Freelancer Limited (FLN) Earnings Call Transcript & Summary

Hello, and welcome to the Freelancer Limited First Half of 2026 Investor Update. My name is Matt Barrie. I'm the Chief Executive and Chairman. With me today, I have Dylan Carter, the Chief Financial Officer; and Andrew Bateman, the Vice President of Products. And as always, you may direct your questions to either myself or to anyone in the room at the end of the call. We're building the Amazon of services at Freelancer. We're in the field of labor payments in freight. We have 3 properties, all of which are market leaders in their own respects. Freelancer is the largest global crowdsourcing marketplace by a number of users with 90 million professionals in the marketplace. We do jobs from $10 to $10 million, and everything from a simple and mainstream is building websites and so forth right up to complex -- solving complex problems to the U.S. government in our Moonshots Innovation Challenge program. Escrow.com is the world's largest online escrow company. We have done USD 8.3 billion of transactions through that to date, and it's for buying and selling any assets or product service online in an untrustworthy environment. And also Loadshift, which is Australia's #1 heavy haulage marketplace, it's going global, and we've done over 1 billion kilometers of freight to date. So we do midweek on -- sometimes on a Wednesday, somewhere around 300,000 to 400,000 kilometers of freight, which is the distance from the Earth to Moon trucking freight. All businesses are solving problems in trillion-dollar markets. They're all broad horizontal businesses that are small businesses and large businesses alike need as well as consumers. And in terms of our business model, it's straightforward, and Freelancer will make 3% from the clients, 10% from the freelancers in every job. And as I said before, we scale up from $10 job to $10 million jobs. We have optional upgrades, memberships and building upgrades and product upgrades. Loadshift has identical business model to Freelancer, and it actually runs in the same technology stack, which is great in terms of synergies. And Escrow, we make a blended average of around 1.3%, 1.4%, 1.5% on the transactions that go through. And we do transactions from $10,000 to $100 million. Our Board is quite illustrious from an operating perspective. I won't go into detail on the bios, but it's myself, Darren Williams, Patrick Grove and Craig Scroggie. And we have a quite diverse management team around the world for those businesses. In terms of the update, our gross payment volume was up 30.9% in Australian dollars to AUD 574.6 million. And U.S. dollars was actually higher at $403.8 million, 45.5%. Our revenue was down 12.1% in Australian dollars at AUD 23.9 million. In U.S. dollars, it was down 2.6%, effectively flat at $16.8 million. The 3 things we're focused on at this point is we've got new financial and product leadership in the group level and at the Freelancer product level and for Dylan Carter, this is his first financial results call with me today, Andrew Bateman, who is in charge now of product for Freelancer. Escrow and Loadshift continue to grow in scale, and execution and cost discipline will drive operating leverage. And in the future, we plan to accelerate that with some breakthroughs that we've made in terms of our AI capabilities in the business. And just 2 little minor points I'll note, while I'm here is that AI now handles 100% of Tier 1 support interactions with an 11% containment rate, which means that 11% of support interaction stays purely within the AI system. And engineering, some of our teams have experienced up to a 40% uplift in releases, thanks to [indiscernible]. On to the segment performance in terms of Freelancer. We had a mixed quarter half in terms of Escrow worked really well. Loadshift went okay, and then Freelancer, we are wrangling product. We've had a number of challenges in Freelancer. So we're confident that we understand what the issues are and we're kind of on top of them and we're solving them, and I'll go through in a bit of detail. As I've mentioned in previous quarters, we've had a parabolic rise in scraping activity hitting our site and remediation work that we took in order to combat that led to some specialty indexation. Also, stability has been mostly rectified from this, and the traffic is above pre-indexation levels and projects are lifting, and we do expect financials to follow that, but I will show you some more detail about that. In the previous quarter, I also talked about how we put in security controls to enhance and reduce our fraud risk, but it also churned some returning customers. And I'll show that we've tuned that away, I think, fairly effectively at this point. And then in the middle of all of that, we've had a failed gateway migration in the first quarter, which has revealed some long-standing issues in terms of acceptance with some gateways, and we have quite a robust plan for rectifying that and actually proving that better than it has been performing for many years. And in fact, I think this has been actually a quite long-standing drag in the business has been our payments acceptance. Now I want to talk a little bit about marketplace dynamics before I get into those issues. Overall, the marketplace dynamics have been fairly static compared to last year. As in the composition of jobs in the marketplace has remained pretty much the same in the first half as in the full year of 2025. Websites, IT and software comprise about 31% of our jobs. That's indeed a little bit higher in the first half, a couple of points. Traditionally, websites IT and software has been 30%, 31%. In recent periods, it dropped a little bit below design media architecture, but it's now regained its top spot. Design media architecture is at 27% and then a long tail, which has remained fairly consistent. The only thing I would probably note is that the data entry and admin is down a couple of percent. AI jobs now represent about 7.4% of GMV measured across the entire skill base. And the composition of jobs from client countries has remained fairly static as well. The United States has actually grown in terms of share slightly. And Germany has gone down a little bit slightly. We are going to, in the second half, improve the experience for non-English experience of the site. We've got some enhancements to the product. We can talk about a bit later on, which will improve non-English. But the U.S. increased in its composition slightly over the first half. The average project value also increased. It does reflect higher value and more sophisticated work and some of our ad targeting. It also reflects Loadshift because it includes all jobs that run on the Freelancer stack. And so that's currently at USD 453 and with about 4,000 job categories to date and climbing in the Freelancer marketplace. So let me go to the challenges, right? It was a challenging half for Freelancer. We've had a 1,300% increase from scrapers over the year in terms of total traffic to the business. And that is a lot of traffic because we are quite a substantial site. Obviously, what the AI scrapers are doing is trying to take content for their training data sets. We're quite a valuable data set because we're the world of work, and we have the largest, obviously, number of freelancers in the world on our platform, nearly 90 million. That introduced over the course of last year. You might remember in previous quarterly updates, we had some stability issues. That has been mostly rectified pushing a lot of this content to the edge. There's still a little bit to go, but pretty much stability is significantly better than it was in the second half of last year, which kind of where it really all kicked off and affected our numbers. And this, unfortunately, has led to a little bit of indexation, as I flagged in previous quarters, in that you've got everyone from Anthropic to the Chinese scrapers to Google and so forth, constantly hitting our site. The -- unfortunately, they do not obey rate limiting directives. And so -- and the Google scrapper comes from the same IP address range as the SERPs indexation. So when you try and attempt to rate limit Google, it actually affects your SEO indexation. What has happened as a result of that is we did lose some SEO traffic through trying to remediate that. The traffic is actually now back and above what it was before that indexation event, but it is -- the traffic composition is slightly different. So projects continue to rise. We continue to work on this. We still have some things to go in terms of improvements. But I think this one is mostly handled and we're kind of seeing how that plays out. I will note that if you check Google trends, and I think it's probably important given the fact there's a whole bunch of narrative around AI out there and AI taking people's jobs, et cetera, and some pretty ridiculous claims from some of the AI companies because they need to justify the trillion dollar valuations. And so they do that by sometimes making claims about the world of work and taking everyone's jobs. I still don't know anyone to this day that's lost their job due to AI directly. But if you will look in the Google trends, you can independently verify this yourself. You can search for Freelancer, freelancer.com across the U.S. and Canada and the U.K., et cetera, and you'll see that the interest in Google is actually climbing for this work probably because we're entering into unstable times in terms of the global economy. The second challenge we had, which I also talked about in the last quarter was security controls. The second half of last year, we implemented enhanced security controls to reduce our risk of fraud. In the process of doing so, I did flag that we did hit some returning customers because the controls were a little bit over eager in terms of blocking the fraud. We stopped malicious actors in their tracks as a result of these controls, and we've been progressively tuning those controls to reduce the volume that has been experienced, but we did churn some clients as a result of that. And you can see here that the controls have come down quite dramatically in this period where it was quite low, we're be testing various other methods of reducing fraud. The issue in the first half that challenged Freelancer was we had one of our gateways fail on us in the first half. I will point out that we have many different gateways on Freelancer for processing incoming payments. We have a specific U.S. gateway, which is on the left-hand side, that's very healthy. For our pristine U.S. traffic, we send it to a particular gateway. We're very careful about what we send there. And you can see there that the recurring volume from that gateway is climbing, which is what you would expect from a healthy business where repeat customers come back and pay more, et cetera, over time as they build the usage of the product. The right-hand side shows what happened in one of the gateways, and it was unfortunate because it was the same time that we were kind of dealing with some of these other issues, where there was a merchant identified consolidation that was happening unbeknownst to us. And it caused some of our users on certain merchant IDs to be unable to recur their transactions. So they had a card in the system. They would hit that card when they would use the site and then those cards started failing, right? We detected that around May 13 of this year, and then we started peeling that volume off and putting it to more healthy gateways, but it has affected the financials for this half, and I'm not happy about it. And -- but we -- you can see here that by really the -- around June, we were taking volume off that gateway and shipping it to other gateways. We have a pretty robust plan in terms of how we're going to enhance our payments moving forward. In fact, what this has actually shown to us is when we dug into the aggregated numbers, we actually found that there were some long-standing issues in a couple of the gateways. We now have -- we have a very good health in our U.S. gateway, in our Australian gateway, in our Apple Pay gateway for taking payments from iOS. We're in the process of deploying Google Pay, which will be live in the next 2 weeks, which will be good for Android. We have an Indian local acquiring system, which is up and running and in test and currently in commission, which will happen over the course of the next quarter, which will take Indian traffic off. And then we have a number of other gateways around the world, which are basically in a reasonable condition or moving to other gateways for local acquiring payment methods. But I believe that after we did some pretty detailed analysis on what was going on here, we think this has actually been a fairly big drag on the business for some time. And I think the good thing is we now know how to look through this. And I think that we'll end up with a much, much, much stronger payment system, and that drag will be taken away from the business once that's done. Under the new leadership, we have got an enhanced product direction. This is just an example of what's currently being AB tested on the site at the moment in terms of a new brand direction. We're building out more specific verticals in areas like robotics, cosplay, electric vehicles, influencer marketing and so forth to really solidify our presence in verticals. I feel like eBay has positioned themselves over time as in cars, collectibles, cameras and so forth. We're doing the same in services. We've had some pretty decent wins in terms of A/B testing various improvements to help surface and match effective freelancers for clients in a very unstructured environment, which I think has been quite a good step forward. And of course, as before, we solve high-value problems in a range of different areas, as I mentioned, AI jobs, it's around 7.4% of our GMV at the moment. We've got Moonshots Innovation Challenges at the very high end for U.S. government and other corporates, which is now live. In fact, we're completing a pretty interesting innovation challenge right now on the distribution of particles from explosions for a big mining company. We do work with field services. I talked about before, with Hewlett-Packard, we've done 90,000 repairs for laptop service and printers, and we're doing some government work with getting people off the dole and getting into the world of work online, for example, with Bahrain. With Escrow.com, we've now done $8.3 billion worth of transactions secured. You can see that a great quarter. We think we will continue to have some good quarters from here. As I'll talk about in a second, we've broken into a new segment with Escrow, which I think is proving to be quite strong even this morning, we had a 7-digit transaction come in for that. And that area is wholesale electronics. So we had healthy volumes across all the major segments, domain names, IP addresses and wholesale electronics had a quite marquee half year, and I think will continue to be strong and become our #2 segment. This is helping secure the trade of things like servers and CPUs and data center equipment and GPUs and so forth. Our volume in Australian dollars up 39% to $517.6 million, in U.S. dollars, which is currency neutral because the vast majority of volume in escrow is in U.S. dollars, up 54.6% to $363.7 million. In fact, I've seen the GMV peak at close to 80% here and there. So we're doing very strongly with Escrow. This is the fourth highest GMV ever. We've deployed a number of new banking integrations, including real-time payments, instant matching of funds to make the transactions speed up. broken into that new market. And we have a new user interface rolling out in the third quarter. We already have some of our customers in that, which will also make it as a byproduct of that a lot faster for us to develop product in combination with our AI software engineering technology activity, where I said before, teams are experiencing up to a 40% improvement in terms of productivity, in terms of releases as a result of using that technology, which feeds directly into our code base. It's our seventh year of profitability, consecutively for Escrow. In terms of Loadshift, we hit the milestone of over 1 billion kilometers of freight posted. We're starting to see a lift in the numbers. And I think this will be a great year for Loadshift because we've got quite a number of product improvements, which are coming out very strongly. The revenue was -- the GMV was up 9.6% to $14.6 million. It was an all-time record for revenue and an all-time record for GMV. As I mentioned in previous quarters last year, we hit profitability in this business. It was a modest profit, but it has across that inflection point. We had record media hits. You may have seen this all over the news from the fuel crisis in Australia. This is an Australian-only business, but it is going global. We're in negotiations with a partner about doing that as we speak. And that fuel cost might be back again. As I noticed today, Saudi Arabia cut its fuel production by about 50% after some refiners were hit. So we might be all over the news again as a result of that because we have established ourselves as an authority in diesel in this country. We've got new product launches, including a new onboarding design for carriers, which increased the bid liquidity. We've got effectively a page now for tracking all the trucks and having a new view into the Loadshift marketplace where you can basically find trucks directly in certain locations and see what the capacity is and book them directly from a map view. We've got a bridge into the public telephone network, which is live as we speak and is currently being tested in the estimates-only field, where we've seen a 5x increase in volume from doing so. This will reduce off-siding and increase security and tracking. And we think it will lift award rate, which will be a leg up again in revenue for this business. And as I said before, progressing plans to expand into other jurisdictions. I'll now pass over to Dylan Carter, who will go through the FY '26 financial performance.

Thank you, Matt. I'll now explain how the operating performance translated into our group financial results. I'll focus on three areas: how transaction growth converted into gross profit, how we managed the cost base, and how the result flowed through to cash and balance sheet capacity. The key dynamic this half was that strong growth in Escrow.com and continued growth in Loadshift were offset by low activity in Freelancer. I'll begin with the headline metrics. Overall, it was a mixed result. Group GMV increased by almost 31% to AUD 574.6 million, or USD 403.8 million. Driven by continued strength in Escrow and ongoing growth in Loadshift. Gross revenue declined, however, 12.1%, or to $23.9 million, or 2.1% in U.S. dollar terms. Freelancer operates a blended take rate of about 13% plus additional add-on fees, compared with Loadshift of about 12.5% and 1.35% for Escrow. Strong growth in the lower take rate Escrow transactions, therefore, did not fully offset the decline in the higher margin Freelancer business. EBITDA was $1.1 million, while operating loss was approximately $0.3 million, effectively year-round breakeven despite the headwinds experienced during the period. So while earnings were below where we wanted them to be, I think the important takeaway is this remains a business with very strong underlying transaction activity, a diversified earnings base and a clear plan to improve execution from here. Now let's take a look at each of those businesses individually. Turning first to the marketplace business. Combined GMV was $57 million for the half, down 14.3%, with lower Freelancer activity, partly offset by continued growth in Loadshift. As Matt outlined earlier, Freelancer was affected by several operational headwinds during the half. Additional conversion friction following strengthening of integrity controls, disruption to the organic channel from search engine re-indexation, and lower payment acceptance following gateway consolidation and other gateway issues. Those issues reduced project activity and conversion. Freelancer GMV declined 20.3% to $42.4 million, while revenue declined 23.2% to $14.9 million, all in Australian dollar terms. On a constant currency basis, the declines were more moderate at 11.7% for GMV and 15% for revenue, with stronger Australian dollar amplifying the reported result. Loadshift continues to grow, with GMV up 9.6% to $14.6 million and revenue increasing 12.2% to $1.8 million. While the take rate remains stable around 12.5%. Across the two marketplaces, the blended overall margin reduced from 31.5% to 29.3%, which includes add-on fees and the take rates of the individual businesses. Add-on fees include things like project upgrades and other services that customers can purchase. The H2 focus is therefore very practical for the Marketplace business, removing those operational frictions identified during the half, restoring the SEO and organic traffic, and also payment conversion. Now turning to the Escrow business. Escrow was the strongest financial performer across the group during the half. GMV increased 39% to $517.6 million. On a constant currency basis, the growth was higher at 54.6%, which is the functional currency of this business, with the stronger Australian dollar reducing the reported result relative to the underlying U.S. dollar transaction activity. Revenue increased 15.1% to $7 million in Australian dollar terms, or 27.7% on a constant currency basis. Gross profit increased 21.8% to $6.5 million, while gross margin improved by about 2.7 basis points due to a more fixed cost base within the cost of sales profile of this business. The difference between GMV and revenue growth reflects transaction mix. Escrow processed a greater contribution from verticals and transaction sizes with lower percentage fees. As a result, the blended take rate reduced from 1.63% to 1.35%. Despite that reduction, the scale of the volume growth delivered strong incremental gross profit and improved operating leverage. We would expect that the take rate to return to its original level, excluding any significant one-off transactions. Growth was broad based across Escrow's core verticals, as Matt mentioned earlier. Domain GMV increased 56%, while IPv4 GMV increased 17%, and we had a standout quarter for wholesale verticals. This reflects continued demand for trusted infrastructure in complex, high-value online transactions. The result really demonstrates the quality of the Escrow's business model, strong transaction growth, a relatively stable cost base, and increasing gross profit contribution as the platform scales. Turning to the group's overall earnings result. Revenue, as mentioned, declined 12.1% to $23.9 million, reflecting the lower contribution from Freelancer. Gross profit declined 10.4% to $20.9 million, while gross margin improved 2.1 basis points -- 2.1 percentage points to 87.4%, supported by the change in business mix, which results in a lower cost of sales. Obviously, the US dollar is the main functional currency for the group, representing over 70%. And so therefore the US dollar growth rates on the earlier slides are a better reflection of the drivers of this business. Against that revenue backdrop, however, in Australian dollar terms, we maintained discipline across the cost base. Employee expenses were 4.4% lower, and hosting costs also declined 5%, reflecting active workforce management over cost, operational efficiencies, and supplier negotiations. These savings were partly offset by higher administrative costs, including increases in AI tooling internally, token usage in our products, and additionally, further compliance costs within the Escrow business. We do see further cost reductions available within this business, and we're working through the details. Overall, operating expenses increased by only 1.8%, despite those additional investments that Matt mentioned that we've been doing over the period. The group generated EBITDA of $1.1 million and recorded a near break-even operating result, with an operating loss of $0.3 million. This compares with operating profit in the prior period of AUD 1.8 million in Australian dollar terms, with the reduction principally reflecting the lower gross profit contribution for Freelancer. Below the operating result, the group did recognize a $2.4 million non-cash impairment of non-core digital assets. This impairment reflects our strategic focus back to the core business and limited current cash generation of those non-core digital assets. Importantly, the carrying value of our core Freelancer, Escrow, and Loadshift businesses remain fully supported by our business plans. The statutory result also booked a $0.9 million foreign exchange gain, arising principally from the translation of foreign-denominated liabilities into Australian dollars. After these items, the group recorded a net loss after tax of $2.1 million. The key takeaway is that we have demonstrated tangible cost discipline, particularly across employee and hosting expenses. The focus from here for the group is to combine that discipline with improved revenue performance and stronger operating leverage, as Matt went through early on. Turning to cash flow, the group closed the half with AUD 17.9 million of cash and cash equivalents, compared with $22.9 million at the beginning of the period. The underlying business generated approximately $0.7 million of positive operating cash flows before working capital. This was offset by a $1.3 million working capital outflow, primarily reflecting the timing of user obligation movements and higher prepayments made during the period. The group also made $2.8 million of lease payments and approximately $0.1 million of other capital expenditure during the period. Foreign exchange movements reduced reported cash by approximately $0.8 million, and a further $0.8 million was used to acquire additional minority interest in Loadshift. The cash movement therefore reflects a combination of the lower operating contribution during the half, working capital timing, lease commitments, and the acquisition of those minority interests in Loadshift. Cash generation remains a key management priority. The focus is on restoring earnings contribution from Freelancer, maintaining the growth and operating leverage of Escrow, and continuing to manage investment and expenditure with discipline. Finally, in terms of the balance sheet, the group held AUD 17.9 million, as I mentioned, at 30 June and remains debt-free. A substantial component of the group's working capital position reflects user obligations held in our platforms. Managing those obligations requires strong treasury discipline and appropriate liquidity across currencies. From here, our capital framework has really three core priorities. First, maintaining sufficient liquidity to meet user obligations as they fall due. Second, restoring consistent positive free cash flow through improved operating performance and disciplined expenditure. Third, allocating capital selectively to opportunities where we see the most attractive risk-adjusted returns. The balance sheet gives us flexibility, but that flexibility must be carefully managed. So our focus is not just to simply deploy capital. It is to improve the return generated on the assets that we already have in the group. So stepping back from the financial result, group transaction activity remains very strong, led by Escrow and Loadshift. We have demonstrated tangible discipline in employee and hosting costs. There's actions underway for Freelancer, which are focused on improving revenue conversion and earnings. With that, I'll hand back to Matt for the H2 priorities.

Okay, thank you. The main thing we've got to do is wrangle the Freelancer products. In the first half of last year, I think the business was in a fairly decent position. We were consistently about two-thirds of the way towards my goal of consistent operating profitability per month. In the second half of last year, however, we started taking a couple of hits. It started with the scrapers kicking off at a tremendous rate, and I will say 5% reduction in hosting costs with a 1,300% increase in traffic is pretty decent management by the engineering team. We've been absolutely smashed. There's actually articles in The Financial Times about Anthropic and what the scraping they do to us, et cetera, and so forth. So we had the scrapers hit us and then we had enhanced fraud controls that we deliberately put in place to protect our users, that unfortunately resulted in blocking of some users. While we're trying to debug all of that, we had a gateway failure issue where one of the gateways decided they wanted to consolidate a number of our merchant IDs. That consolidation, the reason why it was problematic was we're not a SaaS business, so we're not every month having our users make a recurring charge. We have intermittent usage where users kind of go away and come back and go away and come back. That consolidation was done without senior management's knowledge, and that resulted in quite a number of our returning users unable to use their cards. So it took a while for us to be able to figure out what exactly was going on. As part of that, we have unearthed that there has been longstanding payments issues. Some of the gateways over the last 10 years have not been performing well, to be honest. Some gateways are performing well. Our main U.S. gateway for credit cards, as I said before, is a very, very healthy graph. You can see there that you've got initial payments coming in from users, and you've got recurring payments as they kind of repeatedly tap their card. That's the left-hand side graph. Which shows you that despite the rhetoric out in the market, AI is not out there taking jobs and everyone's going to be unemployed and on universal basic income. The users, if they can make payments, they come back and they keep using the product. The usage may be intermittent, but you do have a healthy growing customer base. That's what our main U.S. gateway looks like. That's also what our Australian gateway looks like. That's what our Apple Pay gateway looks like. But when you look at this other gateway, which was a rest of world gateway, you can see that there was a break right over the holiday period. It's an unfortunate time to have an issue like that because it's over the holiday period. So while you're monitoring the stats, you do expect a drop in volume over the Christmas period. And when that volume doesn't come back, you're kind of running around trying to figure out what exactly is the issue. So it's very unfortunate this happened right at the same time as we were excessively blocking our users from our fraud controls as well as dealing with the scrapers, and so on. So this is not good. We're not happy about it. It has meant that we've got now a far more robust plan for how we take payments in the future. In fact, it's actually behind me on the board in terms of, we're splitting up basically all the incoming payments by country, by currency, and ensuring that we have clean, pristine, well-managed gateways handling those payments. Part of the problem we have taking payments as a business is that because we've got so many customers from around the world, we get credit cards and other things from all over the place, right? So unlike a traditional e-commerce business that might be based in Australia where you just see predominantly Australian-issued cards, we get cards from Egypt, Nepal, Bangladesh, U.S. In the U.S., we get food stamp cards attempted. We get German cards. We get Japanese cards. We get cards from everywhere. That is a complex thing to manage because not every gateway will like every card type. Some of our gateways will not accept certain BIN numbers. Some gateways will accept the BIN numbers, et cetera. And so you have quite a complex environment for managing what payments you send where. As we have noticed, if you send cards that are benign to a gateway that doesn't accept them, and those cards are from valid users wanting to do valid payments, the gateways will back off the overall acceptance rate because they see a mix of traffic that while it is valid traffic, it's not something they want to accept, and they'll back off all cards. We think we've had a long-standing problem actually in two of our gateways for some period of time, including the one that broke. That was surfaced quite explicitly after it did break and after we did look into it into some detail. We do think that coming through this, where we have the traffic split off into specific gateways by country, specific gateways by device, for example, Google Pay for the Android devices, et cetera, that we will have much better acceptance. The acceptance we should be getting is north of 85% across all gateways. We're seeing that in our main U.S. gateway. We're seeing that with our Australian gateway. But some of these rest of world gateways have traditionally been in the 30% range. They've been in the 30% range not because people don't have money or because there's fraud on those gateways, but because the acceptance of certain cards issued from certain countries has been pretty poor, and those cards hitting those gateways has led to a generalized back off from those gateways' acceptances. So -- while this is unfortunate and Freelancer has taken a beating over the last 12 months, it is fixable and we know how to fix it. We know how to measure it, the fixes, and we can do it methodically. That is our primary focus is to -- and we're right in the thick of it doing daily stand-ups, turning this all around. We're seeing some improvements already in some of the redirected traffic and the acceptance rates of those gateways. As I've said before, we have an Indian local acquiring method which has gone through the e-mandate integration, which is quite a sophisticated integration you need to do. That's complete now, and we're in the process of commissioning a gateway for India traffic. In the next two weeks, we'll have Google Pay commissioned for Android. Then we have quite a long workflow to go through to clean up the gateways. We've got talks underway with other regional providers of payment methods in various esoteric countries to split off all of that traffic. The other benefit is some of that revenue will be net new that we've never experienced before simply because there's countries where we've just had historically very, very poor acceptance of payments. For example, Egypt and Pakistan and Bangladesh and what have you. We've got a lot of freelancers from those countries wanting to pay for memberships that haven't been able to historically. It's quite a bit of volume there. So there will be net new revenue in addition to acceptance improvements where we will aim to lift those 30-something percent acceptance rates to 80-something percent as much as we can. So it does pain me to come in quarter after quarter and talk about some product issues, but there have been product issues in Freelancer. We do have a new product management. As I said before at the beginning of the call, you may address questions to Andrew Bateman, who is in charge of the Freelancer product. We do have a lot of revenue initiatives. We do have a lot of wins. If you look at the actual numbers underlying and as much as we do with the graphs, we have had quite a good series of A/B test wins for actual front-end interface improvements, UX interface improvements, helping people service good quality freelancers better in a very unstructured environment. Some of those breakthroughs have only been enabled thanks to AI and the ability to look through and dig through someone's entire work history and surface what is relevant for a completely unstructured data request coming in for work. We also have a number of things which are going to be launched in the next quarter around more complex and more sophisticated ability to get things done on the site. I don't want to preempt those product announcements for the next quarter, but you'll be able to get very, very, very sophisticated work done in a way that no other marketplace in the world is doing right now on the site, which will lead to much high value projects. We've also got a vertical strategy which is being deployed as we speak. As of next week, the first of the vertical marketplaces will go out, where we're targeting very contemporary, modern, and sophisticated verticals and really building liquidity in those key pillars where we know we have strength. As I said before, it's a bit like eBay has a strong pillar in computers and cameras and cars and so forth as their initial strategy rolled out. Today in Facebook Marketplace, there's certain categories they're very good at, as is Amazon, as is all the major marketplaces. We're doing the same. And I do think that it's quite transformational what we're doing to the product. If you go look at the reviews and the ratings and the Trustpilot scores and so forth, we are by far the number 1 rated marketplace out there. Every second day or so, the freelancers are telling me that this is the best marketplace out there in terms of the features and the quality. We have come a long way in the product, but there have been some long-standing issues that we've had to solve, and it's not the least in the face of quite a lot of adversity. I will make a comment just about AI because I'm probably going to get one in the Q&A regardless. I will probably do a commentary about this. I do have a long-standing podcast series that I've done on Macro Voices where I talk about AI with Erik Townsend, as well as write a number of essays on the field. I am not seeing AI really come in and wholesale take jobs from any sort of categories. Potentially, there could be an argument for in the area of translation, there may be a little bit of stuff moving to machine translation and so forth. We handle the whole world of work. So we have every type of job you can possibly think of on the site from $10 to $10 million jobs and quite a lot of stuff for doing very sophisticated, multi-part work, which will be coming out in the next quarter that we've been working on for some time. One would think that if certain niche areas were going to be heavily affected by AI in the broader world, you would see that dislocation. For example, the legal industry, if people are getting Claude to draft all of the legal documents, you will probably see that potentially an industry, for example, the legal industry would be quite troubled at this point in time. You would see it reflected across all legal firms. The nature of our work is dynamic and changes over time. When we first started this business, there were only 20 categories of work. There was graphic design, web design, copywriting, translation, data entry, and a few others. We didn't have any mobile phone app development back when we started this business, right? Then, of course, that category came along and you had an explosion in Android jobs and iOS jobs. Over time, the mix of jobs being performed on Freelancer changes. It's a very deflationary business. You get more bang for buck every year in terms of what you can get done and more sophisticated, more complex work gets done. I would never have imagined back 17 years ago that we would be doing work for the U.S. government in modeling shock propagation and satellites and gene therapy in humans, et cetera, and so on. So the nature of work that you can get done on the platform does go up the stack. The mix of work on the platform does change, although I will point out that in the first half of this year, we haven't really seen much of a mix change at all. Website's gone up a couple of points and reclaimed its number 1 field. Data entries dropped a little bit, but we implemented some more anti-fraud things in data entry to solve some spam that was happening there. Pretty much all the other categories are pretty static, and the U.S. as a percentage of jobs has actually nudged up a couple of points. You can compare this to every quarter to the last quarter. We're not really seeing a big change in mix other than AI-related work is about 7.4%. That's across all the areas of work. We don't have it broken down in a separate category, although we may change that in future reporting. So my point is that the type of work coming into Freelancer will change over time, although we haven't really seen a lot in the last half. Our job is to try and allow more sophisticated and complex work to be done, which I think we've done a great job over the last 17 years, going from $50 websites to $7.5 million gene editing innovation contests, et cetera, and so on. And certainly from the experience that we've seen internally at Freelancer, everything that touches AI has just resulted in more work for humans. While we've got all of our tier 1 support now going through, which is the first touch support going through AI at 11% containment rate. Their support team is busier than ever. Going up the stack, we started a whole strategy and operations group. They're managing the AI frameworks. AI is now in the org chart, reporting into people that were frontline support people. Now they're now managers. We're coming up with new operating roles for these agents in operations and sales and other category areas, et cetera. And we have quite extensive plans for where that's all going to go. So all it's done is create more productivity, and as a result of that, it's created more work around the edge cases for humans, including having an engineering team building the harness and the framework and the features and so on. So I think that there's an infinite amount of work in the world, and that's only going to accelerate over time. And at the same time, the skills and the quality of the work that's coming out of a broader freelance community have gone up very dramatically. I've been very acutely monitoring all sorts of categories over the last 17 years in terms of what I see the output from the freelancers. Let me tell you, 17 years ago, if you got a $50 website done, it was pretty much a $50 website. It looked and felt like a $50 website. But there was a whole part of the market that wanted a $50 website, which is your cafes and so forth and so on. Today, you've got very sophisticated work done, very complex work done across a whole range of categories, a whole range of different areas. The nature of that will change over time. And I think we're pretty excited now because the main problem that we have with this Freelancer business, other than wrangling the product, which we, I think increasingly are getting a stronger handle on, although we've taken a few hits in the last half, is that the trillion-dollar problem to solve in this industry is the Nike problem, which is you see the ads of the runner across the icy lake, who's slam dunking a basketball in an urban alleyway. The peak performance of human athleticism. You go to the Nike store, you buy the Nike shoes, you go home, and you're still fat. We have a similar problem, which is starting a business is hard. Our business is around entrepreneurs that want to start businesses, businesses that want to grow their business, and it's the world of work. The more you consume our product, the more that you have a cognitive overhead, in terms of having to manage people and manage complexity, and businesses require a lot of effort to grow and to build, et cetera. The positive thing is, with AI, we can now provide and become that guide in that world of work, through the cycle of starting and growing businesses. We can do so in a universe where we can now assist, no matter what business you've got, we can provide guidance across that. We can provide very specific tailored advice. We can step by step nudge you in the direction of each particular project or task that you want to execute. We can help you find the right freelancers, no matter how weird or esoteric your requirements are, et cetera. So I do think this environment, effectively AI, I see it no different as the computer or the mobile phone or the internet or software. They're all productivity tools. It is a fantastic productivity tool. But all it's doing is lifting the ability of what humans can do, and it's increasing the amount of work that's available in the world. And we do so in a very deflationary platform, being the Freelancer platform, where we connect up the largest and, by the flip side of that, the lowest cost marketplace for human labor in the world, which is now powered by some of the most advanced productivity technology the world's ever seen. So I do think we're in a pretty phenomenal position. You'll have to kind of watch ensuing quarters to show that we can actually wrangle some of the issues we've been facing to be able to believe that story. I completely understand that. It does pain me to come in quarter by quarter and kind of talk about the core marketplace and the issues facing it, et cetera. I can understand that some investors probably think that there's always something new, and there has been over the last 12 months. I do think in the first half of last year, you could see the cash flow generation. You could see the earnings punching through. You could see the operating leverage was really starting to really kick in. And then we had kind of just a couple of issues that rolling after that. But we have got new financial leadership. We have got new product leadership. We have got a very solid plan. I do think the anecdotal experience that we're getting from our customers is significantly better. You can check out the ratings, you can check out reviews, you can go on the platform, you can talk to the freelancers. You can ask them what they think of the platform. You can ask them what they think of the experience. I think they'll all overwhelmingly say this is the best platform out there, full stop, in the world. I put my hand on heart and say that, and you can go and ask the freelancers yourself. But we had our challenges in the first half. We are fixing them. We've got our heads down, and we're working hard on that. On the Escrow business, that business is doing extremely well. You can see from the long-term graph of the volume that goes into this business that this has just been trending up over time. Quarter on quarter on quarter. This is actually a 26-year history of the business. I don't know how many businesses would have published a 26-year history of volume, but it's up and to the right, and we're now at a point now where we've got seventh year of profitability, the scale to be able to really double down on some growth initiatives. I am excited that we are breaking into new verticals. The sales team is particularly excited about quite a number of things that's been happening in import-export, and specifically in wholesale electronics, servers, computers, IT equipment, data center hardware, et cetera. When they started breaking into this a couple of quarters ago, the transactions were in the 5-digit range, and then they were in the 6-digit range. Now they're in the 7-digit range, and we've seen 8-digit transactions being set up, and now we're seeing 9-digit transactions that have not gone through yet, but we see brokers negotiating them, et cetera. So this is -- I think this vertical is really starting to have a very good product market fit, and the sales team, literally before I walked in, they just mentioned another 7-digit transaction just went through this morning. It's a third transaction from a particular customer. We're integrated now in a couple of these wholesale marketplace platforms, the likes of BrokerBid and The Broker Site, et cetera. We're rolling out across all the others. We're seeing repeat usage. We're seeing very strong usage and growth in terms of repeat use and transaction size and good margins. So we had -- obviously from a low base, we had a quite explosive increase for wholesale electronics, but that's increasing. It's not just -- we're not just seeing it in wholesale electronics. In consumer foods, there's a deal that hasn't actually consummated, but there was a 60-page agreement that was signed between all the parties and quite a complex negotiation. It was for moving of food between various countries and securing supply of particular type of products. While it didn't get up, it was substantial in size, and it was the first transaction coming through from a different partner, which specializes in sourcing products in various markets. So we asked -- and that transaction was an 8-digit transaction. So while that didn't get up, we are seeing quite a lot of attractiveness from the participants in this import-export wholesale space. There are a number of attractive features around escrow being faster, quicker, easier to get deals done, as opposed to go through law firms which slow things down. And so I think that we will see continued growth in this business. We have quite a number of features coming out in the next quarter. We have a whole new front-end interface, which is being rolled out. It's already in front of our top guys at the moment. We have some more sophisticated features to allow us to run greater parts of our customers' businesses, whether it's brokers or dealers or what have you. We have a whole interface for them to try. And we have real-time payments now as well. So we're starting to roll out the ability to receive instant wire transfers, which is obviously a very, very big thing. I'll go into Loadshift and the business continues to grow. The marquee things we've been waiting for are now being deployed. So the calling, which I've talked about for a few quarters, is live now for estimate-only calls. This will allow us to basically reduce the off-siting and increase the customer satisfaction of our customers. We have effectively a whole new way of viewing Loadshift, which is the ability to directly see the trucks on a global map, a bit like an Uber-style experience, which I think is going to also lead to volumes. A new onboarding experience for carriers, which has lifted liquidity. As I said before, we're progressing to other jurisdictions, which we're in discussion right now with a major partner. So I think Loadshift will have a great year, and I think Escrow will have a fantastic year, and we really have to put our heads down with Freelancer. I'll now open up to Q&A. You may address the question to myself, to Dylan Carter, the Chief Financial Officer, or Andrew Bateman, the VP of product. Or you may want to follow up, if you wish, by emailing investor@freelancer.com, and we'll arrange a one-on-one at any time after this. We now open up, any questions, please. I do know that it's a lot to put up in a call, so hopefully [indiscernible] in the next few minutes.

[Operator Instructions]

Just give it a few minutes to people ask a question before it kicks off. If you do ask a question, you may reach out to us individually. As I said before, investor@freelancer.com or Dylan@freelancer.com to reach Dylan Carter or myself, Matt@freelancer.com.

Got no questions right now, but maybe we'll give another minute or two.

Okay, we'll give them one more minute. Don't be shy. Okay, Douglas. Doug, you have a question?

I did send it on the chat. I just -- probably either Dylan or Matt, but what's the estimate of the GMV or revenue drag from the payment gateway issues for this half?

Well, I've got a picture of it actually. It's about up to USD 400,000 a month from this gateway.

Yeah, that's from one gateway. We obviously have a number of gateways in the group, performing and nonperforming. So that would be part of -- what's shown there is just part.

That's the main problem.

It's USD 400,000 per month?

Yes. That's it.

And given like you said this has happened -- this issue has been around for a while, what do you think the total opportunity cost us?

So the issue is, as I said before, we've got many gateways. So we've got -- the left-hand side is our main U.S. gateway, which we commissioned in 2025 -- 2024, that operates at about 85% acceptance rate. You can see there that for initial payments coming in that are fairly steady, that the recurring is growing quite strongly. This is exactly what the Australian gateway looks like. This is what the Apple Pay gateway looks like. This unhealthy gateway is a rest-of-world gateway, which you can see prior to that event, the recurring line was not growing as strongly as the left-hand healthy gateway. So even though we had that event that happened in the first quarter of this year where we lost the $400,000 of traffic, prior to that, when we actually dug through the numbers, we were discovering the acceptance rates for many cards were not high that should have been high. What was happening is we've got an adaptive routing network that we built with payments. In fact, we put a patent in on this and were granted a patent. In hindsight, it's probably a bit questionable whether that patent was actually a smart thing to do or not. What we do is, the philosophy previously was when a card comes in, what we do is we route the card to the gateway, which is the lowest cost and the highest acceptance, and then we kind of keep sending the card around until that payment is collected. What we thought it would do is increase the acceptance of that card. So if a gateway decides it doesn't want to accept the card because cards get issued by an issuing bank and then through the gateway, an acquiring bank will take on the risk, and then at that point, the fee is charged, and the payment is made. We thought that we were maximizing acceptance by moving the card around and just finding a gateway that would accept it, and then we dynamically would look at what the acceptance rates are and what the costs are. What we think in hindsight, though, that was doing was bouncing cards around a bunch of gateways that would never be accepted because potentially they're from India and so forth. While in the West, we have networks like Mastercard and Visa and American Express. In places like India, you have card networks like RuPay, where simply if you don't accept that network, you just won't accept the card. So I think and what we've definitely observed and is that when you send cards to a gateway and the gateway doesn't accept them, the gateway will back off on all card acceptances. So that's why you see this recurring line here on the right-hand side, prior to that gateway failure event, was not growing as strongly as the one on the left-hand side. And this gateway was doing 30-something percent acceptance rate across all cards. The assumption the payments team, up until that point in time, was, well, there's a whole bunch of cards of people with no money. There's potentially a bit of fraud mixed in, whatever. And we kind of bounced the cards around, and then we kind of got a few gateways where we were kind of collecting all, everything, and everything was kind of hitting if it didn't work on the other gateways. We took that different philosophy when we deployed the gateway on the left-hand side, where we only sent very pristine U.S. traffic to that gateway and nothing else. The acceptance has been remarkable, and you can see the customers keep coming back, and their cards work. So we think that on these other gateways, which are taking quite a large amount of volume, acceptance rates of 30-something percent should have been closer to 80-something percent. Now, you can only get there if you specifically have the gateways that accept those payment types. So we have embarked, in the last couple of quarters, on splitting off. Having observed prior to this event what was happening on the main U.S. gateway, where we were seeing good recurring, we then basically went to implement an Indian local acquiring method. We had to go through quite a lot of work for that because the Indian regulation is quite heavy with a thing called E-mandate, which requires effectively an online document to be signed for someone to effectively have their card used for recurring membership. We've got that done, and we've got that completed, and there's processes being commissioning. So you've got to be able to have local acquiring methods that would accept the card in the first place. You then take that traffic off the main credit card gateways, which would increase their acceptance rates, and then you'd be very careful about what traffic you are sending to those gateways. Rather than around robbing them and bouncing them around in a dynamic routing method, if they're hitting all the gateways where you do have a failure, you have to be very careful and very specific about where you send the cards. Now not every business would've experienced this because we have a very large global number of participants sending us cards from all places around the world. But we do think that these gateways should not have been performing at the level they were performing at. I think the assumptions that were made previously, where they were just treated as catchalls and if a card didn't accept on sort of the top couple of gateways, it kept bouncing around, was probably a mistake. So we think this has been quite a big drag. I mean if you have a lemonade stand at the front of the office and you try and take payments and 6 in 10 customers you say, go away, I can't take your payment, you're not getting much repeat business. So we do think this has been a drag, and we think the drag has continued since 2013, actually, when we turned on our first gateway, to be honest.

No, No, thanks for that. Maybe I'll just shift to Escrow. Maybe this one's for Dylan. There didn't seem to be much operating leverage, given the revenue growth. Can you just give us some details on the lower take rates? Is that because of certain segments or the transaction size? And maybe touch on the higher SG&A costs for that segment, too, and how do you see that sort of evolving going forward?

So the operating leverage of that business is partly impacted by transfer pricing between the group's shared costs. So when you exclude those, you're actually seeing quite a lot of operating leverage within the direct costs of that business, escrow.

We calculate that based on the proportion of revenue contribution. So if the revenue contribution of Freelancer goes down, then the costs go up on the escrow side as a result of that.

So the underlying is there actually is a lot of leverage in that group, and the way in which our overall group policy sits, that's why you're seeing that dynamic. There are more fixed costs in the cost of sales line for Escrow. A lot of the bank charges are dollar value charges, as opposed to percentage charges on the transaction volume. So we actually do see the cost of sales margin improvement, as you saw in this half.

Cool. No, great. And the lower take rate, is that because of the certain categories or is that because of the transaction size that's occurring?

We price Escrow.com differently based on different verticals, not just the value of transactions. So for example, if you do automotive, typically, the end market is only willing to pay $99 at most for an Escrow transaction, or $49 if you're not doing title transfer. If you do real estate, it's 0.2% plus $250. If you do an airplane, it's 0.35%. If you do a domain name, we can charge potentially rack rate. If it's an IP address, we might do it a little bit cheaper because it's less subjective in the valuation. It's more of a commodity thing. If we're breaking into a new market, such as wholesale electronics, we may discount slightly in order to capture that market. And then, of course, if the transaction size is large, we will discount as well. We only really had 1 major transaction in the half, in the sort of 8-digit range that was of scale. That was actually in the last week of the half. We actually had several more in the sort of $60 million, $100 million range that didn't get set up, that we think might come back in the future. It was fairly diversified across lower value transactions. But there are a range of different markets, and we're breaking into a new market as well, being a wholesale import, export, et cetera. So that's why it's reflected at 1.35%. I will mention that that take rate did get as low as 0.89 in previous years. So when we bought the business, it was 1.45. So 1.38, or was it 1.39?

1.35 is not too far off a good historical rate, particularly for a 4th highest quarter ever. I would expect it to be around the 1.45, 1.5 is a pretty good steady state value for that. We had a higher number in the last period; I think it's pretty healthy and it's certainly not at a heavily discounted because it's pretty close to the long-term average.