Description

Thesis

Growth is re-accelerating and is likely to hit high teens in ‘24 and 20%+ by ‘25 as large covid new customer cohorts lap their large first and second year declines, which recently has slowed the company’s top line. Active buyers will start growing again in ‘25 and the push to mid-market will continue growing the spend per buyer for years. Coupled with further, though smaller, margin improvements in ‘24 due to scale and marketing efficiencies, the company will get back to the Rule of 40 and achieve a higher multiple again. It is too cheap at 12x free cash flow (after adjusted SBC) two years out.

Summary

We bought Fiverr for the first time in late 2019 at around $20. Nathanj posted a good writeup on FVRR that fall. We sold it two years later for close to 10x our money. I wish we sold all of our tech positions around the same time but that’s a different subject altogether. Anyway, we’ve now bought it again as we believe the business will resume faster growth due to lapping Covid cohorts and it has much higher profitability, evident on adjusted EBITDA numbers but not yet in GAAP numbers, due to temporarily inflated stock compensation. 

We won’t spend much time on discussing the freelancing market. Generally it is growing double digits, especially post covid with more remote work and people choosing flexibility. The vast majority of freelancing is done offline (e.g. 90%), and the two dominant global online players for the last 5+ years have been Fiverr and Upwork. TAM estimates for freelancing work are in the hundreds of billions, and Fiverr’s GMV will hit $1.2B this year, so it’s a very small part of the market.

Fiverr specializes in smaller pre-defined “work gigs”, with an average spend/buyer around $270 and has built the most sophisticated marketplace with 700 SKU-like categories of jobs in every possible vertical, easy search functionality, good website, great matching algorithms between buyer and sellers, freelancers ranked by buyers, much more brand recognition now post Covid, large trove of data, very good search engine optimization tools to drive marketing efficiency. The key advantages for the buyers are easy onboarding, large catalog (you find help for what you need), speed to getting the job done (vs. offline or UPWK), clarity of specs & delivery time, low cost (the competitiveness of the marketplace leads to low prices, and geographical arbitrage), great seller ratings system, all demonstrated by a buyer NPS pushing 70. We’ve used it extensively in house over time with good results. The key advantages for the seller are speed (no search costs for new gigs, one can sign up, post a gig, and get a match possibly same/next day), clear specs (no haggling over what has to be delivered), certainty of price (pre-defined) and easy payment (handled by FVRR), as shown by a seller NPS pushing 80. Buyers pay 5.5%, sellers 20%, and a host of mostly seller value-added services (promoted gigs - paid to be displayed higher, but still ranked by their algos; seller plus - traffic & keyword analytics to improve matching; learning and workflow tools), but also a couple on the buyer side (subscriptions, freelancer management system, curated sellers) push the take rate to close to 31%, on which they have had no pressure, and management has signaled continued small increases. 

We will not discuss Upwork here in detail but they have mostly an RFP-based system which is significantly slower to job start, with a low hit rate for freelancers, much higher average job sizes with $5k/spend per client, and longer time to completion. In the last 2-3 years they started a Fiverr-copycat catalog business which is more gig-like but without much impact on Fiverr. Other large possible competitors have failed to make any significant inroads (eg. LinkedIn has a version of a freelancer marketplace that has not done well). Vertical players (e.g. Toptal and others) were more of a concern for us years ago. Horizontal platforms like FVRR and UPWK have demonstrated stable business models, have grown the platform sizes significantly, and freelancers and buyers benefit from being present on fewer platforms from a time management perspective. Further, a more diversified platform is better in the age of AI. 

Fiverr has been taking market share over time. Just simply looking at FVRR share of FVRR+UPWK revenues, they went from 20% in 2017 to 35-36% now. Due to the cohort behavior described below which has an outsized impact on FVRR, they already hit this by 2020-2021, and it has been stable since. We track this by also looking at google trends across a large variety of other marketplaces (eg. FVRR, UPWK, Freelancer, PeopleperHour, Guru, Toptal, etc.) and this shows a probably more accurate picture that FVRR is the leading marketplace with roughly 55% share, followed by UPWK w/ about 33%, and all other sites much smaller in the single digits. 

A push over the last year was redesigning their upmarket offering with better targeted offerings to segment mid and large customers into higher spend categories. This is the new Pro and Enterprise upgraded business suite version of their prior Fiverr Business category, which has grown from about 5% of GMV to 10% in the last 3+yrs and is growing faster with much larger 3x+ spend/buyer, which with the new products is likely to increase repeat buying and we believe it can lead to much higher 10x spend/buyer in this category in the future (e.g. UPWK projects are on average $5k/client per year). The key advantages for the buyers will be access to pre-vetted 1% of talent, larger project sizes and team formats, back-office tools like managing freelancer projects & compliance, having a “concierge” to help guide through all this. A completely separate new acquisition channel is Fiverr Certified, where large tech companies (such as Amazon Ad business, TikTok, Stripe, Monday.com, they have two dozen on the platform now and adding) started co-branded dedicated websites (i.e. marketplaces), where pre-selected Fiverr talent will solve recurring tasks needed for the partner’s own small business customers (eg onboarding, data integration, training, etc). This will start contributing in earnest in ‘24 and will drive active buyers but especially much higher spend/buyer (predefined tasks costs thousands of dollars) at likely higher margins, since some of the marketing costs are picked up by tech partners who push the offering onto their client base. 

To understand the core Fiverr business and the revenue growth potential (ex the new growth initiatives above), we focus on the buyer cohort behavior. The company publishes once a year charts that let you figure this out. Ultimately this is what drives the Active buyers and the Spend per buyer. The cohort work we’ve done drives home the point that Fiverr can reasonably return to roughly 20% topline growth later in ‘24 and into ‘25, a vast improvement over the single digit growth of the last three quarters that has hurt the stock: 

• Pre covid, the year 1 new buyer cohort (which they win via marketing spend), used to decline approx -40% in year 2, -15% in year 3, and then grow 1-2% from year 4 on. The latter is why their repeat buyer business has kept growing and the co now generates over 65% of revenues from repeat buyers vs low 50% a few years ago. This is important because they pay little in marketing costs to retain these buyers, which leads to a lot of operating leverage over time.

• Example: in 2019 we estimate the Y1 new buyer cohort contributed $46m in revenues. The Y2 cohort (prior year new buyers) contributed $20m = this was Y1 2018 * (1-40% decline), roughly. The Y3 cohort contributed a bit over $12m. All other cohorts Y4+ contributed $29m.

• Two positive things happened in Covid, i.e. in 2020, 2021: the Y1 cohort got a lot bigger (because more people bought gigs and provided remote freelancing work), and the large Y2 drop got smaller (because the platform became temporarily even more useful). 

• Example: in 2020 we estimate the Y1 cohort contributed $85m in revenues (a 90% y/y increase vs a more typical 40% increase). The Y2 cohort contributed $34m, a smaller -25% decline vs the -40% typical. So this combo of large Y1 + large Y2 spiked the revenues. In 2021 the Y1 cohort y/y increase was more normal, but to a still very large $125m, and the Y2 cohort decline was still small. 

• In ‘22 things started going “post Covid” as these positives started reverting (and this process will be completed in 2023 by our calcs): the Y1 cohort was flat at a high level (so no growth contribution), and the Y2 decline was close to -50%, worse than the historical -40% trend. In ‘23 we think we will be fully “post Covid”, with the Y1 cohort smaller than ‘22 by double digits, to below a “normalized” level of around $100m spend (due to weaker macro), and a Y2 y/y drop only a bit worse than historical as the reversion continued for part of this year.

• Thus, ‘22 and ‘23 growth rates were reduced from (1) post-covid zero growth or declines in the largest cohort Y1 coming off high levels, and (2) larger than normal drops in the second largest cohort Y2. This brought down the revenue growth substantially.

• It’s important to note that the cohort behavior described above is the main driver of the Active buyer (AB) count, which some are focused on as a problem with the business. As Covid hit, ABs had their largest sequential increase of 0.35m ppl in Q2.20 and this started lapping in Q2.21 (0.2m increase, Y1 drop lap), and declined sequentially for the first time in Q2.22 (Y2 drop lap, -0.05m), the first drop. ABs have been roughly flat since around 4.2m and will very likely stabilize by H1.24 then start increasing, also as FVRR does more marketing.  

• Based on conversations with management we have gotten the sense that a reasonable ongoing expectation of cohort spend remains $100, $60, $50 in the first 3 years, which is in line w/ pre covid actuals and our assumptions above. Management is continuously communicating that cohort behavior is now stable vs pre Covid, and if anything, better with the higher value buyers that have more tools.

• We believe that ‘24 will see double digit growth in the large Y1 cohort, below historical trends, due to the weaker ongoing macro which dampens their marketing spend and the efficiency of the spend. Assuming a macro lift by ‘25 the Y1 cohort should accelerate to a mid 20%s growth as their marketing investments return to a more normal level and efficiency. We also assume continued normalization in the Y2 cohort drop. 

• Running the math on the existing cohorts at unchanged marketing efficiency and basically no impact from Certified, conservatively leads us to high teens growth in ‘24 and low 20%s in ‘25. Example: in ‘24 Y1 110m, Y2 60m, Y3 60m, Y4+200m so roughly $430m in revenues. Of course one can reasonably disagree on our estimates of S&M spend, or the efficiency of S&M spend per dollar of Y1 cohort revenue (how we drive the marketing efficiency), but recent efficiency trends have been in line w/ pre-covid (despite less spend in weaker macro), and we are assuming similar levels for ‘24 (again conservative), and slight increases ‘25+ due to larger scale, more brand awareness, and SEO improvements. 

Stock compensation math is the second insight we have. The idea is simple: stock comp is inflated, so sensible market participants who exclude it from “true profits” believe this company makes no money (just turned GAAP positive net income in Q2.23 first time ever), but this is not true. Like most analysts steeped in the value mindset, we think of stock comp as a real expense. However, in the case of FVRR an adjustment is necessary given the wild rises in the stock price which coincided with multi-year RSU and option grants which get amortized in the stock comp expense over roughly 4 years. 

• The company issued a good chunk of RSUs in 2020 and 2021 (4-5 million over the two years) at very high stock prices. Share comp started doubling in Q4.20 (stock >$150, 11% of revenues vs 7% of revenues in Q3.20), then spiked by Q2-Q3.21 (stock >$200), and this cumulative effect of additional issuance at high prices vastly increased the accounting expense of the share comp. 

• Disclosures are clear that stock comp is recognized over roughly 4 years, measured at fair value on grant date, and then recognized as an expense using the straight line method. So share comp expenses inside the income statement are inflated until 2024-25 due to this accounting method.

• Stock comp has now started declining, -7% in Q1 and -8% in Q2 and we expect it to decline until 2025.

• So what’s the “real” stock comp expense? The difference in basic share outstanding q/q (cleaner) is on average roughly +230k/qtr over the last 10 quarters, quite stable w/ some up/down fluctuations. So let’s call this 1m shrs/yr or about 2.5% of fully diluted count (not far from what the company has been signaling). Thus, the “real” stock comp today is 1m * $29 current pierce = $29m/ year, not $65-$70m they are likely to book this year.

• So a simple proxy for “true FCF” this year would be to take the expected FCF including stock comp for ‘24 and deduct $29m from it.

• We expect this continues declining until ‘24, and as % of revenues the current absurdly high level of 19-20% will drop to a max of 9-10% by ‘25-’26, and then continue lower into the single digits over time. This was 8% of revenues in 2019 and 2020 before the large Covid spikes. Also UPWK, whose stock didn’t have anywhere near the exponential rise of FVRR in 2020-2021 is spending 11-12% of stock comp/revs today and this is also declining now.

The balance sheet is solid with excess cash, as they have close to $700m of cash, growing each quarter, and $460m of debt, a 0% convert at $214/shr which will either be bought back early (trading at 87, <7% yield) or repaid at maturity in Nov ‘25.

Fiverr has had slight beats and raises to the top line guidance in the last two quarters. The stock initially rallied but then gave back the gains. The bear case is a combination of slow growth (especially vs. UPWK), no “real” profits, and AI fears. UPWK shares have outperformed by 30% vs. Fiverr this year, as optically UPWK was more of a “broken” story, coming into ‘23 with a significantly lower valuation, slowing growth and negative margins, and while growth has continued slowing (vs. FVRR starting to accelerate), they delivered large margin improvements which have contributed to UPWK stock outperformance. FVRR is now seen as a “show me story”, with lagging growth and discounted profitability due to large stock comp. 

We have found typical alternative data sets to not be very additive in this case, with wide and fluctuating deltas to reported revenues. Credit card data (various) and web traffic data (eg SimilarWeb) have not been a good predictor. Indicatively, US app downloads (better proxy for demand), DAUs and MAUs are increasing in the last few months vs. declines earlier in the year. Global google trends are also not predictive, and have shown continued double digit growth YTD, with some slowing likely due to the cyclical effects. We look at US business applications (tracked by the Fed), which is a measure of new small business starts, which the company has discussed in the past as correlated to their small business buyer, and indeed we have observed a moderately high (eg 60-70%) correlation with Fiverr’s growth. These applications have accelerated year to date, eg. were negative y/y as of YE ‘22, +4% in Q1.23, +7% in Q2.23, and crossed into double digit y/y growth in June and July ‘23, and weekly data shows continued 12-13% growth into early September. In summary, we believe their Q3 (and 2023) guidance is achievable.

Risks

• Fiverr’s move upmarket doesn’t succeed, or requires more personnel hires, as UPWK is already present there. First, we think the Fiverr model is very different than UPWK’s RFP based business. Second, Fiverr’s core business which drives the cohorts will steer the ship for a while as it’s 90% of the GMV. The company has no plans for a salesforce, and the new midmarket tools are likely to continue accelerating adoption into a larger spend per buyer.

• AI turns out to be a longer term threat. It’s very difficult to disprove that it won’t get worse with future versions of AI and some work would get displaced. We worked on estimating their AI-impacted categories and looking at various seller surveys and expert calls. The conclusion was that the impact would be very small, possibly positive in the ST (mostly due to productivity enhancements over a large number of categories, offset by larger negative impacts in a few categories), and unknown to small negative in the LT. Both FVRR and UPWK management teams are consistently saying that AI is a net benefit to the business. In the end we believe this is true for the next few years as AI tools develop at a high rate, the number of alternative AI providers increases fast and buyers’ search costs also increase (what’s new in AI this week?). Smaller and medium companies (the core target market here) do not have the resources to acquire or develop the expertise in house. It certainly makes sense that freelance talent, which will be expert at incorporating AI in their work as it increases their productivity and thus pay, will provide a good return on investment for this target market.

• Cyclical concerns: The business turns early in the cycle due to small business concentration and signaled a downturn early in the summer of ‘22, which further slowed the growth. Now it’s been stable for a few quarters and will rebound early. The environment has stabilized for the last couple of quarters. FVRR’s hint at some increased spend in Q3-Q4 means management potentially sees green shoots around the corner. In early September at a conference the UPWK CEO also signaled that smaller businesses are “feeling maybe more confident with the macro in Q2” than the previous quarter. The two concerns highlighted recurringly from the buyers perspective were inflation and interest rates, both of which are likely lower in a year. 

• Perennial worry of freelancers taking the work off the system i.e. the gray market. We were more worried about this 3-4 years ago and it’s less of a concern now. The system really does provide significant advantages to the freelancer in terms of search costs & certainty of payment. The company has sophisticated algos that check for this type of behavior and freelancers will see a drop in their “rank” and inbound business if this is suspected, so it prevents some of this behavior. 

Management

We’ve known management for four years and believe they are good at both long term strategy and day to day execution (a problem in many tech firms), and have found them to be honest. Micha Kaufman the CEO is the founder and is a truly innovative product guy. He is a repeat entrepreneur and a relentless innovator that has built the best marketplace offering in freelancing, constantly evolving new features. A recent example is his brainchild Fiverr Neo, an early beta product that uses AI to provide an easy conversational experience to a buyer interested in a new project and improve the matching between buyers and sellers, while seamlessly generating a trove of useful profile data on the buyer. While this is likely not monetizable directly, it will likely improve the customer experience and lead to more repeat business. Micha is certainly a visionary and more prone to being optimistic.

Ofer Katz is a good match to Micha, as a conservative CFO, focused on execution, who has been with the business since 2011, and full time CFO since 2017. In late summer 2022 he started signaling an accelerated increase in margins and they delivered very well against this target. 

Valuation

As the last three years have shown, it’s hard to value growing tech stocks with small GAAP earnings. Our primary valuation method here is discussed above in the share comp math, where we basically calculate a “true” FCF (earning power) adjusted by the correct (smaller) stock comp expense, and get to $1.30, $1.80, $2.40 for ‘23-25. We add back roughly 60% of stock comp in ‘23 and this declines over time to 15% by ‘26. This results in a 20% true FCF margin in ‘25, coupled with roughly 20% growth. This is a Rule of 40 company dominant in a duopoly industry, and should be trading at 30x PE, so our price target is $70. In an upside case we can easily see it at $90-100/share. Our downside is 2x revenues less net cash, or $26. 

Our true FCF number above is roughly 90% of the adjusted EBITDA, so for simplicity one can define the R40 as rev growth + adj EBITDA margin as well. Looking at EV/adj EBITDA we project 10x on ‘24 and 7x on ‘25 so again very cheap. Looking at EV/EBITDA (ex all stock comp, though again we don’t think this is correct) we get 14x on ‘25 and 7x on ‘26.

An alternative we use for tech stocks is EV/GP because we think it normalizes well across business models and gives you a good indication of PE at larger operations scale, when earnings convert from GP at say ~30%. Our rule of thumb is 3x GP is “very cheap” (e.g. 10x PE), and 2x is “stupid cheap”. On this metric FVRR is trading at 3.3x on 23, 2.4x on ‘24, and 1.8x on ‘25. FVRR’s median lifetime EV/GP is 6.2 and UPWK’s is 6.1, so similar. This seems reasonable to us as an intermediate valuation goal for both companies if you believe (like we do) that they basically own the online freelancing market and both will have good growth for years given that the vast majority of spend is now offline, and margins will keep scaling with opex leverage.

A few historical observations that support a re-rating of FVRR in the next year, possibly starting in Q4.23:

• FVRR’s rule of 40 is highly correlated (eg 90%) with its EV/GP, particularly when R40 goes above 30%. In Q4.19 the EV/GP jumped to 5.5x as the R40 went to 31% (from prior qtr 26%). In Q2.20, when covid massively accelerated revs and R40 went to 90%, EV/GP jumped to 16.3x. R40 only hit a max of 100% in Q1.21 but EV/GP peaked at an unbelievable 29x. In Q2.22 R40 dropped to 18% (post covid cohorts slowdown + just 5% margin), and EV/GP fell to 3.3. The R40 bottomed in Q1.23 at 15%, hit 22% in Q2.23, will be 28% in Q3 and pass 30% in Q4. This will likely surpass 40% by H2 2024.

• FVRR’s EV/GP is higher than UPWK’s when FVRR R40 outperforms (again 90% correlation). This is abundantly clear from Q4.19 to Q1.22. We expect that FVRR - UPWK R40 delta will increase to double digits (10%+ in H2.22 and in ‘24), and now right FVRR is trading at a -0.5x discount to UPWK on EV/GP (which doesn’t make sense), and both stocks are at low absolute levels.

Catalyst

• Reacceleration in growth due to cohort normalization, higher spend/ buyer, macro better at some point in ‘24, which also leads to growing active buyers. Our projections, we think, are conservative and lead to only single digit revenue beats vs consensus, but we believe the primary driver of the stock will be proving out the revenue acceleration

• Back to Rule of 40 by H2.24, crosses 30% by Q4.22 which has been an inflection point for the valuation in the past

• Margins continue pushing higher in ‘24 due to growth & marketing efficiency. We bake in 200-250 basis point increases for the next two years, with an EBITDA margin of 21-22% in ‘25.

• Likely to increase the LT profitability guidance over the next 1-2 years. CFO hinted in September at a conference that “we promised 25% margin in the long term. There is a way to go, but it’s not too long anymore”

• AI fears dampen significantly with topline re-acceleration and continued execution