In one of the in-depth interviews — the company is the kind that wants to be named only as “a Bishkek outsourcing firm with foreign clients” — the founder told us a story we kept hearing back in different forms. The firm had been negotiating a contract with a U.S. customer. Everything was lining up: the technical scope, the price, the timeline, the legal counterparty in the United States. Then the customer’s procurement team came back with the requirement nobody had flagged in the earlier conversations. The vendor needed to have at least 20 staff officially employed in-office at the time of contract signing.
The firm had four people. Two founders, two developers. The deal collapsed. Not on capability, not on price, not on language. On headcount, against a number set in a procurement template designed for vendors in countries that produce 20-person teams as easily as Kyrgyzstan produces three-person teams.
That story sat at the centre of what the outsourcing chapter ended up being about. Software development outsourcing in Kyrgyzstan, in 2019, was a sub-sector of roughly 2,000 to 2,500 specialists distributed across 505 registered companies (Ministry of Justice count) — but 87% of those companies had under five staff. The 8% that were High Technology Park residents were the only ones that could field 20 in-office developers for a single contract.
What the count looked like
- Specialists: 2,000 to 2,500 working in the sub-sector — the KSSDA “IT market and IT education services” study gave 180–190 active companies; we found 505 in the Ministry of Justice registry and counted 100 more from open sources.
- Revenue: 606.2 million KGS in 2017 (NSC data), with 42–47% year-on-year growth between 2015 and 2016.
- Gender: women were 6% of developers. Women in the sub-sector were more often found in management (project managers, executives), in training (IT-course teachers), and in QA.
- Regional distribution: 93% of registered firms in Bishkek; a handful in Osh doing 1C implementation work; one in Karakol with headquarters in Chicago.
- Key named players (residents of HTP, mostly): Zensoft, MadDevs, Timelysoft, Spalmalo, Mirsoft, Softline, Max Soft, Attractor Software, Namba Soft, SolTech, Prosoft, Askartec.
The sub-sector’s structure is the part of the chapter most worth re-reading in 2020. Three out of four developers in outsourcing were under-30. The senior tier was thin, and getting thinner — which the chapter on the labour market in the previous post has already documented. The HTP residents were the bridge between the small-firm reality and the international-customer expectation. The residency rules of the HTP — 80% of services exported, non-cash payments only, 5% income tax for resident employees, 12% insurance premium of average national wage — were both subsidy and constraint. They subsidised the firms that could already export. They locked out the firms that wanted to but couldn’t quite get there yet.
The HTP’s rising share
Here is what the HTP residency was doing to the revenue distribution over time.
The story under this chart is one of selection, not of growth. The HTP didn’t make resident companies grow faster on its own. What it did was attract the firms that were already on a growth trajectory and could meet the 80% export rule, and concentrated the sub-sector’s revenue in their hands. In 2014, the typical HTP-resident contract was around $5,000–$10,000 from a small overseas client. By 2017, the same firms were signing larger contracts with the same kinds of clients — small overseas start-ups, mid-sized companies outsourcing a slice of their stack — and competing for them, mostly, against firms from Belarus, Ukraine, and India.
The 80/20 rule had a side-effect the chapter spent some time on. Several respondents told us they had turned down or carefully avoided local Kyrgyz customers, because for every Som earned domestically, they needed to find four times as much in foreign revenue to keep the export balance. The result was that the best outsourcing firms in the country were structurally disincentivised from working with Kyrgyz banks, ministries, or large local enterprises that might have wanted to digitise. The non-resident 87% picked up that work instead — at lower prices, with thinner teams, and often without written contracts.
The salary picture
The pay structure was the part of the chapter that respondents read first, every time. It is also where the brain-drain story sharpens into a single chart.
Two things to read in that chart. The first is that, by Central Asian standards and by South Asian standards, the Kyrgyz junior rate of $300–$600 is genuinely competitive. The price advantage the chapter listed as the country’s first competitive advantage is real at the entry level. The second is that the gap opens up the further up the experience ladder you go. A Kyrgyz senior at a non-HTP firm earning $1,000 is being paid less than half of what their Belarus or Ukraine counterpart is earning. The gap between Kyrgyz senior and US average — roughly 9× — is the gap that the green card was equalising for the developers who left.
The asymmetry between junior and senior pay is what made the headhunting dynamic the chapter described possible. HTP-resident firms had signed a gentlemen’s agreement among themselves not to poach each other’s developers — but the agreement did not extend to non-resident firms, and the foreign customers hiring remote developers from Kyrgyzstan were not part of any local agreement at all. So the senior layer was being thinned from both above (Green Cards, Kazakhstan, Russia) and laterally (foreign customers hiring directly through Toptal, EPAM, Railsware).
The Upwork channel
Roughly 10% of the sub-sector’s specialists were freelancers in 2019. About 270–300 KG freelancers were active on Upwork at any given time.
The interesting tail is not the median Upwork freelancer at $10–15/hour. It is the people at the top. The KG freelancer on Upwork who has lifetime-visible earnings above $300,000 was based in Bishkek and worked in enterprise systems integration. The one with $200,000+ lifetime earnings was, the focus-group respondents told us, based in Karakol — the same town we had been in for the labour-market focus group a few weeks earlier, where students were enrolling in short-term certificates instead of finishing their degrees. The Karakol freelancer had not gone through any of the formal pipeline the rest of the chapter is about. They had found foreign customers, built a portfolio in English on Upwork, kept their contracts, and reinvested into building their own micro-firm in the Issyk-Kul oblast — a chain of decisions made entirely outside the institutional system the report was studying.
That pattern — high-end freelancers operating downstream of the regional institutional infrastructure, not because of it — was one of the things the report flagged for the recommendations section. The country had a freelance base earning roughly $1.39 million USD in all-time visible Upwork earnings as of February 2019, almost none of it visible to the Kyrgyz banking system, the tax authority, or NSC’s labour statistics.
The development companies in three groups
The chapter sorted the 505 outsourcing firms into a typology that became useful later for the recommendations:
-
HTP residents — 45 firms at the time of the research (8% of total). LLCs or Individual Entrepreneurs. Average company age 5+ years. Export-oriented, foreign clients, 1% turnover paid to the HTP Directorate in lieu of most taxes. Many had been in operation for over a decade. These were the firms with the headhunting agreement, the foreign-customer relationships, and the salaries at the top of the ladder.
-
Non-residents working with foreign clients — a smaller group, want to be HTP residents but cannot hit the 80% export threshold, or cannot field the 20-person staff foreign customers require. Several respondents from this group described the bind directly: they did some local work to cover OPEX, which kept them under the 80% export rule, which kept them out of the HTP, which kept them away from the larger contracts they wanted.
-
Non-residents working locally — 80% of the population, average company age 1–3 years, small orders ($2,000 per contract typical), turnovers of $5,000–7,000. Listed themselves on Google Maps and 2GIS. Worked on trust, often without written contracts, often without a website. Named in the chapter: Prosoft, Sunrise, Goog Studio.
-
Foreign-registered firms with KG operations — companies registered in the United States or the UK with offices in Bishkek. B12, OpenCBS. Plus international firms hiring KG developers remotely without any local presence: OptDyn, EPAM, Railsware, Toptal.
-
Freelancers — ≈10% of the sub-sector total, split roughly evenly between foreign-facing (270–300 active on Upwork) and locally-facing (about 300, on freelance.kg / diesel.elcat.kg / lalafo.kg / rgb.kg).
What foreign customers actually asked for
When we asked outsourcing firms what their foreign customers’ decisive selection criteria were, the answers ranked almost identically across the survey and the IDIs:
- English-language fluency — 59% of mentions. Far ahead of the rest.
- Communicative skills of the executing team — repeated by multiple respondents: ability to sell, ability to defend a position, ability to keep the customer informed.
- Reliability — 7.7% of mentions, surfaced more strongly in IDIs than in surveys: portfolio, founders, average age of the team.
- No special requirements — for the 10.3% of firms whose customers were in Kazakhstan or Russia.
“Knowledge meeting market requirements” and “low prices” appeared in the survey but neither were ranked first by respondents who actually worked with foreign clients. The technical capability was assumed. The differentiation was on the communication layer.
The pattern below that surface — and the chapter spelled this out — was that Kyrgyz developers had a marginally better English level than their post-Soviet competitors in Belarus and Ukraine, but Kyrgyz outsourcing firms had weaker sales-house functions. A Kyrgyz developer could communicate with the customer once a relationship was running; a Belarus firm could close the customer in the first place. That’s where the Kyrgyz pipeline kept leaking.
The geography of the customer base
Forty percent of foreign customer revenue, in 2019, came from the United States. The next tier — Ireland, Kazakhstan, Russia, China, Japan, the UK, Tajikistan, Australia — together made up another 40%. The remaining 20% was scattered across the EU, South America, Southeast Asia, Sri Lanka, and several African countries.
Customers rotated. The same firm rarely held the same customer for more than 9 years, which sounds long but is in fact the duration of one or two consecutive multi-year contracts in a sector where typical engagements run 3–24 months. Long-running customers came from Upwork relationships that had matured beyond the platform: a small-scope Upwork contract led to trust, trust led to a side-contract, the side-contract migrated to direct billing, the direct billing extended for several years.
Freelancers chose countries by solvency, the chapter noted bluntly. US, China, Western Europe for those who could work in English. Russia and Ukraine for those who couldn’t — at thinner margins, with frequent reselling, and with intermittent payment friction.
The problems firms named
The eight cross-sector problems from the previous post all show up here, but the outsourcing-specific texture is worth getting on the record. Three sub-clusters of problems emerged:
Problems that hit firms working in the local market. Local customers without the ability to write a Terms of Reference. “They do not want to pay, they do not know how to work, they do not know what Technical Assignment is, that is, it is a huge stress to work with local customers.” Education that did not match market requirements. Gender and age prejudice — one female founder’s quote: “Our office is small and the team is small, we are all still young and we are not taken seriously and we are rated cheaper. It is very hard to work with customers, I am a young woman and I look like a teenager.” Difficulties with legal contract registration for short-term work: “Most recently, we received an order for 3 months of work, found an IOS developer, and tried to arrange it, but our lawyer could not find a way to do it.”
Problems that hit firms working in foreign markets. Education mismatch, lack of qualified and experienced personnel, expensive specialists, irregular professional events, low country recognition, the 80/20 balance, weak communication skills, brain drain, high cost of licensed software (Jira at $10/user/month was named explicitly: acceptable in the US, expensive at scale in the CIS), and the size problem — even the largest Kyrgyz firm with 50+ developers was small by international standards, where signing major contracts required teams of 200+.
Problems that hit freelancers specifically. Low country awareness as an outsourcing destination. Banking delays of a week or more through correspondent banks. Payoneer commissions of $15 per $1,000 transferred, and 0.5% on currency exchange. A working pattern several IDI respondents described directly: “Money goes long. 2–3 days before Payoneer and the same until the bank. More profitable than taking a Payoneer card.”
What the firms themselves said they needed
When asked what would help, the outsourcing-firm responses clustered into five recommendation areas — and these are the ones the chapter put forward to USAID:
To increase the supply of specialists in the short term: training centres run by existing companies on the basis of universities and vocational colleges (the Belarus model was named explicitly); a migration programme to accelerate residence permits and work permits for IT specialists from elsewhere in Central Asia.
To increase supply in the long term: curriculum updates with the Ministry of Education on blockchain, AI, big data, NLP, business analysis, data science; a mechanism for training existing university teachers, with foreign specialists brought in for long-term residencies; communication and English-language coverage at undergraduate level.
To meet international quality standards: open regional certification centres for the C language (CPP Institute), Python (Python Software Foundation), and for IBM/Microsoft data-science and AI tracks.
To raise international recognition of Kyrgyzstan as an IT destination: HTP to obtain accreditation from international outsourcing associations (gsa-uk.com, iaop.org); publication in international professional venues (clutch.co, hackernoon, themanifest.com); participation in international outsourcing exhibitions.
To raise average profitability: soft-skills training for existing players (pitching, networking, business correspondence); English-language courses during working hours; ISO-9001-2008 certification for HTP residents.
To increase cash flow from foreign sources: Ministry of Economy to formally enable Kyrgyz citizens to accept payments to domestic-bank accounts from Behance.net, Upwork, and similar international platforms, with the legal classification of the income clarified.
The last of these became, more than any other recommendation in the report, the one that the pandemic settled. By the second half of 2020, the conversation about Upwork income recognition had moved from “this is a niche complaint from 300 freelancers” to “this is a significant fraction of how the country’s IT workers are now being paid.” We will pick that up in the recommendations post.
What the outsourcing chapter left unsaid
There is one observation from the field that did not make it into the official chapter and is worth getting on the record here. The Karakol freelancer earning $200,000+ on Upwork, and the Bishkek freelancer earning $300,000+, were both, at the time we identified them, paying no Kyrgyz income tax on that revenue — because the legal infrastructure to formally recognise Upwork income as Kyrgyz-source income did not exist. They were not evading. They were waiting for a category that fit. The State Tax Service had a 6,388-person register of people paying voluntary patents under “computer and photocopy services” in 2018; a meaningful fraction of those, in the IDIs, were software developers using that patent because they could not find a better one.
The total annual income tax forgone, on this single channel, was probably between $1 million and $3 million USD by 2019. Compared to a sector that contributed 3% of GDP, the forgone revenue is modest. But the lesson the chapter drew was that the recognition problem and the tax-fairness problem are the same problem viewed from two ends. Until the income can be formally recognised, the developer cannot get a bank loan, cannot prove their salary for visa purposes, cannot serve as a Kyrgyz reference for an inbound foreign client — and the state cannot tax them. Nobody wins.
The next post in the series picks up software publishing, which is where the outsourcing firms’ product ambitions actually live.
Source: USAID Enterprise Competitiveness Project (2019). Analysis of the Information Technology Sector in the Kyrgyz Republic. Implemented by KG Labs Public Foundation; commissioned by USAID ECP / Nathan Associates / ACDI-VOCA, June 2019. Outsourcing sub-sector covered in primary research January–March 2019 across Bishkek, Osh, Jalal-Abad, Karakol: in-depth interviews with founders, developers, HTP officials, KSSDA representatives, and freelance practitioners; surveys of 106 IT sector respondents; Upwork profile scrape, February 2019.