Posted on May 29, 2019 by zeljkos
TLDR: Data from two small software vendors shows that conversion rates for direct software sales differ between countries by two orders of magnitude. For each thousand new website visitors, the top group of countries had 9.14 paying customers, while the bottom group had only 0.19 paying customers. Both companies used in this research share some similarities: they sell technical B2B software in the English language, use SEO, and have products in the $100 to $2000 range. Therefore, be careful when extrapolating this data to other businesses.
A few months ago, I was on a meetup where the sales manager for a small press clipping SaaS was retelling his experiences in different countries. The audience was surprised when he explained that the same sales approach that worked in Belgium, completely failed in Germany or Portugal. He also said that they blocked their website for large Southeast Asian countries because they were receiving large amounts of inquiries but never selling anything there.
A lively Q&A discussion ensued. One person was so agitated that they told the speaker he should be fired for failing to sell in the huge German-speaking market. Others defended the speaker, saying that German markets have a preference for local solutions in their own language. Most people disagreed if blocking an entire country was a good practice. Yes, India and the Philippines are buying much less software than the US, but if someone from that country wants to buy, why reject their money?
I realized I can actually answer those questions. Companies don’t publish this kind of data because they consider sales reports confidential, and because reporting on differences between countries can sound judgmental. However, I have my own source of data as I own two small software companies. For the judgmental part of the problem, I will present you with all the data and you are free to make conclusions yourself.
Facts about the companies:
- GemBox Software sells software components to .NET developers, and has six employees.
- TestDome sells automated pre-employment tests, and has nine employees.
- Both companies sell internationally via an English-only website. We don’t have localized pages.
- Both companies have a mailing address in London and are registered in the EU (Croatia).
- Customers purchase products directly with a credit card, in transactions ranging from $100 to $2400, using FastSpring as a payment gateway.
- The main sales channel is SEO, and both companies have hundreds of landing pages.
- Part of the sales also come from word-of-mouth and free referrals.
- Both companies only have inbound sales, meaning that they follow up with promising leads.
This is quite a typical setup, I know many other small software vendors that are organized in the same way.
I used two main sources of data:
- Google Analytics, where we exported New Users per country for a period of two years.
- FastSpring sales data, where we exported the number of sales per country for the same period.
As always in data science, 90% of the work was cleaning the data and figuring out what is relevant:
- Unfortunately, Google Analytics reports don’t give ISO country codes or even ISO country names. Instead, they use their own list of politically-adjusted country names. I created a mapping table to convert those to ISO country codes and names.
- Sales amounts per country turned out to be unreliable for countries with only a few customers. For example, a single large customer from Malta was repeatedly buying large packs, so they significantly increased the total sales amount for Malta. For that reason, I decided to use the number of unique new customers per country as a basis for ranking.
- Some smaller countries don’t have enough data to be statistically relevant. Therefore, I applied a filter to only include countries with more than 2000 visitors (New Users in Google Analytics) for both companies.
- Since there are two companies, the question was how to merge their data? Since median conversion values were almost similar (1.05 for GemBox vs 1.06 for TestDome), and the ranking of countries was similar, I decided that the resulting conversion value for each country is an average of both companies.
With the above adjustments, I got a list of 84 countries. The strongest economies are mostly at the top of the list, with few exceptions.
What convinced me that the data is good enough for publishing is the Slovenia-Croatia-Serbia test. That is a trick I use to see if global research is actually valid for small sample sizes. Slovenia, Croatia, and Serbia are quite small (2, 4, and 7 million people, respectively), they don’t have a strong economy, and are otherwise very similar. However, Slovenia is a bit more developed than Croatia, and Croatia is a bit more developed than Serbia. When you look at the majority of well-known indices (for example GDP per capita, Human Development Index, literacy, etc.), they align in that exact way. And, without any data mingling, the same alignment showed in the list of conversion rates.
Still, I am not comfortable assigning each country an exact ranking. The research was done with only two companies, and data totals are small. For a more fair presentation, I have grouped all countries into six groups: A+, A, B, C, D, F.
Without further ado, here is the chart with all 84 countries:
For the more data inclined, here is a spreadsheet with the data used to produce the chart:
I was quite surprised when I examined the list. I knew different countries had different inclinations for buying software online, but the difference is staggering. Per 1000 website visitors, the A+ group (Luxembourg, New Zealand, etc.) produced 9.14 new customers, while the F group (Philippines, India, etc.) produced only 0.19 new customers. The difference between the top and bottom group is 48 times, almost two orders of magnitude. Let’s take a look at each of the groups.
Group A+ contains only four countries:
|Luxembourg, New Zealand,
9.14 customers per thousand visitors
A few things to notice.
Funnily enough, only one A+ country has English as a primary language (New Zealand). It seems that there is no language barrier in a developed country where a large part of the population knows English as a second language. It may actually be a benefit, because non-native English speakers who search keywords in the English language are more likely to be highly educated, with university degrees and office jobs.
All countries in the A+ group are small, which might also be a benefit. Users in small countries often don’t expect to find a local software solution and are used to searching for software internationally.
And finally, all A+ countries are service-oriented economies and have a very high hourly wage. Luxembourg, for example, has an average hourly wage of 45 EUR/hour. It doesn’t make sense to spend days solving a problem when you can purchase existing software that is cheaper and does the task better.
Selling in A+ countries is a digital marketer’s dream. 9.14 purchases per thousand visitors means around 1% of visits result in a purchase, they buy fast and don’t have time to haggle over the price and send sales inquiries. Unfortunately, that dreamland is quite small, with only 16.7 million people.
On the other hand, Group A is much larger, with 530 million people, but half the conversion rate:
|Australia, Sweden, Switzerland, Finland,
United, Kingdom, Netherlands, Austria,
Guatemala, Ireland, United States, Canada
5.17 customers per thousand visitors
Total population: 530 million people
Here we find the usual suspects: large English-speaking countries (US, UK, Canada, Australia) and countries in Northern Europe. But why is Guatemala there, being neither and having a GDP per capita of only $4470? The number of visitors from Guatemala was small, but we still had six unique customers for GemBox and seven unique customers for TestDome. It may be a statistical anomaly, or there is something special about the English-speaking minority there, I don’t know.
The heavyweight in this group is the United States with 329 million people, more than all other A+ and A countries combined. That explains why the US is the center of the software universe and why the old startup advice is to focus on the US first. 38% of all our sales comes just from the US. Unfortunately, the US is also the most problematic to sell in this group.
When I started my first company, I made the mistake of making the initial price list in EUR. Although the payment processor offered buying in a local currency, US companies were constantly asking for a fixed price quote in USD. After we switched our price list to USD, we didn’t have any issues. For some reason, EU countries don’t have a problem with the price being in USD and local price in EUR changing daily.
That is not the only problem; US companies disproportionately ask for custom EULA and other legal documents often. I am a bit puzzled by this. In most cases I don’t understand what benefit they get out of minor legal changes, as our software always comes with an AS IS clause, and our maximum monetary liability is the purchase price. Still, many US companies are happy to spend more on lawyer fees than the actual price of the software they are buying.
If we were ever to open a foreign subsidiary, it would be in the US. But I am not sure that would influence our US conversion rates, as most of our customers don’t know or care about our location. FastSpring, our payment gateway, is in Santa Barbara (California), so US customers can purchase with local bank transfers and US purchase orders.
Next is Group B, with a conversion rate half that of the previous group:
|Germany, South Africa, Slovenia,
Estonia, Belgium, Israel, China,
France, United Arab, Emirates, Italy,
Czechia, Uruguay, Spain, Singapore,
Cyprus, Hong Kong, Greece
2.33 customers per thousand visitors
Total population: 387 million people
A few interesting notes here.
France and Germany are both in this group, although developed countries. I suspect that is because they have large local software markets. You can find a software solution in French or German, with local support.
The outlier in this group is China, but don’t be fooled. China has a good conversion rate because we rarely get visitors from China. Google Analytics shows that we get more visitors from Singapore than from all of China, which is debatable as many Chinese users use a VPN. I considered excluding China from this analysis, but I decided to leave it so you know what to expect when you see China in your website analytics.
Following is Group C, with a conversion rate of 1.23:
|Chile, Lithuania, Mexico, Mauritius,
Slovakia, Costa Rica, Argentina,
Lebanon, Croatia, South Korea, Latvia,
Brazil, Poland, Japan, Portugal,
1.23 customers per thousand visitors
Total population: 877 million people
The biggest surprise here is Japan. 126 million people, developed economy, and a lot of foreign corporations, but they still don’t buy software in English. There is a large barrier to entry in the Japanese market, and it seems that you need a completely local approach.
Then there are EU countries like Lithuania, Slovakia, Croatia, Latvia, Poland, and Portugal that have low conversion rates.
Group D has a conversion rate of only 0.66:
|Malaysia, Taiwan, Thailand, Colombia,
Romania, Hungary, Bulgaria, Azerbaijan,
Russia, Saudi Arabia, Turkey, Peru, Serbia
0.66 customers per thousand visitors
Total population: 516 million people
Here are the worst-performing EU countries: Romania, Hungary, and Bulgaria. Large countries like Russia, Thailand, and Turkey are also here.
But, they are still better than Group F, which is not converting at all:
|Sri Lanka, Kenya, Venezuela, Vietnam,
Indonesia, Nepal, Philippines,
Bosnia & Herzegovina, Ukraine, Jordan,
Dominican Republic, Ecuador, Egypt,
Bangladesh, Pakistan, Belarus, India,
Iran, Morocco, Tunisia, Armenia, Algeria
0.19 customers per thousand visitors
Total population: 2.68 billion people
This group only has 0.19 paying customers per thousand visitors, a full 48 times less than Group A+. Unfortunately, many countries in this group are very populous, like India (1.34 billion), Indonesia (286 million), Pakistan (204 million), Bangladesh (166 million), and the Philippines (107 million). To get a sense of scale, both GemBox and TestDome have more customers from Luxemoburg (0.6 million) than from Indonesia (268 million). Software is eating the world, but it seems that it is eating Western countries first.
India deserves special analysis. It is a huge English-speaking country but our companies get only 0.12 unique customers per thousand visitors. This is not a statistical fluke, as we get many visitors from India. “I don’t understand,” said our inbound sales guy, “we get many inquiries from India and I schedule many demos to prospective Indian customers, but those leads never materialize into sales.” In the end, we just stopped treating any emails or demo requests from the Indian subcontinent as leads, and we just tell them that they can purchase online. I am not sure why conversions there are so low. Maybe it’s because our prices are too high for that market, or businesses in India just don’t buy much software.
Because of that, I feel the utmost respect for Indian software companies that are able to sell in their own country. There is a common feeling that the US is a very hard market. As Sinatra sang, “New York, If I can make it there, I’ll make it anywhere…”. In the world of global software sales that seems to be false. Our data shows that the US is in the group of easy countries. But if you can sell your software to price-sensitive, mobile-first customers in Delhi, Islamabad, Jakarta, or Manila, then you can really sell it anywhere.
If you are a startup in India selling software online, do yourself a favor and focus on the US first. As an example, only recently I realized that Visual Website Optimizer, an A/B testing tool that we love and use extensively, comes from Delhi, India. But the contact phone listed on the top of their menu is a US toll-free number.
If a country didn’t make this list, it means our websites didn’t have at least 2000 visitors from that country in the last two years. That can be because a country is very small, or it has very little targeted keyword searches in English. In any case, don’t expect noticeable sales from other countries.
Since differences between countries are so large, and we are a commercial enterprise, we don’t have much choice but to treat countries differently:
- Instead of relying on anecdotal stories we hear at meetups or on Internet forums, we can estimate how prospective a lead is based on data from that country.
- We plan to do targeted pay-per-click campaigns in selected countries from Groups A and A+, as they are more likely to convert.
- Our support is putting more effort into answering questions from free users in high-converting countries.
I am not too happy about this, but there is no sense in arguing with the data. Of course, we will update our data every few years and change accordingly.
Disclaimer: Again, this is only data for GemBox and TestDome, so it doesn’t apply to cheaper consumer-oriented software, or more expensive enterprise software. But, I would love to live in a world where many companies publish their anonymized conversion data in this format, so average conversion data is known for each industry. Wouldn’t it be great if you could compare your funnel performance with similar companies? As far as I know, even Buffer, a company famous for its transparency, doesn’t publish a breakdown by countries.
If you like this, share or, even better, publish your own data.