Outsourcing Map 2027: the map search engines hand to software buyers, from Riyadh to Madrid

A buyer in Riyadh, Dubai or Doha types “software development company in” and the engine finishes the sentence. We asked that question 3,816 times, in four languages, on two engines and in eight markets. The result fits in one line: the list proposed in Riyadh is the same as in London and New York. India holds 37 % of the English map, the Arabic map is Egypt and Saudi Arabia, the French map is France then Morocco and Tunisia, and Algeria is practically absent. Here is what the map shows, what it does not say, and what a buyer can ask for instead.

Symloop Research10 min read
Outsourcing Map 2027: the map search engines hand to buyers3,816 queries, four languages, two engines, eight markets: where suggestions place a software partnerSTUDY · SYMLOOP RESEARCH · 2027Outsourcing Map 2027:the map search engines hand to buyers3,816 queries, four languages, two engines, eight markets:where suggestions place a software partnerRIYADH, DUBAI, DOHAthe same list as London and New YorkONE MAP PER LANGUAGEIndia 37 % in English · Algeria absentsymloop.com

Short answer. Search engines do not hand out one map per country, they hand out one per language. In English, 37 % of the places Google proposes are in India and 17 % in the United States; a Saudi buyer finds their own country in 2 % of the suggestions, an Emirati in 5 %, a Qatari in 4 %. In Arabic, the map is Egypt, 28 %, and Saudi Arabia, 27 %, then Syria and the UAE. In French, it is France, 56 %, then Morocco, 15 %, and Tunisia, 9 %. In Spanish, Latin America, 75 %, then Spain. The list is the same everywhere: the suggestions Google proposes in Saudi Arabia, the UAE, Qatar and the United Kingdom are identical to those of the United States, letter by letter; only the first line changes, and not in every country. Algeria is named once in 96 in French, and never in English, Arabic or Spanish.

This study is for anyone choosing a vendor from Riyadh, Dubai, Doha, Paris or Madrid, and for any vendor sitting outside the map. A suggestion is a phrase an engine proposes to finish yours: it is neither a search volume nor a ranking of quality. It only shows what the engine puts in front of the buyer before they have read a single page. We have an interest, and we state it: Symloop is an Algerian software company, so we are in the country the map does not show; that is why we wanted to measure the map rather than comment on it. The method and its limits are written below, and the set of 12,230 suggestions is downloadable.

Four languages, four maps

For each language, the figure counts the places named in Google’s suggestions and assigns them to their country.

Figure 1 · Share of place mentions per country in the suggestions Google proposes for seeds such as “software development company in”, in four languages. India holds 37 % of the English map, ahead of the United States; the Arabic map is split between Egypt and Saudi Arabia; the French map is France, then Morocco and Tunisia; the Spanish map is Latin America. Algeria is named once, in French. Suggestions collected on 16 November 2026; the Arabic, French and Spanish samples are small and read as an order, not as fine percentages.

The English map rests on 762 place mentions. India holds 284 of them, 37 %, with its cities at the top of the list: Bangalore, Chennai, Hyderabad, Pune, Noida, Ahmedabad. The United States follows at 17 %, then the UAE and Pakistan at 5 %, the United Kingdom and Qatar at 4 %. Saudi Arabia weighs 2 %. Algeria does not appear, Morocco appears once. The Arabic map is smaller, 82 mentions, and reads as an order rather than as fine percentages: Egypt 28 %, Saudi Arabia 27 %, Syria 11 %, UAE 10 %, Jordan and Iraq 5 % each. No Maghreb country is on it.

The French map, 96 mentions, is first of all domestic: France 56 %, with Paris, Nantes, Lyon. Then come Morocco, 15 %, and Tunisia, 9 %, with their cities, Casablanca, Rabat, Tangier, Agadir, Sfax, Sousse; then India, 7 %. Algeria is named once, by the suggestion “boite de développement informatique alger”. Across the whole collection, all seeds together, Morocco is named in 17 distinct suggestions, Tunisia in 14, Algeria in 3, one of which has nothing to do with choosing a vendor. The Spanish map, only 28 mentions, looks to Latin America, 75 %, then Spain, 25 %.

01

The same map in Riyadh, Dubai, Doha, London and New York

The figure takes the collection market by market, and compares each market with a control market that has no reason to look in the same place.

Figure 2 · The same collection, market by market. The letter-expanded suggestions Google proposes in Saudi Arabia, the UAE, Qatar and the United Kingdom are identical to those of the United States; in Arabic, those of the three Gulf countries are identical to Egypt’s. The leader’s share does not move, the buyer’s own country stays small in English, and the first line is local only on some seeds.

On Google, the 1,309 English suggestions obtained by adding one letter to the seed are identical in Saudi Arabia, the UAE, Qatar, the United Kingdom and the United States: the similarity index is 1.00 in all four comparisons. In Arabic, the 136 suggestions of the three Gulf countries are identical to Egypt’s. India’s share is 37 % in all five English markets, and the largest gap observed on any country’s share, between a market and its reference, is 0.2 point. What changes is the share of the buyer’s own country, and in English it stays small: 2 % seen from Riyadh, 5 % from Dubai, 4 % from Doha, against 17 % for an American buyer.

On Bing, the list varies more from market to market, with similarity to the United States between 0.51 and 0.62, but the map stays the same: India 34 to 35 %, United States 17 to 19 %, UAE 7 to 8 %, with a largest gap of one point between markets. Bing’s Arabic sample is too small to read, about twenty mentions. We observe these lists; we do not explain how the engines work inside. The simplest hypothesis is that a suggestion reflects what is typed in a language everywhere in the world, and that English software queries are typed a great deal from India; we did not measure that.

«In Riyadh, Dubai, Doha, London and New York, the engine proposes the same list: it does not hand out one map per country, it hands out one per language.»
02

The first line is local, the rest is not

The figure shows what Google proposes first when the buyer has typed the seed and a space, before any letter.

Figure 3 · What Google proposes first once the buyer has typed the seed and a space. Localisation exists, but it fits in one line: Dubai gets it on four English seeds of six, Riyadh on one, Doha on three; in Arabic, Riyadh gets it on three seeds of four, Dubai and Doha on none. From the second typed letter on, the list is the same everywhere.

Localisation exists, and it fits in one line. In Dubai the first suggestion is local on four English seeds of six; in Doha on three; in Riyadh on only one, “mobile app development company in saudi arabia”, the other five starting with India, Hyderabad, Kolkata or Malaysia. In Arabic, Riyadh gets a Saudi first line on three seeds of four, while Dubai and Doha get none: they are offered Egypt, Saudi Arabia or Jordan. As soon as the buyer types one more letter, the list is everyone’s list again. For a vendor, this means there is one local place per seed, and the rest of the page belongs to those the whole world already types.

03

What the map does not say, and what travels in its place

The map says where many people look; it does not say who delivers. The figure lists what a buyer can check alone, whatever the vendor’s country.

Figure 4 · A suggestion says nothing about a vendor’s quality; it says what many people typed. What the buyer can ask for instead fits in five kinds of proof that cross borders. The accented layer is the one this study practises: a method, data, a result that can be re-run.

For the buyer. A suggestion is not a recommendation. Before keeping a country or a company because it appeared by itself, ask for five things: reviews counted by a third party, named work with a client you can reach, a published measurement with its method, a live product you can open, and the people who will do the work with code ownership written in the contract. It is the same discipline as comparing three quotes and remote acceptance: what can be verified counts, what is declared does not. A project stuck with a vendor who vanished, described in this guide, almost always starts with a choice made on the first page.

For the vendor outside the map. Waiting to be completed by an engine is pointless: the list is global and it is already taken. What crosses borders is proof the buyer can check without believing you. It is also what conversational assistants read when asked for a list of vendors: reviews they can count, clients they can name, figures they can quote. A measurement that can be re-run, like an evaluation set before production or like this study, is proof of that kind.

«A suggestion says what many people typed, not who delivers. What crosses borders is reviews counted by a third party, named work, and measurements anyone can re-run.»
04

Method, limits and data

Everything above can be re-run. Here is how the collection was built, and where it stops.

  • A suggestion is not a volume. It says neither how many people search nor who is best; it says what the engine proposes.
  • One day, one machine, no account. A signed-in person sees suggestions personalised by history and location; the collection describes the starting list, not every screen.
  • Seeds shape the result. Other seeds would give other counts; the order of countries is more robust than the percentages.
  • Three samples are small. 82 mentions in Arabic, 96 in French, 28 in Spanish: read them as a ranking, not to the decimal.
  • The list of cities is finite. A place it does not know is not counted; Latin America is grouped as one region; the Arabic word “عمان”, which means Oman or Amman, is counted apart.
Figure 5 · The accented beat is the control markets: without the United States, the United Kingdom and Egypt, we would have concluded that the Gulf “searches in India”, when the same list is proposed to everyone in the same language. The 3,816 queries were sent on 16 November 2026, none failed, and the set of 12,230 suggestions is downloadable.

Engines and markets. Google’s suggestion service, queried with the market’s country code and interface language, and Bing’s, queried with the market’s locale, with no account, no cookie and no history, on 16 November 2026. Five markets studied: Saudi Arabia, the United Arab Emirates and Qatar in English and Arabic, France in French, Spain in Spanish. Three control markets: the United States and the United Kingdom for English, Egypt for Arabic. Seeds. Six in English: “software development company in”, “mobile app development company in”, “best software company in”, “app developers in”, “web development company in”, “it outsourcing company in”. Four in Arabic: “شركة برمجة في”, “شركة تطوير تطبيقات في”, “أفضل شركة برمجة في”, “شركة تصميم مواقع في”. Eight in French, including “agence développement application mobile”, “société développement logiciel”, “développeur offshore” and “nearshore développement”. Seven in Spanish, including “empresa desarrollo software”, “outsourcing desarrollo software” and “nearshore desarrollo software”. Plus 28 intent seeds, on price, choice and reviews.

Expansion and classification. Each place seed is sent alone, followed by a space, then followed by every letter of the alphabet, 26 Latin or 28 Arabic letters, so that the engine enumerates its places instead of giving only ten. That makes 1,908 queries per engine, 3,816 calls, no failure, and 12,230 distinct suggestions per market and language. A list of about sixty countries and their main cities assigns each place to a country; job queries and noise are removed from the maps. A language’s map is each country’s share of Google’s place mentions, read in a reference market, the UAE for English, Saudi Arabia for Arabic; the similarity between two markets is the Jaccard index of their lists. The file of suggestions, with market, language, engine, recognised places and classes, is downloadable here. A next Symloop Research study, in preparation, will cover measured mobile experience in the same countries.

Is this a ranking of the best countries to outsource to?
No. It is a record of what two engines propose when someone starts looking for a vendor. A country can dominate the map because many people type it, for good or bad reasons, and an absent country can have excellent companies.
Is a suggestion a search volume?
No. A suggestion is a phrase an engine proposes to complete yours. The engines do not publish the exact rule; all we know is that a proposed phrase was judged frequent and relevant enough to be shown. We count phrases, not searches.
Why is the list the same in Riyadh and in New York?
We observe it without being able to explain it from the inside. On Google, the suggestions obtained by adding a letter are identical in all five English markets; only the first line, before any letter, is sometimes local. On Bing the lists differ more, but country shares stay within one point.
Why does India hold 37 % of the English map?
The simplest explanation is that the list reflects what is typed in English around the world, and that English searches for software companies come a great deal from India. We did not measure it: it is a hypothesis, not a result.
Algeria is absent: does that mean there are no companies?
No. It means that few people type these seeds followed by an Algerian city, in these languages, often enough for the engines to propose them. The map measures a search habit, not a supply. Morocco and Tunisia, by contrast, are established on the French map.
Can the study be re-run?
Yes. The seeds, the markets, the date and the letter-expansion rule are written above, and the file of 12,230 suggestions is downloadable. Suggestions change over time: a collection redone in six months will give other lines, and the comparison will be a result in itself.
What should a buyer ask a vendor for instead?
Five kinds of proof that do not depend on the vendor’s word: reviews counted by a third party, named work with a client you can reach, a published measurement, a live product you open yourself, and the named team with code ownership in the contract. Ask three vendors from three countries the same questions.
Symloop Research · study no. 1 · open data

You are choosing a software partner from Riyadh, Dubai, Doha, Paris or Madrid: ask for the five proofs, from us as from anyone

The dataset is downloadable and the method is written above. For a replication on your sector or country, or for a project, write to us. We work from Algiers and Sétif, in French, Arabic and English.