Coursera (COUR) 2026 年第三季法說會：隨 Helix 成為優先事項利潤率擴大

重點摘要

• Coursera 公布 2026 年第三季營收為 3.77 億美元，年增 94%，主因為 5 月 11 日完成交易後納入了 Udemy 的業績。在標準化基準下，營收則下滑約 3%。
• 調整後 EBITDA 達到近 6,200 萬美元，佔營收的 16.3%。列報的利潤率較去年同期擴增超過 800 個基點，不過其中包含了約 1,000 萬美元的季度收購會計權益。
• 公司預計在年底前實現超過 1.5 億美元的年化總成本節省。管理層仍按計劃在交易完成八個月後達成 1.15 億美元的年化淨協同效應目標，優於原先規劃的 24 個月。
• 企業部門營收為 2.01 億美元，在標準化基準下年增 2%。淨留存率微幅提升至 93%，但管理層表示這仍遠低於其長期目標。
• 消費部門營收在標準化基準下下降 9%，儘管付費訂閱用戶數成長 30% 至超過 160 萬。訂閱人數的成長被單項證書訂閱及 Udemy 交易型產品的疲軟所抵銷。
• Project Helix 是公司的首要產品重點。Coursera 計劃於 2027 年上半年對企業客戶全面開放，隨後再審慎向一般消費者推出，但在 2027 年的初步展望中，假設 Helix 不會帶來顯著的營收貢獻。

核心財務數據

業務與營運表現

在多年期合約的支持下，企業部門仍是 Coursera 最穩健的營收來源。該公司在本季度結束時擁有 11,897 家企業客戶，年減 4%。超過 90% 的客戶流失來自年合約價值低於 50,000 美元的帳戶。管理層還表示，北美大型企業客戶的留存率明顯強於中型市場及非北美帳戶。

企業部門毛利為 1.63 億美元，部門毛利率為 81%。Coursera for Campus 與 Coursera for Government 仍佔企業業務較小部分，但管理層形容其展現出健全的成長。

消費部門毛利為 1.16 億美元，部門毛利率為 66%。消費訂閱營收約為 1.31 億美元，佔該部門營收的 70% 以上。Coursera Plus 持續成長，但新增訂閱用戶主要集中在每用戶平均收入較低的地區。

管理層將消費端壓力部分歸因於流量、SEO 和探索模式的變化，以及消費者信心較弱和高通脹影響了可支配支出。各類技能的需求也有所轉變：Python 依然受歡迎，但管理層指出，隨著 AI 改變學習者所尋求的技能類型，其熱度已低於三、四年前。

Project Helix 旨在結合技能識別、適應性學習、實作練習、評量與能力驗證證明。Coursera 正與特定的客戶和合作夥伴共同開發該平台。管理層表示，Helix 的投資已納入目前的營業費用結構中，在推出前無需擴編團隊。

管理層業績展望

針對 2026 年第四季，管理層預期：

• 營收介於 3.58 億美元至 3.66 億美元。
• 調整後 EBITDA 介於 5,500 萬美元至 5,900 萬美元。
• 調整後 EBITDA 利潤率約為 16%。
• 企業部門成長大致平穩，而消費部門因行銷漏斗指標放緩而將進一步減速。

針對 2026 全年，Coursera 將列報營收展望收窄至 12.29 億至 12.37 億美元。對應的標準化範圍為 15.03 億至 15.11 億美元。管理層維持全年調整後 EBITDA 利潤率約 14% 的目標。

Coursera 未發布正式的 2027 年財測。其初步模型假設為各部門第四季的軌跡將大致代表 2027 全年的標準化成長：企業營收大致平穩至微幅下滑，消費營收則出現雙位數低段的衰退。

這些假設包含了潛在的營收負協同效應，但未納入交叉銷售貢獻或顯著的 Project Helix 營收。管理層預期 2027 年調整後 EBITDA 金額將面臨壓力，主因營收假設較弱，且 2026 年確認的每季約 1,000 萬美元收購會計權益有所遞減。公司預計加速成本節省與持續的費用紀律將能抵銷這些逆風。

風險與關注事項

• 消費端領先指標仍低於管理層預期，單次購買、交易型產品及單項證書訂閱均面臨壓力。
• 企業淨留存率為 93%，仍低於管理層的長期目標，而學習與發展（L&D）預算持續面臨壓力。
• Coursera 對 2027 年的初步假設顯示企業部門成長有限，消費部門將出現雙位數低段的衰退。
• Coursera 與 Udemy 重疊客戶可能產生的營收負協同效應已納入 2027 年規劃中，儘管管理層表示截至第三季實務上尚未出現此情況。
• 由於收購會計影響、併購費用以及營收逆風帶來的營運資金壓力，自由現金流預計將持續低於調整後 EBITDA。然而，管理層預期公司在 2027 年仍將保持強勁的現金產生能力。
• Project Helix 將分階段推出，投資步調將根據客戶採用情況、使用率及商業化實據來調整。

分析師問答重點

管理層表示，消費端策略仍以訂閱制為核心，特別是 Coursera Plus 與 Udemy 個人方案（Udemy Personal Plan）。公司指出，與單門課程訂閱相比，Coursera Plus 能提供更好的客戶價值與更高的客戶終身價值。

針對企業需求，管理層強調 Helix 旨在解決勞動力轉型問題，而非僅作為一般的學習福利。早期設計合作夥伴反應積極，公司研究顯示客戶付費意願更高，儘管目前尚未公布定價與套裝方案。

管理層表示，營收負協同效應目前仍屬理論層面。截至目前，同時使用兩個平台的客戶在續約時並未出現可測量的負協同效應，但公司已保守地將潛在重疊因素納入其 2027 年的初步假設中。

公司尚未將淨留存率季增至 93% 視為一種趨勢。管理層認為最大的執行機會在於中型市場客戶及北美以外地區，這些領域的留存率弱於北美大型企業。

完整電話會議逐字稿

完整財報電話會議逐字稿

管理層陳述

Operator

Ladies and gentlemen, thank you for standing by, and welcome to Coursera's Third Quarter 2026 Earnings Call. [Operator Instructions] This call is being recorded. [Operator Instructions] I would now like to turn the call over to Cam Carey, Vice President of Investor Relations. Mr. Carey, you may begin.

Cam Carey

Good afternoon, and thank you for joining us for Coursera's Third Quarter 2026 Earnings Conference Call. Today, I'm joined by Greg Hart, our President and Chief Executive Officer; and Mike Foley, our Chief Financial Officer. Following the prepared remarks, we'll open the call for questions.

Our earnings press release was issued after market close. It is available on our Investor Relations website at investor.coursera.com where this call is being webcast live and where today's materials, including our quarterly shareholder letter have been posted.

During this call, we will present both GAAP and non-GAAP financial measures. Reconciliations of non-GAAP measures to the most directly comparable GAAP measure can be found in today's release and supplemental materials. Unless otherwise specified, all financial results discussed are presented on an as-reported basis and all growth percentages refer to year-over-year change.

As-reported results include Udemy only for the periods following the transaction closed on May 11. References to normalized performance include Udemy's pre-close results and are provided solely for illustrative comparability. All business metrics reflect a unified reporting methodology. Finally, all statements made during this call relating to future results and events are forward-looking statements based on current expectations and beliefs. Actual results and events could differ materially due to a number of risks and uncertainties, including those discussed in our earnings release, shareholder letter and SEC filings.

Please refer to today's earnings press release for more details on our forward-looking statements. And with that, I'll turn it over to Greg.

Gregory Hart

Thank you, Cam, and good afternoon, everyone. We appreciate you joining us. Q3 was our first full quarter as a combined company, giving us a clearer view of both our progress and the work ahead. We delivered $377 million of revenue and nearly $62 million of adjusted EBITDA. Mike will walk through the details, but the results show 2 things. We are strengthening our financial profile as committed.

Reported adjusted EBITDA margin expanded by more than 800 basis points year-over-year, doubling to approximately 16%. At the same time, our top line trajectory, particularly in consumer, remains below our growth ambition. We are staying focused on what we can control, and that starts with the integration.

Since May, we have simplified the combined organization, integrated our go-to-market teams and aligned around a unified product vision and road map. The pace matters as much as the actions, and we are moving faster than planned. We now expect to have more than $150 million of annualized gross cost savings in place by year-end. Those savings underpin the transaction's $115 million annual run rate net synergy target, which we are on track to deliver just 8 months after close compared with our original 24-month plan.

As we have said, integration is the foundation, not the finish line. Its purpose is to create the operating discipline, focus and investment capacity to build what comes next. Improving our growth trajectory will require disciplined execution and products that create meaningfully more value for customers. That is where I want to focus today, the problems our customers face and the platform we are building to solve them.

We are in the middle of one of the largest technology build-outs in history. Goldman Sachs estimates global AI-related investment will reach roughly $1 trillion in 2026. That investment is moving far faster than workforce readiness. Earlier this year, Deloitte found that 43% of business leaders say their organizations are highly prepared on AI technology infrastructure while only 20% say the same about talent, a gap of more than 2:1.

That gap is why, as I said in our last call, AI raises the bar for learning. Infrastructure alone does not produce a return. Companies can deploy the technology, but people still have to redesign workflows, exercise judgment and know when to trust the output and when to question it. Investment in people now needs to catch up with investment in technology.

We recently published our 2026 Global Skills Report, now in its eighth edition. For the first time, it draws on learning signals from the combined Coursera and Udemy ecosystem, and our data shows how learners are already responding. Learners are not choosing between technical AI skills and human skills. They're building both.

Across our platforms, the GenAI enrollment rate increased from 25 per minute in 2025 to more than 40 per minute in 2026. At the same time, critical thinking enrollments grew over 100% year-over-year. That combination makes sense. As AI becomes more capable, people need the fluency to use it and the judgment to frame the right problem, evaluate the output and lead through change.

Learners and educators face a related challenge. AI is extraordinarily good at giving us answers, but getting an answer is not the same as helping us learn or build a durable skill. Research from the University of Pennsylvania illustrates why the design matters.

In a field experiment with nearly 1,000 high school math students, access to a standard chat-style interface improved performance on practice problems by 48%. But when the AI was removed for the exam, those students performed 17% worse than the control group. The encouraging news is that a tutor designed with learning guardrails largely mitigated that negative effect.

That distinction is central to our strategy. If the value of online learning is simply access to information, AI makes that value easier to replicate. The durable value lies in trusted expertise, effective practice and feedback and credible proof that people can do the work. We believe AI can strengthen each of those when it is grounded in learning science and designed to build real capability.

Coursera and Udemy's history is rooted in expanding access to learning at global scale. We now need to build on that foundation to address the tougher workforce questions businesses face. What skills do we need to achieve our business goals? How do we help each individual build and apply them quickly in the context of their daily work? And how do we know with verifiable evidence that capability has improved.

I believe solving these problems is one of the defining opportunities of the AI era. In September, at Forward, our annual customer event, we gave customers their first look at Project Helix. Project Helix is our adaptive AI-native skills platform that draws on the combined strengths of Coursera and Udemy to connect trusted learning to verified proof of skills. Helix is the clearest expression of the unified platform strategy I described last quarter.

The experience starts with the outcome a customer wants to achieve, for example, preparing a workforce for a new AI-enabled process or helping an individual move into a new role. From there, Helix is designed to identify the skills required, assess the learner's current capability and assemble an adaptive path through trusted learning and practice with credible evidence of what the learner can do.

Three capabilities make that possible. First, skills intelligence tied to business priorities. We are building Helix to translate a goal expressed in natural language into the skills and development path needed to pursue it. Second, personalized and applied learning. The aim is an experience that continuously adapts to the learner's role, goals, progress and organizational context, bringing relevant learning and practice into the flow of work.

Third, proof. Graded assessments, hands-on practice and recognized credentials provide evidence of demonstrated skills that learners and leaders can use. We believe connecting trusted learning, relevant context and credible proof at enterprise scale is what will differentiate Helix. The value of that connection can compound over time as the system learns which skills matter, how people build them and what evidence best demonstrates readiness.

How we build Helix is just as important as what we build. Our forward keynote was a preview. We are taking a phased approach, working with a select group of customers and partners to shape the experience and targeting broad enterprise availability in the first half of 2027. Helix has to earn the right to scale. That means putting new learning experiences in customers' hands, measuring whether they solve the problem better than what exists today and investing in what works.

Alongside that road map, we continue to invest in the Coursera and Udemy products customers rely on today. Many of the capabilities we are shipping this year deliver value to customers while laying the groundwork for Helix. Let me give you 2 examples.

First, moving from practice to proof. Completion and learning hours are easy to report. Verified skills are not. Created role play places a learner in a realistic scenario such as a tough negotiation and uses interactive learning with an AI persona to evaluate the learner against goals set by the course author.

Created Labs are built to do the same for technical skills, such as a developer deploying a web application in a live sandbox environment. The goal is evidence of what a person can do, not just that they finished. Second, learning in the flow of work. Employees increasingly turn to AI tools for answers, and those answers may not lead to trusted learning or practice.

Our integrations with ChatGPT, Claude and Copilot are designed to recognize a learning moment and point the learner to a relevant course, video or guided practice. On Udemy, connectors can also work in reverse. When a tool identifies that a sales representative needs coaching, they can send that insight to Udemy to create a ready-to-use role play.

Helix is being designed to connect these capabilities around the customer's goal so that learning, practice and assessment work together as one system that turns learning data into action, understand the goal, identify the skill, build it, apply it and prove it. That system is only as strong as the ecosystem behind it.

As AI makes access to information more abundant, we believe trusted expertise becomes more essential. Our ecosystem provides that expertise, along with the context, responsiveness and verification to keep pace as skilling needs change. Today, Coursera and Udemy have reached more than 310 million learners and serve more than 1.6 million paid subscribers and nearly 12,000 enterprise customers.

Our instructor and partner communities contribute complementary forms of a valuable asset, human expertise. Udemy brings a global community of nearly 100,000 instructors who can respond quickly as tools, roles and skills change.

Coursera brings deep relationships with more than 425 universities, leading companies and professional institutions that provide trusted instruction and recognized credentials. Together, they provide a broader, more responsive foundation for learning, practice and assessment that is central to Helix.

In Q3, we expanded our ecosystem and tested new ways to connect its expertise. At Forward, we announced 16 new university and industry partners and 25 new professional certificates. New partners included Harvard Business School Online in leadership and Business Strategy, GitHub in agentic AI and modern software development and the American Medical Association and Taipei Medical University in Healthcare and the Life Sciences.

Existing partners are moving quickly as well. Earlier this month, Anthropic nearly doubled its Coursera catalog with 10 new courses for developers, small business owners, creators and teachers. Anthropic plans to publish more courses each month to help learners stay current as AI tools and techniques evolve. We are also beginning to test how expertise can move across platforms.

Dr. Ryan Ahmed, a professor and Udemy Instructor, who has taught several hundred thousand learners, adopted a series of courses covering Claude into a specialization on Coursera. This is one of our first experiments bringing a Udemy Instructors expertise to Coursera's learner base, and it became one of our top new launches of 2026. A second experiment, AI Series, helps learners connect the latest AI skills to their roles.

These learning paths combine content from multiple partners and are reviewed by practitioners in each field. Learners receive a curated plan to track their progress with the right content in the right sequence for their chosen field. We started with AI series for data analytics, business leadership, software development and cybersecurity and plan to expand to more fields in Q4.

Both experiments are informing Helix, how expertise is best leveraged across platforms and how content from different experts comes together around the learner's needs. Today, we are using expert curation. Over time, we intend to make these capabilities systematic with Helix using stronger signals of internal and external skill demand to intelligently assemble learning, practice and proof around each learner's needs and each organization's goals.

The architecture is intended to be flexible enough to support that ambition as technology, the labor market and skills needs change at an increasingly rapid pace. Our instructors and partners are also helping shape that design. Earlier this month, at our first partner and instructor Forum in Madrid, more than 80 leaders from our university, industry and instructor communities work directly with our team on the future of AI-native learning.

They see real opportunity in the combination and several have already joined our design work. As Helix develops, we want to give these experts better ways to turn their expertise into adaptive learning, practice, assessment and proof and reach the right learner at the right moment. That combination, trusted institutional expertise, practitioner speed, broad distribution and increasingly intelligent orchestration through Helix is what we believe can make the ecosystem more valuable to all participants over time.

The progress this year sets up our next phase. In May, we closed the transaction. In June, we established the financial and integration framework for the combined company. Since then, we have accelerated integration and previewed the product strategy that work is meant to enable. As we look to 2027, our priorities are straightforward. First, complete the major integration milestones with speed and discipline, protecting continuity for customers, learners, instructors and partners while improving commercial execution.

Second, build Helix alongside our customers, progressing through alpha and beta programs toward our target of broad enterprise availability next year. Third, turn our greater scale, data and investment capacity into new products that drive stronger customer outcomes. Integration creates capacity. Product execution turns that capacity into customer value and customer value is what unlocks more durable growth over time.

With that, I will turn it over to Mike to walk through the financial results and outlook. Mike, please go ahead.

Michael Foley

Thank you, Greg, and good afternoon, everyone. Q3 was a solid quarter. We delivered on our revenue and adjusted EBITDA goals, moved quickly on integration and continue to execute against the financial framework we established in June. We reported total revenue of $377 million, up 94% year-over-year, driven by the inclusion of Udemy's results beginning May 11.

On a normalized basis, revenue declined about 3%, consistent with the near-term revenue trends we outlined earlier this year. As we discussed in July, the drivers of our revenue performance continue to vary across our segments. Enterprise remains our most durable revenue source, anchored by multiyear contracts even as growth moderates.

At the same time, consumer revenue remains in transition. Subscription momentum continued, but softer leading indicators and pressure in onetime purchases and transactional offerings more than offset that growth. Even so, the mix of our revenue continues to shift towards recurring revenue. Enterprise and consumer subscriptions now represent nearly 90% of total revenue, up 500 basis points year-over-year on a normalized basis.

For the remainder of my remarks, I'll discuss our non-GAAP financial measures unless otherwise noted. Gross profit was $214 million, representing a gross margin of 64%, driven by the contribution of Udemy's higher-margin enterprise revenue and continued rate improvements. Operating expenses totaled $186 million or 49% of revenue, down 120 basis points year-over-year, reflecting sales and marketing efficiency and reduced duplication across corporate functions.

Net income was $57 million or 15.2% of revenue, and adjusted EBITDA was nearly $62 million or 16.3% of revenue. As a reminder, adjusted EBITDA includes a purchase accounting benefit of approximately $10 million per quarter in 2026. In June, we said this combination would produce a stronger earnings profile through higher gross margins, disciplined cost actions and synergies.

Only a portion of that early benefit is reflected in our reported results, but the work behind it is now well advanced. We conducted extensive pre-close planning. We moved quickly to simplify the combined company and remove duplication across overlapping functions, and we integrated our go-to-market organization, aligning global account coverage with customer continuity as the priority.

As Greg mentioned, based on the actions completed and underway, we now expect to have in place more than $150 million of annualized gross cost savings by year-end. As a reminder from our June call, our $115 million annual run rate net synergy target reflects those gross savings, less approximately $35 million of potential gross profit impact from revenue dis-synergies embedded in our 2-year framework.

Those revenue dissynergies are ultimately an estimate. Our focus has been on what we can control, executing the cost actions required to deliver our gross savings commitment. As I've said, synergies are table stakes, and I'm pleased with the team's speed and discipline in compressing a 24-month plan to 8 months, particularly while the top line remains under pressure. Our primary goal for 2026 has been to move through the most disruptive parts of integration thoughtfully and quickly so we can increasingly focus on commercial execution, product delivery and the most important driver of future revenue, customer value creation.

That shift is critical to improving our growth trajectory over time.

Turning to cash performance and capital allocation. We reported nearly $5 million of free cash flow, which included approximately $28 million of merger and integration-related costs, consistent with the onetime cash expenses outlined in our July outlook. We ended the quarter with approximately $776 million of unrestricted cash, cash equivalents and marketable securities with no debt.

The lower balance reflects repurchases, our strategic investment and onetime cash expenses. Our capital allocation priorities remain the same: complete the integration, fund Project Helix and our product road map and return capital when we see attractive value. In Q3, we repurchased approximately 18 million shares for just over $100 million. Through today's call, total repurchases under our $500 million authorization now stand at more than $215 million.

Turning to Enterprise. Enterprise revenue was $201 million, up 214% year-over-year as reported and 2% on a normalized basis. We ended the quarter with 11,897 enterprise customers, down 4% year-over-year. More than 90% of logo churn was concentrated in accounts with less than $50,000 in annual contract value. Our net retention rate was 93%, a slight improvement from last quarter, but still well below our long-term ambition.

Demand varies by use case and region, while customer continuity remains the priority following the integration of our go-to-market organizations in July. Segment gross profit was $163 million, and segment gross margin was 81%, benefiting from stronger content economics. As we look to Q4, our enterprise priorities are executing the year-end renewal cycle, working with our early project Helix design partners to incorporate customer feedback directly into what we build and making it easier for our sales teams and channel partners to bring the complementary offerings of Coursera and Udemy to customers around the world.

Turning to Consumer. Consumer revenue was $176 million, up 35% year-over-year as reported and down 9% on a normalized basis. Consumer gross profit was $116 million and segment gross margin was 66%, driven by stronger content economics and subscription mix. Consumer subscription revenue was approximately $131 million, representing more than 70% of segment revenue. That includes Coursera Plus, which continued to perform well across monthly and annual offerings as well as Udemy's early but fast-growing subscription mix.

We ended the quarter with more than 1.6 million paid subscribers, up 30% year-over-year. Subscriber growth continued to outpace subscription revenue growth, largely reflecting the geographic mix of new subscriber additions. Core subscriptions remain in our growth engine, but that growth is more than offset by pressure in individual certificate subscriptions and Udemy's transactional offerings.

Leading indicators remain softer than we want, and our outlook does not assume an immediate rebound. As Greg discussed, we are advancing an ambitious product road map. That ambition comes with trade-offs, and we're making them deliberately guided by 2 principles: First, pace milestones with proof. Project Helix will roll out in stages through 2027, beginning with enterprise, and we will invest behind each phase as evidence from customers and learners supports it.

Second, growth with returns. In consumer, we continue to hold paid marketing to a strict return threshold rather than pursuing volume without attractive economics. As positive signals emerge, we will be in a strong position to invest behind them while remaining within our margin framework.

Now moving to our financial outlook. For the fourth quarter, we expect revenue of $358 million to $366 million. Consistent with the slowing top line trends we've discussed throughout the year, we expect enterprise growth to be approximately flat and consumer to decelerate further given softer leading funnel indicators.

For adjusted EBITDA, we expect between $55 million and $59 million, achieving the 16% Q4 margin target we established in June. For the full year 2026, we're narrowing our as-reported revenue outlook within the prior range, now anticipating $1.229 billion to $1.237 billion, equivalent to $1.503 billion to $1.511 billion on a normalized basis. We remain confident in the approximately 14% full year adjusted EBITDA margin target, which we raised last quarter.

As a reminder, this as-reported margin is higher than the normalized equivalent and it also includes a meaningful purchase accounting benefit that will moderate in future years. Today's shareholder letter provides the corresponding normalized view and additional modeling considerations.

Before I close, I'd like to share a few directional points on 2027. Consistent with our usual cadence, we're not providing formal 2027 guidance today. We will provide it on our Q4 call after completing our annual planning process and executing the year-end enterprise renewal cycle. But with the integration progressing faster than planned and Project Helix beginning to roll out next year, we want to provide a directional baseline.

For modeling purposes, our starting assumption is that each segment's fourth quarter revenue trajectory is broadly representative of its normalized full year 2027 growth rate. For Enterprise, that implies approximately flat to modestly declining revenue. For consumer, a low double-digit year-over-year decline.

These assumptions account for potential revenue dis-synergies as overlapping customers combine over time and do not rely on any cross-sell contribution. Importantly, they also do not rely on a material revenue contribution from Project Helix. We expect Helix to begin rolling out in 2027. Our initial focus will be getting it into customers' hands, learning quickly from feedback and establishing clear adoption usage and commercial trends.

Said differently, we're not assuming a Helix-driven growth contribution before the evidence supports it. But we are investing and operating with the expectation that Helix will strengthen our offerings and expand the value we deliver to customers through a more integrated and capable skills platform. On profitability, we will remain highly disciplined and provide our full year 2027 adjusted EBITDA margin target next quarter after completing our annual planning process and getting better visibility into the top line.

There are a few important considerations to keep in mind. Transaction timing, along with the seasonality of adjusted EBITDA will place approximately 1 point of pressure on the reported annual compare. Additionally, our early revenue assumptions, together with the moderation of the approximately $10 million per quarter purchase accounting benefit included in 2026 adjusted EBITDA will be a headwind to adjusted EBITDA dollars in 2027.

We intend to offset those headwinds through the integration and synergy actions we've accelerated and the cost discipline we've demonstrated in 2026 and prior years. Coursera has expanded its full year adjusted EBITDA margin every year as a public company. We're proud of that track record, and we remain committed to strengthening the earnings potential of the business. We're funding Project Helix along with continued investments in our existing offerings within that margin framework.

Over time, unifying our platforms and a more efficient AI native architecture creates further opportunity for leverage. Finally, as we establish the compensation framework for the combined company, our objective is to significantly reduce stock-based compensation as a percentage of revenue and make meaningful progress lowering our annual dilution rate.

Future share repurchases are incremental to that work, not a substitute for addressing dilution structurally. Let me close with 3 priorities we have asked investors to measure us against. First, execution, protect customer continuity, complete the major integration actions with discipline and deliver the full synergy target by year-end. Second, financial quality, translate that execution into stronger margins, cash generation and disciplined management of dilution. Third and ultimately most important, product and commercial execution, turn the scale, data and investment capacity of the combined company into a more valuable platform for learners and customers through Project Helix.

I'm pleased with the speed of the integration and the accelerated delivery of synergies. The next phase is proving that this stronger foundation can translate into better customer outcomes and over time, more durable growth. Let's now open the call for questions.

Operator

[Operator Instructions] We will now take our first question from the line of Ryan MacDonald with Needham & Company.

分析師問答

Ryan MacDonald

Maybe to start, Greg, with just the commentary around the consumer segment and what you're seeing. Maybe just provide a bit more color in terms of the breakdown of maybe where you're seeing some of the softer demand as you think about relative to sort of the core subscription business within consumer relative to the one-off transactional in the Degrees business? And then sort of how you're thinking about sort of magnitude of investment within your framework and how you're allocating across those 3 components of consumer as we look to try to maybe get that business back in the right direction.

Gregory Hart

Thanks, Brian, for the question. So a couple of thoughts. First, I think -- and all businesses are seeing this in different ways in the consumer segment. You've got 2 things going on, one that's been happening for longer, one that is more recent. First one is ongoing changes in traffic patterns and SEO and discovery that all consumer businesses are seeing. Second one is worsening consumer confidence amid high inflation environments and what that does to discretionary spending.

And so I think we are subject to those things like other consumer businesses. For us, we continue to focus on the transition to an increasingly subscription business. You saw that in the shareholder letter and in Mike's comments as well. C+ and the Udemy Personal Plan are where both of those individual platforms are focused. We expect that to continue.

We will, in Helix, combine the offerings. We have not yet shared exactly what that will look like, but subscription will obviously be the primary focus of Helix, and we expect to continue to do that. Some of the changes that are happening within subscription on the Coursera side are a decrease in subscriptions to individual courses, what we refer to as terminal subscriptions where you subscribe to a specialization that you pay for until you complete the specialization, while Coursera Plus continues to grow.

We expect that, that trend will continue. It's been going on for many quarters. It's sort of been a structural change in the business for quite a long time, and we see that continuing to trend in that direction. It's -- C+ is a better value for customers, and we have higher LTV with C+ as well. So we are supportive of that trend. Overall, with Helix, we're really focused on investing in a platform that we believe will do a much better job of delivering for both our consumer and enterprise segments.

Ryan MacDonald

Appreciate the color there. And then maybe just touching on the enterprise. Obviously, Helix being the focus in the near term. I'm curious, as you think about with the early customers that you're implementing Helix at, can you just talk about some of the deliverables or sort of points you're focused on in terms of what -- how you're measuring success there and sort of delivering an ROI and then how you think about packaging that up in the sales motion within enterprise?

Because it feels like at least from the HR Tech Conference last week where we were at that L&D seems to be sort of a lower priority within the enterprise right now. Just curious how you're thinking about positioning Helix and sort of the ROI you're looking to generate there.

Gregory Hart

Well, one of the core principles of bringing Coursera and Udemy together was you have 2 very complementary platforms that are investing in a lot of the same things. And you have 2 content catalogs that have amazing instructors, both from universities and industry partners, but then obviously from working subject matter experts on the Udemy side.

By bringing them together, the goal was not just to create sort of a better content catalog, but to create an experience that delivers a much better job of what enterprises really want, which is not just a check box that an employee completed of course, but proves that they've gained the skill that the employer is trying to develop in that specific area, that team, et cetera.

And so we see our strongest engagement with our enterprise customers when it is an engagement that is much more focused on what we would call a functional buyer than just a general L&D benefit. Workforce transformation is an underinvested thing. You see stats on the level of investment that's going into AI infrastructure, and those are massive amounts. But then you see in parallel, people saying that, well, we're confident in our tech stack with AI, but we're not confident in our workforce's ability to actually leverage that successfully, and that's where the skill delivery is really important.

And so Helix is focused around that, a personalized and adaptive learning path that does a much better job of delivering skills in a verified way that enterprises can trust. And so a lot of the engagement with our design partners has been around that theme. They are very positive on the things that we're building in Helix and show higher willingness to pay in all the research that we've done on the things that we're building into Helix. We don't have anything to share on sort of pricing and packaging yet for Helix. There's a lot of work obviously going on around that. But until we actually roll that out, we'll be keeping that to ourselves.

Operator

Our next question comes from Brian Peterson with Raymond James.

Brian Peterson

So I wanted to ask about some of the dissynergies that you guys have mentioned. I know you guys have sized that. But I'd love to understand the timing of those and how those may have come in relative to your expectations? And is that something that's mostly expected in 2026? Or is that something that's contemplated in some of the initial commentary that you gave for 2027?

Michael Foley

Yes. Thanks, Brian. This is Mike. So the original kind of modeling for dissynergies really contemplated where customers are purchasers of both platforms today. And we did some modeling around that. So far, frankly, we just haven't seen any revenue dissynergies come through, even with renewals of those types of customers. We're sort of building in some expectation for that for next year in the sort of broad indications we gave you for 2027.

But frankly, it still remains kind of a theoretical calculation. We're not seeing it in practice from a revenue dis-synergy perspective. So nothing really in 2026. And again, sort of theoretical calculation for it in 2027, but it remains to be seen if they actually play out.

Brian Peterson

Got it. And just as a follow-up, I saw that the enterprise NRR ticked up a couple of hundred basis points sequentially. I'm sorry if I missed this. Could you talk about what drove the improvement this quarter?

Michael Foley

Sure. Yes. So it's up a couple of points. It's -- we don't consider it a trend as yet. I think there's a lot of room for improvement there clearly. Just a couple of piece of color that might help. If you look at the NRR that we have on our larger enterprise customers, it's significantly higher than the average of [ 93 ] in particular in North America is very considerably higher on the larger enterprise. And equally, it's quite a bit lower on the mid-market customers and geos outside of North America. So I think that speaks to the opportunity in NRR around execution through 2027 as we bring those teams together and start to execute as one team. But I wouldn't point to anything major for the uptick for this quarter as a trend just yet.

Operator

Our next question comes from Stephen Sheldon with William Blair.

Stephen Sheldon

Just one for me. I know there are a lot of moving pieces as we think about the setup for next year. But just on free cash flow generation, I'm just curious, I guess, how you're thinking about that with all the moving pieces. Should we still expect relatively healthy levels, I guess, of free cash flow generation or conversion from EBITDA? Maybe just help us think high level about free cash flow generation next year.

Michael Foley

Yes, sure. I'll give you some perspective on that. One of the differences, of course, between the EBITDA line and the free cash flow line has for the last couple of quarters and will continue to a lesser extent, the purchase price adjustment that we've talked about, which is really around the capitalized commissions. So that will continue to be a difference, which makes free cash flow lower than adjusted EBITDA. We also, of course, have some expected ongoing merger-related costs into at least the first half of next year.

And then the final piece I would point to is just just by nature of our business, which is relatively cash upfront on both enterprise and also to some degree in consumer. As we face the revenue headwinds that we are anticipating on the current platforms for next year, I do think there'll be some structural difference in there as we can see this quarter on the working capital side. So essentially, bridging between EBITDA and free cash flow, we expect free cash flow to be somewhat lower. We still do expect to be significantly cash generative next year, all of that said.

Operator

Our next question comes from Nafeesa Gupta with Bank of America Securities.

Nafeesa Gupta

So first, Greg, just curious to understand, you mentioned SEO and inflation as headwinds for consumer. Are you also seeing any early evidence to suggest that AI will ultimately reduce the amount of traditional skills where training is required? Like is there any difference in the popularity of your courses over time with AI? And then I have another question I'll ask later.

Gregory Hart

Yes. Great question, Nafeesa. So a couple of thoughts on that. One, I think there is less demand for some skills that had been very popular on Coursera and Udemy a few years ago. There is still demand for them. So like a good example would be there was a lot of interest on both platforms 3 or 4 years ago in Python -- in learning Python. That is still a popular topic on both platforms, but not as popular as it had been previously.

And so I think that a natural part of the evolution of AI and the technical skills that it can do is going to naturally change the types of skills that people learn. I am not as concerned in the long run about people learning those skills directly from AI. I mean there's a lot of evidence. I cited it in my scripted remarks that just using an AI chatbot actually is harmful for learning and mastery of the skill. And we think our differentiated content and certainly the pedagogy behind all of that is a real differentiator for us, particularly since it also comes from very trusted human educators coming from top universities, industry partners and subject matter experts on the Udemy side. So hopefully, that's helpful there.

Nafeesa Gupta

Yes. And the second question is, how should we think about magnitude and trajectory of Helix investment? Does the investment peak around the 1H launch next year? Or it will continue ramping as adoption grows?

Gregory Hart

Well, the investment is already baked into all of our operating -- our OpEx and all the metrics that you see. So we're not growing -- we're not expanding the team between now and when we roll it out.

And so the rollout, as Mike mentioned in the scripted remarks, will be on a measured basis. So in the first half of next year, we will launch a wide basis to our enterprise customers. Consumer will follow later. And then that -- particularly on the consumer side, but also on the enterprise side, that rollout will really be measured and pace based on the signals we're seeing in the metrics. But I want to make very, very clear that we are not anticipating increases in OpEx due to Helix. It is already contained within the business today.

Operator

Our next question comes from Brian Smilek with JPMorgan.

Unknown Analyst

This is Vignesh on for Brian. I guess across the different initiatives you guys laid out, Project Helix [indiscernible], Learn Vector and others, I guess, where are you most focused on today and see the biggest opportunities kind of to drive incremental growth over time?

Gregory Hart

Project Helix, absolutely. I mean that is the future of the business and the future of the 2 platforms. Right now, we have Coursera and Udemy as individual separate platforms. We have purposely not poured energy into bringing content from one to the other because we want to do that in a platform that we believe is better suited for what our customers need on both the consumer and enterprise side. That is Project Helix. And so that is the #1 priority for the business. We're continuing to invest in Coursera and Udemy, as we referenced in some of the remarks, both because we want to continue to improve the customer experience there, but also because we're building things that we will then leverage within Helix.

Operator

Our next question comes from Devin Au with KeyBanc Capital Markets.

Devin Au

I just want to follow up on the better enterprise performance in the quarter. I don't know if you already talked about this, but could you kind of speak to how the different segments, business versus campus versus government have kind of trended in the quarter? And can you also just kind of speak to how your conversation with the business customers and how are you thinking about maybe like R&D budget and spend intentions for next year?

Gregory Hart

So a couple of thoughts. One, CC, as we call it, Coursera for Campus, Coursera for Government, but both of those continue to be smaller parts of our enterprise business, but growing at nice healthy rates in both cases in the quarter. We believe that we have a very differentiated offering for each of those segments relative to others, particularly on the campus side. Enterprise is still the majority of our enterprise business, so CP as we would call it.

As Mike mentioned a little bit earlier, NRR up slightly sequentially from 91% in Q2 to 93% in Q2. Both of those figures are well below where we want them to be. We don't view that as a trend that we're banking on right now. We think that the right thing to get NRR to where it needs to be is really going to come as we roll out Helix and as that does a better job of addressing our enterprise customers' needs. L&D budgets, I think, still are pressured broadly, particularly as enterprises invest in -- but we also believe that enterprises, as they invest in AI need to make sure that their workforce investment keeps pace with the technology investment in the AI itself so that people are actually able to take advantage of that AI. And as Mike also mentioned, we do see some good signals.

Mike mentioned that our large enterprise NRR is substantially higher than that 3% -- it's actually very much a latter. So the larger the enterprise, the better the NRR. So the lowest NRR that we have is our SMB segment. And the highest NRR that we have is the large enterprise segment. So I think that speaks to the value that we provide to large complex organizations that are trying to do a lot. And then you also mentioned that we had really positive NRR for Americas large enterprise in particular. And so my viewpoint is we are still underpenetrated in pretty much every sector but there's some interesting positive signals in some of the largest potential opportunity spaces out there.

Devin Au

Great. I really appreciate the details there. Maybe just a quick follow-up on enterprise again on the initial guide for flat to down slightly that you haven't really seen any revenue dissynergies yet. It seems like that segment on a normalized basis in 3Q is still up a little bit. But I guess you're maybe taking in some conservatism? Or is it mostly just kind of expecting the R&D budget and the overall environment continue to maybe soften in '27?

Gregory Hart

Well, yes, on a normalized basis, the enterprise business was up 2% year-over-year. The from the NRR at 93%, overall growth at 2% churn from a logo perspective, Mike mentioned this in the scripted remarks, 90% of the logo churn that we saw in the quarter was from accounts with an ACV under $50,000. So we really see strong retention across our larger segments, which ties back to the NRR comment I made. I don't know that there's anything that we would say that would draw a trend overall, but we sort of expect the same type of performance that we saw in this quarter to continue on a forward-looking basis. We haven't -- to the point that Mike also made in the scripted remarks, we have not baked in any upside from Helix into our 2027 forecast.

Cam Carey

Thank you, Devin, and thank you, everyone, for joining today's call. A replay will be available shortly on our Investor Relations website. We appreciate you joining us.