Trusted by 50+ Korean brands entering the U.S. marketSchedule your free consultation
← Back to Insights
Analysis

How to Read Mathpresso’s QANDA Growth Like an Operator, Not a Fan

Mathpresso is a Seoul-based startup founded in June 2015 that operates the AI-based learning app QANDA. Public sources show fast revenue growth through 2023, continued operating losses, and multiple funding events including a 10 billion KRW equity investment from KT in September 2023. The useful takeaway for US-entry planning isn’t hype. It’s how to sequence proof points, unit economics, and localization decisions when your product is AI-heavy and your distribution depends on trust.

This is a sequenced playbook you can run as you study Mathpresso’s path and translate it into your own edtech or AI-driven scale-up plan.

Phase 1. What do we actually know about Mathpresso from verifiable sources?

Mathpresso is best described, from the evidence, as a Korean edtech company built around QANDA, with documented growth and documented losses. The core corporate facts are stable across sources: the company was founded in June 2015, is headquartered in Seoul, and lists co-CEOs Lee Jong-heun and Lee Yong-jae. Those basics are consistent in THE VC’s company profile and a 2021 press kit snippet. (THE VC profile for Mathpresso; QANDA official site and press/recruiting hub)

QANDA is described as an AI-based learning platform focused on problem solving. NextUnicorn’s summary is concrete: it claims the app provides a solution plus similar problems, video explanations, concept books, and lectures. That implies QANDA isn’t only a “solver.” It’s a content stack built around a high-frequency user intent: “I’m stuck on this problem.” (NextUnicorn company info for Mathpresso)

Financial performance is not fully disclosed across official filings in the packet, but Unicorn Factory Data Lab reports consolidated revenue of 24.2 billion KRW for the prior year, up 42% year over year, alongside an operating loss of 19.4 billion KRW, down 20% versus the prior year’s loss. It also presents a revenue trajectory from 2020 to 2023 (0.5 billion KRW, 2.1 billion KRW, 10.7 billion KRW, 17.0 billion KRW) and separately cites the 24.2 billion KRW figure. The exact mapping between the series and the “prior year” headline isn’t fully clear from the excerpt, so treat the direction as the reliable point: strong top-line growth with losses that are shrinking but still material. (Unicorn Factory Data Lab profile on Mathpresso)

On funding, Unicorn Factory reports a Series B in 2019 that included the global investor Legend Capital, and an equity investment of 10 billion KRW from KT in September 2023. KT’s stated intent in that coverage is tied to education-focused AI model development and service enhancement. That matters for operators because it signals what strategic investors tend to pay for: not “AI in general,” but productizable capability and distribution adjacency. (Unicorn Factory on Mathpresso funding and KT investment)

One more constraint: the Ministry of SMEs and Startups (MSS) press release on the Unicorn Bridge program provides cohort-level averages and program terms, but the visible body text does not list individual company names. The prompt states Mathpresso is in that cohort, yet that link is only said to exist in a non-visible attachment. You can use MSS data to understand what “potential unicorn” support looks like in Korea, but you shouldn’t treat cohort averages as Mathpresso’s own metrics. (MSS press release on the Global Unicorn Vision Declaration Ceremony and Unicorn Bridge)

Phase 2. How did QANDA’s product shape a globalization path?

QANDA’s product shape pushes you toward global demand because the problem it solves is universal, but the trust requirement is local. A math learner in Vietnam and a math learner in California both want fast help, yet they judge quality differently and sit inside different curricula and payment habits.

The sources point to three product layers that matter for global scaling:

  • A solution search function, described in the 2021 press kit snippet as a “solution search service,” later strengthened with a question-answer service. This implies a loop: search, ask, get validated help. (QANDA press/recruiting hub referencing company materials)
  • A broader learning bundle, per NextUnicorn: similar problems, video explanations, concept books, lectures. That’s how you turn a single transaction into longer retention. (NextUnicorn on QANDA feature bundle)
  • Model quality improvements, per Unicorn Factory: using generative AI to improve accuracy and speed of Q&A, with claimed impact on satisfaction, loyalty, and purchase conversion in global markets. Even if you treat “conversion improvement” as directional, “accuracy and speed” is the operational core. (Unicorn Factory on generative AI improving QANDA Q&A)

Here’s the unhedged opinion: most edtech global plans over-index on translation and under-index on error tolerance.

In math help, one wrong answer can kill trust for weeks. If your AI pipeline can’t hold accuracy under a new curriculum, new problem formats, and new user handwriting patterns, “localization” is theater. QANDA’s emphasis (in secondary coverage) on accuracy and speed is exactly the right axis to obsess over for cross-border expansion. (Unicorn Factory on accuracy and speed improvements)

Phase 3. What does the financial profile imply about scaling an AI-heavy edtech business?

Mathpresso’s reported numbers imply a familiar AI scaling pattern: revenue can grow fast, but profitability lags because model development and platform operations stay expensive. Unicorn Factory reports consolidated revenue of 24.2 billion KRW (up 42% year over year) alongside an operating loss of 19.4 billion KRW, improving versus the prior year. That is not a “burn equals bad” story. It’s a capacity story: product R&D, compute, content ops, and tutoring supply (if the 1:1 service is significant) can expand faster than monetization maturity. (Unicorn Factory Data Lab on revenue and operating loss)

Unicorn Factory attributes performance to growth in a 1:1 remote tutoring service and stabilization of an online academy business in Vietnam. For operators, that’s a major signal: the company is not relying on one monetization channel. It’s mixing app-driven demand with human-supported services and region-specific business lines. (Unicorn Factory on tutoring growth and Vietnam stabilization)

If you’re a Korean SME founder planning US entry, steal the logic, not the exact mix.

In the US, human-tutoring economics, labor compliance, and customer acquisition costs can look very different from Korea or Vietnam. But the principle holds: if your core AI feature creates high-intent sessions, you can attach higher-ARPU services later. What you can’t do is pretend the attachment will work without proving the core session quality first.

Phase 4. How did Mathpresso appear to use partnerships and “signal events” to reduce go-to-market risk?

The sources show two kinds of signals: institutional validation and strategic capital. Both reduce perceived risk for the next stakeholder, whether that stakeholder is a parent, a school, a platform partner, or a future investor.

Partnership and capital signals you can verify

Institutional program signals, with clear boundaries

MSS’s Unicorn Bridge program is designed to support “potential unicorns” with up to 1.6 billion KRW in government support over two years and special guarantees up to 20 billion KRW over two years, subject to budget. MSS also reports cohort-level averages: about 38.4 billion KRW average private investment raised, implied average corporate value around 180.1 billion KRW, average sales of 24.0 billion KRW, and average employment of 106. These are cohort averages, not Mathpresso’s numbers, unless you confirm the company appears in the non-visible attachment. (MSS Unicorn Bridge press release and cohort averages)

For US entry planning, these programs matter less as “funding” and more as a narrative asset: they help you explain to US counterparts why your company has staying power. But US partners will still ask for local proof.

Phase 5. What should a Korean SME copy, and what should you avoid copying?

You can copy Mathpresso’s sequencing of proof points. You should not copy its surface-level tactics without matching the underlying constraints.

  • What to copy (principle) | What it looks like in the sources | What not to assume
  • Build a core session users repeat | QANDA is centered on solving a problem and supporting Q&A, then extending to content formats like videos and similar problems. (NextUnicorn; press kit snippet) | That “more content” alone drives retention without accuracy and speed. (Unicorn Factory’s emphasis)
  • Expect losses during capability build-out | Revenue growth with operating losses that are shrinking. (Unicorn Factory) | That losses are acceptable without a clear plan to reduce them via pricing, efficiency, or mix shift
  • Use credible external signals to lower friction | Strategic investment from KT; global investor participation in Series B. (Unicorn Factory; QANDA news list) | That a Korean signal automatically translates to US trust without US references and compliance readiness
  • Invest in R&D with a traceable focus | THE VC notes national R&D projects in 2022 related to NLP, machine learning, and mathematical intelligence. (THE VC) | That “AI R&D” is a positioning statement rather than measurable improvements users can feel

The practical operator move is to define your “globalizable core session” in one sentence, then measure it like a factory line. For QANDA, the sources suggest that session is: user submits a problem, receives a correct and fast response, then continues learning via related materials. Your product may be different, but your core session has to be equally concrete. (NextUnicorn on QANDA’s layered learning outputs)

Phase 6. If you’re planning US entry, what does this case imply for your execution plan?

The QANDA case implies that US entry for AI-driven edtech is an operations problem disguised as a marketing problem. Your first bottleneck won’t be English copy. It’ll be whether your output is trusted under US conditions.

Run this sequence before you hire a US sales lead or sign a big channel partner:

  1. Define a single “trust metric” tied to correctness and speed. QANDA’s reported focus on accuracy and speed is the right model. Pick your equivalent and instrument it. (Unicorn Factory)
  2. Choose one US wedge segment you can serve without curriculum sprawl. The sources mention Vietnam business stabilization, which hints at focused execution by market. Don’t start with “all US students.” (Unicorn Factory)
  3. Design your attach motion early, but don’t force it. QANDA’s growth is partly attributed to a 1:1 remote tutoring service. That’s an attach that monetizes high-intent demand. Your attach could be test prep, teacher tools, or licensing, but it must connect to the same user intent. (Unicorn Factory)
  4. Plan for capital and signal strategy. KT and Legend Capital are signals in the record. In the US, equivalent signals might be district pilots, university partnerships, or credible edtech distributors. Don’t wait until the end to think about proof points. (Unicorn Factory)

In Prime Chase Data’s work with Korean operators, the teams that move fastest in the US aren’t the loudest. They show one tight wedge, one trust metric, and one repeatable motion from user intent to revenue. Then they scale.

Frequently asked questions

What is Mathpresso?

Mathpresso is a Seoul-headquartered Korean startup founded in June 2015 that operates the AI-based learning app QANDA, with co-CEOs Lee Jong-heun and Lee Yong-jae listed in public company profiles. Sources include THE VC and a 2021 press kit snippet.

What is QANDA known for, based on public descriptions?

Public descriptions characterize QANDA as an AI learning platform that provides solutions to problems and extends learning with related materials such as similar problems and video explanations. This is described in NextUnicorn’s company info and reinforced by a press kit snippet describing solution search and Q&A services.

Do we know Mathpresso’s revenue and profitability?

One secondary source, Unicorn Factory Data Lab, reports consolidated revenue of 24.2 billion KRW for the prior year with a 42% year-on-year increase and an operating loss of 19.4 billion KRW. Other sources in this packet don’t provide audited financial statements or a complete breakdown.

Is Mathpresso confirmed as a Unicorn Bridge company?

The MSS press release confirms 50 companies were selected for the 2026 Unicorn Bridge program but does not list company names in the visible body text, pointing instead to an attachment that isn’t shown here. Without that attachment, you shouldn’t state the selection as confirmed from the visible MSS text alone.

What’s one actionable lesson for US market entry from this case?

Treat “trust under local conditions” as the first operational milestone, not a later brand problem, because QANDA’s reported improvements center on accuracy and speed and its growth is tied to services that monetize high-intent sessions. That sequencing travels better than any single tactic.

Sources

Original MSS overview