is about to order from a shop abroad, and the questions in their head are about delivery, returns and payment, so answer them before they ask. The payment side is its own subject in the article on DACH payments in Solidus.

Leave fake urgency out. This is not taste, it is law. German unfair competition law lists, among practices that are always unfair, the false statement that goods are available only for a very limited time in order to push the consumer into an immediate decision.²⁴ The rule comes from the EU's Unfair Commercial Practices Directive, so it applies in Austria too, and in Ireland.²⁵ "Only today" is fine when it is true. When it is a habit, it is a breach.

Be careful with badges. The old version recommended ISO or TÜV badges as trust cues. The same German list forbids using a quality mark without the required authorisation, and a sustainability seal that rests on neither a certification scheme nor a state scheme.²⁴ Show a certification you actually hold. Never one you would like to hold.

Worked example: rewriting Marble City's ad

The Irish ad reads: "A proper cuppa, no bother. Last chance for 20% off." It works in Kilkenny, where the tone lands and the offer is real.

The German ad Clodagh's translator writes, rendered back into English for this article: "Loose leaf Assam from a family estate, 250g." Then the price in euro with delivery included, and: "Delivered to Germany and Austria in four to six working days. Free returns within 14 days." The headline names the tea. The offer line appears only on the days the offer actually runs, with the end date stated. The image shows the tea, the tin and the scale of the bag. No exclamation marks, no emoji, nothing to decode.

The test is the same as for the targeting: run the new ad against the translated Irish one for four weeks, judge it on purchases in Clodagh's own database, keep the winner.

Retargeting reaches only the people who said yes

Retargeting means showing ads to people who visited your site. A website custom audience in Meta is built from pixel events, and you set how long someone stays in it. Meta's API takes a retention between one and 180 days.²⁶ The old version of this article recommended 14 or 21 days for German B2B buyers. For a shop I start at 30 days and adjust from the data.

The consent rule shrinks this audience twice. Only visitors who accepted ever enter it, because the pixel never loaded for the others. And by Meta's description, people who chose less personalised ads or no ads are not shown ads based on your website data.¹⁵ So the retargeting pool in Germany is a fraction of your visitors, and a small pool is easy to overwhelm.

The number to watch is frequency, the average number of times each person saw the ad:

$$f = \frac{I}{R}$$

$f$ is the frequency, $I$ is the number of impressions and $R$ is the reach, the number of distinct people who saw the ad. It is one count divided by another, which is Foundations: ratios and rates again.

Worked example: Marble City's retargeting budget

In October, 7,500 different visitors reached Marble City's German pages. At a 45% consent rate, $7{,}500 \times 0.45 = 3{,}375$ of them can enter the 30 day audience. Say Meta manages to reach 2,500 of them in a month.

Clodagh first sets €300 a month for retargeting. At a cost of €8 per 1,000 impressions, a figure I have made up for the example, that buys €300 divided by €8, times 1,000, which is 37,500 impressions, so $f = \frac{37{,}500}{2{,}500} = 15$.

Fifteen times a month, every month, to the same people. That stops being a reminder and starts being a nuisance. At €100 a month, the same arithmetic gives 12,500 impressions and a frequency of five. Clodagh takes €100 for retargeting and moves the other €200 to the broad ad set, where it can find new people. Retargeting reminds. It never introduces, and it should never nag.

What I would set up for an Irish shop, in order

If Marble City came to me, this would be the order of work, and it is the same for any Irish shop taking Meta ads into Germany and Austria.

First, the banner and the pixel: two equal buttons on the first layer, the pixel not loaded until accept, Google's tags in basic consent mode. Then test it in a private window with the network tab open, on the German pages, without touching the banner, and confirm nothing goes to Meta.

Second, the server events: the Conversions API with a shared event_id, sending only consented orders, with refused orders kept and renamed in the shop's own system.

Third, the structure: one German ad set for both countries, an optimisation event the budget can feed, broad targeting carried by the ad, a small retargeting set with a calculated frequency.

Fourth, the ads: written in German, concrete, priced, no fake urgency, no borrowed badges.

Fifth, the measurement: every month, Meta's figure next to the database count, per market, so nobody cuts Germany because Meta sees less than half of it.

None of this is exotic. It is the discipline you use at home, adjusted for a market where fewer people are on social media to begin with and the platform sees less of them. If you are planning the move into those markets more broadly, the mistakes I see brands make on the way into DACH is the wider map.

Sources

Footnotes

1. Statistisches Bundesamt (Destatis), 59 % der Bevölkerung zwischen 16 und 74 Jahren in Deutschland nutzen aktiv soziale Medien, press release, March 2026, https://www.destatis.de/DE/Presse/Pressemitteilungen/2026/03/PD26_N014_63.html

2. Statistik Austria, IKT Einsatz in Haushalten 2026, tables for people aged 16 to 74 (table T04, online activities), 2026, https://www.statistik.at/fileadmin/pages/284/IKTH2026_TABELLEN_EXTERN_16_74.ods

3. Eurostat, Individuals: internet activities (isoc_ci_ac_i), participating in social networks, share of individuals aged 16 to 74, 2025, https://ec.europa.eu/eurostat/databrowser/view/isoc_ci_ac_i/default/table?lang=en

4. Meta, About estimated audience size, Meta Business Help Centre, https://www.facebook.com/business/help/1665333080167380

5. Meta, Get started with the Meta Pixel, Meta for Developers, https://developers.facebook.com/docs/meta-pixel/get-started

6. Telekommunikation Digitale Dienste Datenschutz Gesetz (TDDDG), § 25 Schutz der Privatsphäre bei Endeinrichtungen, Bundesministerium der Justiz, gesetze im internet, https://www.gesetze-im-internet.de/ttdsg/__25.html

7. Telekommunikationsgesetz 2021 (TKG 2021), § 165 Datenschutz, Allgemeines, Linde Verlag law database, https://linda.lindeverlag.at/Dokument/537514_165/

8. European Communities (Electronic Communications Networks and Services) (Privacy and Electronic Communications) Regulations 2011, S.I. No. 336 of 2011, regulation 5, Irish Statute Book, 2011, https://www.irishstatutebook.ie/eli/2011/si/336/made/en/print

9. Data Protection Commission, Guidance note on cookies and other tracking technologies, DPC, April 2020, https://www.dataprotection.ie/sites/default/files/uploads/2020-04/Guidance%20note%20on%20cookies%20and%20other%20tracking%20technologies.pdf

10. Court of Justice of the European Union, Press release 99/19, Judgment in Case C 40/17 Fashion ID v Verbraucherzentrale NRW, 29 July 2019, https://curia.europa.eu/jcms/upload/docs/application/pdf/2019-07/cp190099en.pdf

11. Meta, General Data Protection Regulation: cookie consent (Meta Pixel implementation), Meta for Developers, https://developers.facebook.com/docs/meta-pixel/implementation/gdpr

12. Google, Consent mode overview, Google for Developers, Tag Platform, https://developers.google.com/tag-platform/security/concepts/consent-mode

13. Meta, Conversions API, Meta for Developers, https://developers.facebook.com/docs/marketing-api/conversions-api

14. Meta, Handling duplicate pixel and Conversions API events, Meta for Developers, https://developers.facebook.com/docs/marketing-api/conversions-api/deduplicate-pixel-and-server-events

15. Meta, Facebook and Instagram to offer subscription for no ads in Europe (originally published 30 October 2023, updated 4 December 2023 and 12 November 2024), Meta Newsroom, https://about.fb.com/news/2023/10/facebook-and-instagram-to-offer-subscription-for-no-ads-in-europe/

16. European Commission, Commission finds Apple and Meta in breach of the Digital Markets Act, press release IP/25/1085, 23 April 2025, https://ec.europa.eu/commission/presscorner/detail/en/ip_25_1085

17. European Commission, Meta commits to give EU users choice on personalised ads under Digital Markets Act, Digital Markets Act news, 8 December 2025, https://digital-markets-act.ec.europa.eu/meta-commits-give-eu-users-choice-personalised-ads-under-digital-markets-act-2025-12-08_en

18. Meta, About the learning phase, Meta Business Help Centre, https://www.facebook.com/business/help/112167992830700

19. Meta, Basic targeting, countries, Meta for Developers, Marketing API, https://developers.facebook.com/documentation/ads-commerce/marketing-api/audiences/reference/basic-targeting.md

20. Meta, Advanced targeting, locales, Meta for Developers, Marketing API, https://developers.facebook.com/documentation/ads-commerce/marketing-api/audiences/reference/advanced-targeting.md

21. Meta, Multi language ads, Meta for Developers, Marketing API, https://developers.facebook.com/documentation/ads-commerce/marketing-api/multi-language-ads.md

22. Bundesamt für Statistik (BFS), Sprachen, main languages of the permanent resident population 2024, https://www.bfs.admin.ch/bfs/de/home/statistiken/bevoelkerung/sprachen-religionen/sprachen.html

23. Eidgenössischer Datenschutz und Öffentlichkeitsbeauftragter (EDÖB), Merkblatt: Verwendung von Cookies und ähnlichen Technologien im Zusammenhang mit Online Tracking, 27 March 2026, https://www.edoeb.admin.ch/dam/de/sd-web/ByXs53ExpIEy/20260327%20Merkblatt%20Cookies_DE.pdf

24. Gesetz gegen den unlauteren Wettbewerb (UWG), Anhang (zu § 3 Absatz 3), numbers 2, 2a and 7, Bundesministerium der Justiz, gesetze im internet, https://www.gesetze-im-internet.de/uwg_2004/anhang.html

25. Directive 2005/29/EC on unfair business to consumer commercial practices, Annex I, point 7, EUR Lex, 2005, https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32005L0029

26. Meta, Website custom audiences (retention_days), Meta for Developers, Marketing API, https://developers.facebook.com/documentation/ads-commerce/marketing-api/audiences/guides/website-custom-audiences.md