Groupon Inc. (GRPN): The AI-Native Marketplace Turnaround

Groupon, incorporated in Delaware and headquartered in Chicago, operates a two-sided marketplace that connects consumers with local merchants, goods retailers, and travel sellers across North America and a dozen international markets. The economic engine is the Local business, which sells prepaid vouchers and experiences to consumers and collects a merchant take rate, with Goods and Travel functioning as higher-margin, lower-frequency complements. Roughly half of the revenue base is international, which introduces both currency translation noise and a genuine geographic optionality that North American peers lack. The model is asset-light by design: Groupon does not own inventory, does not employ the merchants, and carries minimal physical real estate, a structure that is why a company with under $500 million of revenue can still clear $75 million of adjusted EBITDA at full-year guidance.

The strategic context has shifted materially in the past two years. Under CEO Dusan Senkypl, who took the helm in early 2024 after a board process, the company abandoned the old growth-at-any-cost playbook and pivoted to an AI-native operating model branded Project Foundry, a full-stack rebuild of how Groupon acquires merchants, personalizes the consumer experience, and allocates marketing spend. That pivot explains both the restructuring plan approved in May of this year and the stock's re-rating from its $9.17 low to a $29.90 high before the most recent pullback. The strategic question is no longer whether Groupon is a viable business, which the cash flow profile answers affirmatively, but whether the AI rebuild converts a stabilized, flat-revenue franchise into a compounding one.

The competitive landscape is the quiet reason the turnaround matters. Groupon sits between two much larger forces: local-discovery platforms that own the search and social layers, and the loyalty and rewards ecosystems of national consumer brands. The McDonald's loyalty partnership disclosed in the latest quarter is best read as a proof point in that space, a channel that puts Groupon in front of a national audience it could not have reached through its own paid media. The strategic posture is therefore defensive on the North American Local base and offensive in international markets, where the new consumer platform is driving revenue growth of eight percent in the latest quarter and where supply expansion is being sequenced city by city.

The product stack is thinner than it looks. At the consumer layer sits the Groupon application and website, a discovery surface where personalization determines which deals a given user sees and in what order. At the merchant layer sits a self-serve onboarding flow, a pricing and inventory toolset, and a set of marketing campaigns that bundle a merchant's deal into Groupon's email, push, and paid channels. The intellectual property is not in any single feature but in the matching layer between the two, the data pipeline that learns which customer segment responds to which deal in which neighborhood. That matching layer is the asset Project Foundry is rebuilding, and it is the reason the company can talk credibly about an AI-native architecture rather than a bolted-on chatbot.

The new consumer platform, whose rollout nears completion, is the most concrete product event of the period. It replaces a decade-old codebase with a modern stack that improves conversion on nearly every surface, a change management has tied directly to the return of organic channels to growth in the latest quarter and their acceleration to double-digit growth in July. The mechanism matters: organic growth is the cheapest growth a marketplace can buy, because it is not paid for in media spend, so a platform that lifts organic conversion lifts the adjusted EBITDA margin line for line without adding to the marketing line. Personalization and trust-and-quality capabilities scaled in the same quarter, and management has disclosed removing hundreds of low-quality deals in a deliberate curation exercise that trades short-term unit volume for longer-term purchase frequency, the metric Senkypl describes as the north star of the new acquisition engine.

The moat is narrow but real, and it is worth naming precisely. It is not network scale in the way a payments network or a social graph has scale; Groupon's consumer base of 16.1 million active customers is a decent asset but not a defensible one. The durable elements are the merchant data graph, the purchase-frequency model that optimizes each marketing dollar against a customer's expected lifetime value rather than a seven-day return, and the international footprint that no pure North American competitor replicates. SumUp, the payments business in which Groupon holds a minority stake, is a non-core asset that management has said it would monetize opportunistically, an option that adds value to the balance sheet without changing the operating thesis. The moat is therefore a data moat layered on top of a marketing-execution moat, and both are only as strong as the frequency data that feeds them.

The second-quarter print, released in early August, is the anchor for this analysis. Global revenue came in at $124.7 million, down one percent year over year, while billings fell a similar one percent on a reported basis. North American Local revenue declined two percent, a drag management attributed to softness in Health, Beauty and Wellness, partially offset by strength in Things to Do. International Local revenue rose eight percent, and excluding the Giftcloud deconsolidation it rose nine percent, the clearest evidence that the new consumer platform is lifting conversion in markets outside North America. Unit sales fell seven percent year over year to 8.5 million, a decline management frames as customers purchasing fewer but higher-value local vouchers, a shift that is consistent with the curation strategy but is also the single line item that keeps the top line from growing despite the eight percent international revenue gain.

The profitability line is where the restructuring plan changes the picture. Adjusted EBITDA for the quarter came in at $14.8 million, at the high end of the guide, which is the number the market prices. The GAAP loss from continuing operations widened to $1.5 million from income of $20.6 million a year earlier, a swing that is entirely below the operating line. The driver is $3.2 million of restructuring charges recorded in the quarter, along with interest expense and non-operating items. That gap between the GAAP result and the adjusted result is the single most important read of the quarter, because the adjusted number is the one the market prices and the GAAP number is the one that reflects the one-time cost of the turnaround. Operating cash flow from continuing operations was $18.1 million and free cash flow was positive $15.0 million, both comfortably above the level needed to service the debt and fund the AI infrastructure spend. The cash balance of $226.3 million as of the quarter end is net of roughly $90 million of debt, leaving a net cash position that removes any near-term liquidity concern from the thesis.

The May restructuring plan is the named event that re-set the full-year cost base. The board approved a reduction of up to 400 positions globally, with the majority of the cuts expected to complete by the end of the third quarter. The company estimated total pre-tax charges in the range of $7.0 million to $13.0 million, of which the majority is cash severance. The payroll actions produce $20.0 million to $25.0 million in annualized cost savings, a run rate that is the real point of the plan. In the current fiscal year the company expects to reinvest up to half of those savings in marketing and AI infrastructure, so the net savings are closer to $5.0 million. The plan also came with an explicit raise in full-year adjusted EBITDA guidance from a prior low end to a new range of $75 million to $80 million. The fact that the company raised the target in the same quarter it took the charge is the single most important signal in the plan, because it shows management is confident the savings are real and the growth is coming. The sequencing of the charge and the guidance raise is what makes this a genuine cost event rather than a communication exercise, and that same-quarter pairing is the trigger for the stock's sharp run in the weeks after the announcement.

The full-year guide is the frame against which every quarterly print should be read. Revenue guidance of $513 million to $523 million implies roughly four to six percent growth at the low end and roughly ten percent growth in the second half at the high end, a trajectory that requires the July momentum, which management described as mid-single-digit billings growth accelerating to double-digit organic growth, to hold through the holiday quarter. Adjusted EBITDA of $75 million to $80 million on that revenue base is a margin of roughly fourteen to fifteen percent, which is the highest sustained margin in the company's history and is the number the market is currently pricing into the share price. Free cash flow guidance of at least $60 million for the full year is the balance-sheet number that keeps the debt story benign.

The forward question is whether the second-half acceleration is real or a single-quarter seasonal artifact. Management has given three specific signals: July billings growth accelerated to mid-single digits, organic channel revenue grew double digits in July, and managed channel revenue rose strong double digits on the customer data platform scaled the prior quarter. None of those three signals is a full quarter of data, which is the honest read of the forward evidence. The Q3 guide of $128 million to $130 million in revenue, up four to six percent, is the first number that tests whether the July print was a trend or a bump, and it lands in late October.

Project Foundry is the operating risk that dominates the forward story. The program is roughly four months old as of the second-quarter call, and management has described it as the most consequential work underway, but the full scope of the AI-native rebuild, including the merchant acquisition engine, the personalization layer, and the customer data platform, is not expected to complete its first phase until the end of 2027. That timeline matters because the company has already taken the restructuring charge and the full-year guidance raise on the promise that the program produces the savings and the growth. If the 2027 completion slips or the savings do not land at the guided rate, the adjusted EBITDA margin that the market is pricing, roughly fourteen to fifteen percent, has to come from somewhere else, and the only somewhere else is further cost cuts that could damage the merchant and customer experience.

The appointment of Aditya Rajkumar as Chief Operating Officer, effective August 3 of this year, is the management event that matters most for execution. Rajkumar oversees the marketplace and merchant operations, which is precisely the function that has been underperforming, with North American new merchant acquisition running below expectations and the Health, Beauty and Wellness category soft. The CEO has publicly acknowledged that the North American Local supply engine has been running behind expectations and named Rajkumar, along with a senior marketing executive, as the pair tasked with driving that engine forward. The thesis here is that the turnaround has a data and marketing layer that is working and a supply layer that is not, and the COO hire is the mechanism by which management intends to fix the supply layer. The risk is that a new COO in a mid-restructuring company does not produce visible supply improvement for at least two quarters, which pushes the real test of the supply fix into the fourth quarter or later.

The second-half guide implies roughly six percent revenue growth at the low end and roughly ten percent at the high end, a spread that is wide enough to absorb a flat third quarter and still deliver the full-year number if the holiday quarter is strong. That is the structural asymmetry that makes the forward story investable: the company has de-risked the downside with the restructuring and the cash balance, and the upside is concentrated in a single seasonal quarter that the new platform is positioned to capture. The execution risk is not that the company fails, it is that the timing of the supply fix and the personalization-driven frequency gains do not line up with the holiday quarter, which would push the margin inflection into fiscal 2027 and leave the current share price without a near-term catalyst to validate it.

The largest risk is that the North American Local base, which is roughly half of the revenue, does not return to growth in the second half. The category softness in Health, Beauty and Wellness is not a demand problem in the abstract; it is a supply-freshness problem that the company has acknowledged, and the fix depends on a new COO who has been on the job for fewer than six weeks. If the third-quarter print shows North American Local revenue still negative, the second-half growth narrative loses its main pillar and the stock, which has re-rated roughly three times from its low, has little support below the cash-and-EBITDA floor.

The second risk is the debt and equity structure. Total debt of $312 million against a cash balance of $226 million leaves a modest net cash position, but the stockholders equity line is negative as of the most recent balance sheet. The negative equity is a reflection of years of accumulated losses and share-based compensation, and it is roughly $72 million in magnitude. The negative equity is not a liquidity event, the company is generating positive free cash flow and has no near-term maturity pressure, but it does mean that a sustained revenue miss that erodes the adjusted EBITDA guide would have to be managed against a balance sheet that has no cushion. The two debt instruments that define the refinancing timeline are the 2027 Notes and the 2030 Notes, and any credit spread widening tied to a growth disappointment would show up as a higher cost of the debt already on the books.

The third risk is the AI strategy itself, which is simultaneously the thesis and the largest single point of failure. Management has disclosed that Project Foundry is evaluating additional cost-reduction and automation actions that are subject to board approval and are expected to complete by the end of 2027. That second wave of automation is not in the current guidance, which means it is upside optionality if it works and a further restructuring event if it is needed. The consumer-facing AI personalization layer is the part that drives the frequency metric, and if the purchase-frequency model does not produce the repeat-purchase improvement that the marketing ROI model assumes, the marketing spend that management plans to scale has a lower return than the model suggests, which compresses the adjusted EBITDA margin from the top rather than the bottom.

The fourth risk is international, which is the part of the business that is actually growing and the part that carries the most structural uncertainty. The eight percent international revenue growth in the latest quarter is real, but it is also exposed to currency translation, to local merchant supply constraints that are harder to fix remotely than the North American supply engine, and to the regulatory environment in a dozen jurisdictions. A meaningful currency headwind in the second half, or a supply constraint in one of the larger international markets, would shave the top end of the full-year revenue guide without any change to the North American base. The downside scenario is therefore a stock that holds the cash floor, misses the second-half growth inflection, and drifts back toward the range where it traded in the first half of the year, a level that still represents a meaningful premium to the low the stock printed in the prior year.

The valuation framework starts from the cash flow, not the multiple. The company is guiding to $60 million or more in free cash flow for the full year, a number that sets the floor for the balance sheet. The adjusted EBITDA guide of $75 million to $80 million on a revenue base above half a billion is a margin profile that no pure local-discovery peer matches. The share price and the share count put the market capitalization at roughly $777 million. Against net cash of roughly $90 million, the enterprise value is approximately $687 million. That implies an enterprise value to full-year adjusted EBITDA multiple of roughly eight to nine times at the midpoint of the EBITDA guide. That multiple is the anchor for the whole valuation, and it is the number the rest of this section is built around. It is not cheap against a growing software peer, but it is reasonable against a company that is delivering that EBITDA and generating the cash flow to service its debt.

The bear case prices the company at the cash floor. If the second half does not accelerate and the full-year adjusted EBITDA lands at the low end of the guide, the market re-prices the EBITDA multiple down to roughly six times on $75 million. The resulting enterprise value is roughly $450 million, which works out to a share price in the single digits after adding back the net cash. The result sits close to the $9.17 low and represents a downside of roughly fifty percent from the current price. The bear case is not a liquidation scenario; it is a scenario in which the restructuring savings land but the growth does not, and the market assigns the multiple of a flat-revenue business rather than a turnaround business.

The base case is the current guide held and the EBITDA multiple held at roughly eight times. Full-year adjusted EBITDA of $77 million at eight times is an enterprise value of roughly $616 million. Add the net cash of roughly $90 million and the equity value lands at roughly $706 million. That works out to roughly $17 per share, only modestly below the current price, which tells the reader that the stock is already pricing the base case. The base case is the case in which the restructuring works, the international growth holds, and the North American supply fix is real but slow, and it is the case that leaves the stock roughly where it is.

The bull case requires the second half to land at the high end of the guide and the market to re-rate the EBITDA multiple to roughly eleven times on the strength of the full-year EBITDA of $80 million. Eleven times on that EBITDA is an enterprise value of roughly $880 million. Add the net cash and the equity value lands at roughly $970 million, or roughly $24 per share. That is the level the stock touched in July before the pullback, and it is the number that the current thesis is built to reach. The bull case is therefore not a re-rating to software-multiple territory; it is a re-rating from eight times EBITDA to eleven times EBITDA, which is a meaningful but not outsized move and one that is supported by the international growth and the margin trajectory. The framework conclusion is that the stock is priced at the base case, the downside to the cash floor is roughly fifty percent, and the upside to the bull case is roughly twenty-five percent, which is an asymmetric profile that favors the stock only if the reader believes the second-half acceleration is real. The move from eight times to eleven times EBITDA is significant but within the range the market has paid for comparable consumer turnaround stories, and it is the multiple shift that the current thesis is built to capture.

The quarter revealed that Groupon's turnaround is real on the cost side and unproven on the supply side. The adjusted EBITDA of $14.8 million and the positive free cash flow of $15.0 million confirm that the restructuring is working and that the business is cash-generative at scale, but the one percent revenue decline and the seven percent unit sales drop confirm that the marketplace has not yet found the growth engine that the stock re-rating assumes. The honest characterization is a company that has fixed its P and L statement and is now working on its top line, and the top line is the part that the market is paying for.

The counterargument deserves direct statement. A skeptical reader can argue that the current price already reflects the turnaround, that the EBITDA multiple of roughly eight to nine times is full for a company with flat revenue, and that the AI-native narrative is a story layer on top of a business whose fundamental constraint is a North American local supply base that has been shrinking. That argument has merit. The stock tripled from its low on the strength of the restructuring announcement and the Project Foundry narrative, and a flat third quarter would remove the near-term catalyst that justifies the premium. The case against the stock is not that it is broken; it is that it is expensive relative to the evidence, and the evidence so far is a cost fix without a confirmed growth inflection.

The monitoring items that define whether the thesis holds are the third-quarter North American Local revenue growth rate, the full-year adjusted EBITDA landing within the guide, the new merchant productivity improvement in North America, the international revenue growth rate excluding Giftcloud, and the completion timeline for the Project Foundry first phase. Each of those five items is a binary signal that the reader can check against the next two quarterly prints, and none of them requires a new disclosure to verify. The stock is a second-half event, and the next two prints are the only data that matters for the judgment rendered here. If the third quarter shows North American Local returning to positive growth and the international base holding above five percent, the stock has a clear path to the bull case. If it does not, the cash floor is the only support that matters, and the stock should be treated as a value position rather than a growth position.