Testing before scaling: how I use Austria or Switzerland as a bridge market before entering Germany

Germany is the destination, and a poor place to learn

When an Irish business decides to sell into the German speaking world, it almost always means Germany. That is reasonable. Germany had 83.3 million residents at the end of June 2026,¹ roughly nine times Austria, and more than four times Austria and Switzerland put together. The money is there. So is everyone else's budget, and that is the problem. A first German campaign is a public exam in a language you have not sat before, with the most expensive clicks in the region and no idea yet whether your proposition, your prices or your delivery promise will land.

I am Austrian, I have lived in London for years, and I have watched brands from Ireland do the same thing in the same order: translate the Irish site, switch the Google Ads location to Germany, spend for two months, pull out. The lesson they bring home is "Germany is hard". The lesson they should bring home is "we tried to learn in the most expensive classroom going".

What I do instead is learn in Austria, sometimes in Switzerland, and enter Germany with the answers. This article is how that works, where it misleads, and the arithmetic that decides whether a test market can answer your question at all. The view is my own, from my own market, and I say so where it is opinion rather than fact.

A bridge market is a classroom, not a smaller Germany

A bridge market is a market that speaks the same language as your destination, shares a good part of its buying habits, and is small enough that a mistake costs a month rather than a year. You go there to answer specific questions, carry the answers across, and keep the bridge market as a paying market afterwards.

The questions a bridge market can answer well are the ones about your offer: which value proposition people respond to in German, which price they accept, which trust signals they look for, how your delivery promise holds up when a parcel leaves the country. The questions it cannot answer are the ones about Germany itself: how crowded the German auctions are, how a German returns culture treats your margin, what a German works council thinks of your software. For those, read the mistakes I see brands make on the way into DACH and the legal and tax side in the cross border setup for Germany. This article stays on the test.

Worked example: the brand I will use throughout

Throughout, I will use an invented business: Leeside Outdoor, a waterproof clothing brand in Cork run by a founder called Róisín. Jackets at around €160, a ten year repair promise, an Irish cost per order of about €38 from Google Ads. Nobody real, and every number is made up so that every step can be shown. Two winters ago Leeside tried Germany cold: €18,000 over two months, 190 orders, a cost per order of €95, and a quiet retreat. Róisín is sure the jackets are good enough for the Alps. They have survived a good few Kerry winters, after all. She is not sure what she did wrong. The rest of this article is the second attempt.

Austria and Switzerland in numbers, and the one difference that matters most

Austria had 9,217,573 residents on 1 July 2026 on Statistik Austria's preliminary count.² Switzerland had 9,048,900 permanent residents at the end of 2024, the Federal Statistical Office's provisional figure.³ Germany, as above, 83.3 million.¹ So each bridge is about a ninth of the destination, big enough to buy a statistically honest sample, small enough that a wrong price for a month is a bruise rather than a wound.

The difference that matters most is not size. It is that Austria is in the EU and Switzerland is not. Austria has been a member state since 1 January 1995 and in the eurozone since 1999.⁴ Switzerland manages its relationship with the EU through what its own foreign ministry calls the bilateral approach, a set of agreements "in clearly defined areas",⁵ and it keeps its own currency, the Swiss franc.

For a shop in Ireland that difference decides what each bridge can teach:

Austria is a dry run for Germany's rules, and an easy one. A parcel from Cork to Linz never meets a customs officer, and neither does a parcel to Leipzig. Both are sales inside the single market, priced in the same euro you charge at home. What changes is VAT. Since 1 July 2021 an EU business selling goods to consumers in other member states charges the VAT of the customer's country once those sales pass €10,000 a year across the EU, and the One Stop Shop lets it declare and pay all of that VAT in one quarterly return instead of registering in every country.⁶ For a business established in Ireland the Union scheme of the One Stop Shop is registered with Revenue, electronically through the VAT OSS section of ROS.⁷ Whatever you set up for Austria, the OSS registration, the Austrian VAT in the checkout price, the delivery and returns promise, is the setup you will use in Germany. Austria lets you break it in where fewer people are watching.

Switzerland is a dry run for customs. It is the one market in this article where a parcel from Cork leaves the EU and crosses a real border. On the Swiss side the parcel meets Swiss import VAT at 8.1%, with duties and tax amounts of up to CHF 5 simply not levied, which at the standard rate is a goods value of about CHF 62.⁸ A foreign mail order seller whose small consignments reach CHF 100,000 a year is treated as supplying inside Switzerland, must register for Swiss VAT, and must appoint a tax representative in Switzerland.⁹ Swiss data protection has its own act, the Federal Act on Data Protection, in force since 1 September 2023. It applies to "circumstances that have an effect in Switzerland, even if they were initiated abroad", and a private controller abroad has to appoint a representative in Switzerland when its processing of Swiss residents' data is large scale, regular and high risk.¹⁰ The EU GDPR you already work under does not cover it. None of this is frightening. All of it is paperwork that a small premium market lets you practise on.

Worked example: Róisín picks her bridge

Róisín's three open questions are about her offer, not about customs: which proposition, which price, which delivery promise. Her product is EU bound in the long run, and the OSS registration she sets up for Austria is the same one that will carry her German sales. So Austria is the bridge. She parks Switzerland for later, as the place to test a premium price in francs once the German launch is paid for. Had she sold a €400 jacket to a mountain audience, I would have argued the other way round.

Which bridge answers which question

In my experience, and it is experience rather than a survey, the two bridges are good at different things. Austria tells you how a German speaking buyer reads your proposition and your price inside EU rules, with the same payment habits you will meet in Germany, which German buyers' preference for invoice and direct debit covers in depth. Austria also shares Germany's reading of trust: the Impressum link, the returns policy in full, the delivery date stated as a date.

Switzerland tells you whether a premium price holds, whether your landed cost promise survives a real customs border, and how your copy reads to a buyer who is used to paying more and expecting more. It also tells you something about Germany's south, which in my experience behaves more like Switzerland than like Hamburg on price, though I would not build a forecast on that.

Neither bridge tells you what the German auction will cost. Austria's clicks are, in my campaigns, noticeably less expensive than Germany's for the same keywords, which is why the learning is affordable there, and exactly why the cost per order you read in Austria is a floor for Germany, not a forecast. Carry the ratios across, never the absolute numbers. More on that below.

German is three Germans, and copy is where the bridge first bites

Here is the part that Irish teams underestimate, and the part I know from the inside. Standard German written in Germany is understood everywhere. It is not read as local everywhere. Austria has its own codified standard: the Austrian Dictionary, the Österreichisches Wörterbuch, first published in 1951 on behalf of the Federal Ministry for Education as the official rulebook for spelling in Austrian schools and offices.¹¹ When Austria joined the EU, a protocol to the accession treaty listed specific Austrian terms and gave them "the same status" and "the same legal effect as the corresponding terms used in Germany".¹² A country does not negotiate a treaty protocol over the word for apricot unless the words matter to the people reading them. Anyone in Ireland who has read Irish copy written in London will know the feeling.

Switzerland is the same story with different words. The Swiss Federal Chancellery's own writing rules for federal texts tell civil servants to use Swiss expressions as a matter of course, and are careful to say that these Helvetisms "are not dialect words" but "regional variants of the standard language", adding that Austriacisms and Teutonisms exist in just the same way. Their example list pairs Ferien with Urlaub, Spital with Krankenhaus, parkieren with parken, Konsument with Verbraucher.¹³ And the whole document is written without the letter ß, Grossbuchstaben and Fliesstext throughout, because Swiss standard German writes ss where Germany and Austria write ß.¹³

So the copy rule for a bridge market is this. Write once in standard German, properly, by someone who writes German for a living. Then let a native of the bridge market change the handful of words and the number format that mark it as local: in Austria the months (Jänner, not Januar), the everyday nouns, and the currency symbol before the amount with a space, where Germany puts it after. In Switzerland the ss, the Swiss nouns, and prices in francs. In my experience the handful is small, perhaps 20 words on a whole shop, and it is the difference between "a German brand" and "a brand that is actually here". Then carry the standard German body into Germany untouched. The deeper question of tone, why easy Irish warmth can read as flippant in German, has its own article.

Worked example: Leeside's two headlines

Róisín's Irish headline is "Lighter than you think. Drier than you need." Her German translator renders it faithfully. Her Austrian reviewer, a hill walker from Graz, reads it and asks what the jacket is made of. That exchange becomes the test. Variant A keeps the translated Irish line. Variant B leads with the facts an Austrian buyer asked for: the three layer membrane, the ten year repair promise, and "Austrian VAT included, free returns, no surprises". Both run in the same standard German, with the Austrian word list applied to both, so the only thing that differs is the proposition.

Targeting a bridge market in Google Ads, so the test measures the bridge and nothing else

A test market only works if the traffic in the test is actually from the test market. Two settings do most of the work.

The first is location. Google Ads lets you target countries, areas within a country such as regions, cities or postcodes, and a radius around a point. It offers two ways to decide who counts as being in the location: "Presence or interest", which Google recommends and which reaches people who are in the location or have shown an interest in it, and "Presence", for people who are actually there.¹⁴ For a test market, I set Presence. A German in Munich searching for an Austrian ski resort has shown an interest in Austria, and I do not want her in my Austrian sample. The recommendation is right for selling. It is wrong for measuring.

The second is language. Google decides what language a person uses from the language of their search, their settings and preferences, and on the Display network from the language of the pages they visit, and someone can count as speaking more than one.¹⁵ Google has also announced that from September 2026 the campaign level language setting disappears for Search campaigns, and ads match on the language they are written in, with ads in the language of the search term given priority.¹⁵ In practice this means the German copy finds German speakers on its own. In Switzerland, where German is the main language of more than 60% of the population and French, Italian and Romansh make up the rest,¹⁶ I go a step further and target the German speaking cantons as locations rather than the whole country, so that a French speaking click in Geneva never lands on a German page and gets counted as a failure of the German copy.

Worked example: Róisín's campaign settings

One campaign, location Austria, Presence only. Ads and landing pages in German with the Austrian word list. Two ad groups that differ only in the landing page headline, A and B, with the same budget cap each. The Irish campaign carries on untouched, as the control for seasonality. Nothing else changes for the length of the test.

The sample a test market needs, before it starts

Here is the idea that saves the most money, and it is the one teams skip. A small market does not need a small sample. The number of visitors you need to tell two conversion rates apart depends on the rates and on how sure you want to be, and not at all on how many people live in the country. Austria and Switzerland work as bridges because they are big enough to supply the sample, not because the sample is smaller there.

For two proportions, the visitors needed per variant are:

$$n = \frac{(z_{\alpha/2} + z_{\beta})^2 \left[p_1(1-p_1) + p_2(1-p_2)\right]}{(p_1 - p_2)^2}$$

Here $n$ is the number of visitors each variant needs, $z_{\alpha/2}$ is the z value for your significance level (1.96 for 5%, two sided), $z_{\beta}$ is the z value for your power (0.84 for 80%), $p_1$ is the conversion rate you expect from the current version, and $p_2$ is the rate the new version would need to reach for you to care. The difference in the denominator is squared, so halving the difference you want to detect quadruples the sample. The derivation, and why 80% power is the usual compromise, is in Foundations: sample size, and the broader question of how much data a campaign needs is in how much data you actually need.

Worked example: how many Austrians Róisín needs

Róisín's Irish site converts paid traffic at about 2.0%. She decides that variant B is only worth carrying into Germany if it lifts that to 2.6%, six tenths of a point, which on her margins is the difference between a campaign that pays and one that does not. So $p_1 = 0.020$ and $p_2 = 0.026$, with $z_{\alpha/2} = 1.96$ and $z_{\beta} = 0.84$.

Step one, the bracket: $0.020 \times 0.980 + 0.026 \times 0.974 = 0.0196 + 0.0253 = 0.0449$.

Step two, the top: $(1.96 + 0.84)^2 = 2.8^2 = 7.84$, and $7.84 \times 0.0449 = 0.352$.

Step three, the bottom: $(0.020 - 0.026)^2 = 0.006^2 = 0.000036$.

Step four: $n = 0.352 / 0.000036 \approx 9{,}800$ visitors per variant, so about 19,600 paid visitors in total. At the €0.42 average cost per click Róisín sees in Austria (my made up figure, in the range I would expect), that is roughly €8,200 of clicks. In Germany, at a made up €0.68 per click for the same keywords, the identical learning costs about €13,300, and every one of those clicks lands on a page she does not yet trust. Austria is about 9.2 million people. Finding 19,600 German speaking visitors with an interest in waterproof jackets over six weeks there is routine. That is the whole case for a bridge market in one line: the sample is the same, the price of the sample is not.

Reading the result honestly

When the sample is in, the question is whether the difference is real or luck. The tool is the two proportion z test:

$$z = \frac{p_1 - p_2}{\sqrt{p(1-p)\left(\frac{1}{n_1} + \frac{1}{n_2}\right)}}$$

Here $p_1$ and $p_2$ are the observed conversion rates of the two variants, $n_1$ and $n_2$ are the visitors each received, and $p$ is the pooled rate, all orders divided by all visitors, used because under the null hypothesis the two variants share one true rate. A $z$ beyond about 1.96 in either direction is significant at the 5% level. The null hypothesis, p values, the two kinds of error and why you must not peek daily are all in Foundations: hypothesis tests, and the fuller treatment with factorial designs is in statistical testing: sample size, lift and full factorial designs.

Worked example: Róisín's test reads

After six weeks each variant has 9,900 visitors. Variant A, the translated Irish headline, produced 198 orders, a rate of $198 / 9{,}900 = 0.0200$, so 2.0%. Variant B, the facts first headline, produced 267 orders, $267 / 9{,}900 = 0.0270$, so 2.7%.

The pooled rate is $p = (198 + 267) / 19{,}800 = 0.0235$.

The standard error is $\sqrt{0.0235 \times 0.9765 \times (1/9{,}900 + 1/9{,}900)} = \sqrt{0.02293 \times 0.000202} = 0.00215$.

So $z = (0.0270 - 0.0200) / 0.00215 = 3.26$, with a two sided p value of about 0.001. The difference is real, and larger than the 0.6 points she set out to detect. Variant B goes to Germany. Had $z$ come out at 1.4, the right move is not "run it another week and see", which is peeking. It is to accept that the two propositions are close, pick on cost, and spend the next test on price.

The cost of learning per market

Everything so far reduces to two rates, and both are the plain ratio from Foundations: ratios and rates. The first is the number everyone reports:

$$\text{CPA} = \frac{\text{spend}}{\text{orders}}$$

Here $\text{CPA}$ is the cost per acquisition, $\text{spend}$ is what the campaign cost in euro over the period, and $\text{orders}$ is the number of paid orders it produced in the same period. The second is the number nobody reports, and the one that decides where to test:

$$C_{\text{learn}} = 2n \times \text{CPC}$$

Here $C_{\text{learn}}$ is the cost of one answered question, $n$ is the visitors per variant from the sample size formula above, the 2 is there because a test has two variants, and $\text{CPC}$ is the average cost per click in the market where you run it. Since $n$ does not depend on the market and $\text{CPC}$ does, the market with the lower click price is the less expensive classroom for exactly the same lesson. Where $n$ comes from is set out in Foundations: sample size. The division is the same in both formulas and the trap is the same too: compare rates only when the denominators are comparable, which is why a month of Austrian CPA sits next to a month of German CPA in the chart below, and never next to a fortnight.

Worked example: Leeside's cost per order across the year

Here is Róisín's year in one chart, every figure invented.

Along the bottom are the five phases, in the order they happened, except that the Irish bar is the steady baseline. Up the side is the cost per order in euro, spend divided by orders for each phase. The Irish bar, €38, is where the business lives. The cold German attempt, €18,000 for 190 orders, is €95 and is the bar that nearly ended the expansion. Austria's first month, with the translated headline, the Irish prices carried across to the cent and the delivery charge added at the last step of the checkout, ran at €9,000 for 122 orders, €74. Austria's third month, after variant B won and the prices became round local prices with Austrian VAT and delivery included, ran at €9,000 for 191 orders, €47. The German relaunch, two months at €30,000 for 556 orders with the Austrian learnings carried across untouched, came in at €54.

Three things to take from the chart. The drop from €74 to €47 inside Austria is the learning, and it cost about €8,200 in test clicks to get. The gap between €47 in Austria and €54 in Germany is the auction, more competitors bidding on the same words, and it is the part a bridge market cannot remove. And the gap between €95 and €54 for the same country is the whole argument: Germany was never the problem. Arriving without answers was.

The sequence, top to bottom

Put together, the bridge market method is a short loop that you run once per question and then keep running, because a bridge market that stays live is a permanent test bed for every change you will ever want to make in Germany.

The diagram reads top to bottom. The decision in the middle is the one place teams go wrong, in both directions: declaring a winner early because a week looked green, or abandoning a test because a week looked grey. The sample size step above the decision exists so that the decision is made by the number you wrote down before the test, not by the mood on the day.

Where the bridge misleads

Austria is not Germany with fewer people, and I say that as an Austrian. Prices that read as fair in Vienna can read as high in Berlin and low in Munich. A delivery promise of three days from Cork to Graz says nothing about three days to Rostock. Trust signals carry across well, payment preferences carry across well, propositions carry across well. Absolute costs do not, and neither does the strength of the competition in the auction.

Switzerland misleads in the other direction. It is wealthier, and a price that holds in Zürich is not evidence that it holds in Dresden. Use Switzerland to learn whether a premium story works at all, and to practise your customs paperwork where a mistake is one parcel rather than a container. Do not use it to set German prices.

And both bridges share one trap with every test market: the people who buy from a new foreign brand in month one are not the people who buy in month twelve. Early adopters forgive a clumsy checkout. The mass market does not. That is why the cost per order you carry into Germany is a ratio between variants and phases, not a figure to put in a forecast. The measurement side of that, tracking one customer across three German speaking markets without the attribution falling apart, is its own article, and for a software product rather than a shop the buying process itself changes shape, as B2B SaaS sales in Germany explains.

What to do on Monday

Write down the three questions Germany has to answer for your business, and be honest about which are about your offer and which are about Germany. For the ones about your offer, pick a bridge: Austria if you want the German rules under the same euro and the same OSS return, Switzerland if your price is premium or your customs process is untested. Have your German copy written once, properly, then localised by a native of the bridge market for the handful of words that matter. Work out the sample before the campaign goes live, set location targeting to presence only, and change one thing per test. When the number you wrote down arrives, read it with the z test and act on it either way.

Slower is faster here, and the arithmetic says so. The cold German attempt in my invented example cost €18,000 and taught Róisín nothing she could use. The Austrian test cost about €8,200 and cut her German cost per order by more than 40%. The bridge did not make Germany easy. It made Germany a market she entered already knowing the answers, which is the only way I have ever seen a brand from Ireland enter it well.

Footnotes

1. Statistisches Bundesamt (Destatis), Current population, population by nationality and sex, quarterly figures (83,298,787 on 30 June 2026), Destatis, 2026, https://www.destatis.de/EN/Themes/Society-Environment/Population/Current-Population/Tables/liste-current-population-basis-2022.html

2. Statistik Austria, Population at the beginning of the year and by quarter, preliminary results for 1 July 2026 (9,217,573), Statistik Austria, 2026, https://www.statistik.at/en/statistics/population-and-society/population/population-stock/population-at-beginning-of-year/quarter

3. Federal Statistical Office (FSO), Switzerland's population passed the 9 million mark in 2024, provisional annual results of the Population and Household Statistics, press release of 3 April 2025 (9,048,900 permanent residents on 31 December 2024), Swiss Confederation, 2025, https://www.admin.ch/en/nsb?id=104703

4. European Union, Austria, country profile (EU member state since 1 January 1995, member of the single currency area since 1 January 1999), European Union, https://european-union.europa.eu/principles-countries-history/eu-countries/austria_en

5. Federal Department of Foreign Affairs (FDFA), Switzerland's European policy, the bilateral approach, Swiss Confederation, https://www.europa.eda.admin.ch/en

6. European Commission, The One Stop Shop: the Union scheme for intra Community distance sales of goods, the EUR 10,000 threshold extended to distance sales of goods from 1 July 2021, and the quarterly OSS return, European Commission, https://vat-one-stop-shop.ec.europa.eu/one-stop-shop_en

7. Revenue, Union scheme, One Stop Shop (registration for suppliers established in Ireland through the VAT OSS section of ROS, intra Community distance sales of goods, the single quarterly return), Revenue Commissioners, https://www.revenue.ie/en/vat/vat-ecommerce/union-scheme/index.aspx

8. Federal Office for Customs and Border Security (FOCBS), Online shopping, mail order, importation by companies: duties and VAT on goods ordered abroad, the CHF 5 rule and the 8.1% standard rate, Swiss Confederation, https://www.bazg.admin.ch/en/courier-by-post-online-shopping-mail-order-importation-companies

9. Federal Tax Administration (FTA), Mail order trade and platform taxation: the CHF 100,000 threshold on small consignments and the Swiss tax representative, Swiss Confederation, https://www.estv.admin.ch/en/vat-mail-order-trade-and-platform-taxation

10. Swiss Confederation, Federal Act on Data Protection (FADP, SR 235.1), consolidated text in force since 1 September 2023, Articles 3 and 14, Fedlex, https://www.fedlex.admin.ch/filestore/fedlex.data.admin.ch/eli/cc/2022/491/20230901/en/html/fedlex-data-admin-ch-eli-cc-2022-491-20230901-en-html.html

11. Austria-Forum, Österreichisches Wörterbuch (Austrian Dictionary), English entry in the AEIOU encyclopedia, https://austria-forum.org/af/AEIOU/%C3%96sterreichisches_W%C3%B6rterbuch/%C3%96sterreichisches_W%C3%B6rterbuch_english

12. European Union, Protocol No 10 on the use of specific Austrian terms of the German language in the framework of the European Union, annexed to the 1994 Act of Accession, Official Journal C 241 of 29 August 1994, EUR-Lex, https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:11994N/PRO/10

13. Schweizerische Bundeskanzlei, Schreibweisungen: Weisungen der Bundeskanzlei zur Schreibung und zu Formulierungen in den deutschsprachigen amtlichen Texten des Bundes, 2nd edition 2013, corrected 2015, section 3.6 Helvetismen (paragraphs 361 to 363), Swiss Federal Chancellery, https://www.bk.admin.ch/bk/de/home/dokumentation/sprachen/hilfsmittel-textredaktion/schreibweisungen.html (PDF as filed by the Canton of Zug: https://kr-geschaefte.zug.ch/dokumente/11187/3468-2-17179_Beilage2-Schreibweisungen-Bund.pdf)

14. Google, About location targeting, Google Ads Help, https://support.google.com/google-ads/answer/1722043

15. Google, About language targeting, Google Ads Help, https://support.google.com/google-ads/answer/1722078

16. Federal Department of Foreign Affairs (FDFA), About Switzerland: Languages (four national languages, German the primary language of more than 60% of the population), Swiss Confederation, 2024, https://www.aboutswitzerland.eda.admin.ch/en/language