Building a diversified crowdlending portfolio: projects, platforms, sectors, what 776 scored projects show
Diversifying a crowdlending portfolio is not about adding more lines: it is about making sure that no single borrower, no single platform operator and no single asset class can sink the whole thing. The order of priority our data supports is this: first the number of distinct borrowers, then the number of genuinely independent operators, and finally the mix of asset classes and yield levels. Ten projects from the same developer, or two platforms sharing one licence, are not worth ten lines or two platforms.
According to the Forvis Mazars / France FinTech barometer, as relayed by JeChange and Fortuny Conseil, 25 to 30% of French real estate crowdfunding projects were more than six months late at the end of 2025 [9][10]. The AMF, France's market regulator, considers that late repayments are "largement sous-estimés" (largely underestimated) in the statistics platforms communicate [1]. And in 2026, Tudigo went from court-supervised recovery to judicial liquidation [14][15]. Concentration is one of the few risks an investor can control directly. To measure it, we draw on 776 projects scored by CrowdPickr between 6 May and 2 November 2026, across 23 platforms.
Why a single project is a binary bet
A crowdfunding loan or bond has an asymmetric profile: the upside is capped at the advertised rate, while the downside can be the entire capital. The AMF reminds investors that they may "perdre tout ou partie du capital investi ou des sommes prêtées" (lose all or part of the capital invested or the sums lent) and that "le paiement des intérêts puis le remboursement in fine du capital ne sont jamais garantis" (interest payments and the final repayment of capital are never guaranteed) [5].
Market figures show this risk is not theoretical. In its analysis published on 20 January 2025, the AMF states that about 30% of projects in progress at 31 March 2024 on the ten largest platforms were delayed [1]. Two years after launch, 36.9% of 2021-vintage projects were late, against 20.1% for the 2020 vintage [1]. For the 2019 vintage, the displayed delay rate was 15.2%, but it "jumps to 60.2%" once late repayments are included [1].
The barometer relayed by Fortuny Conseil, a wealth advisory firm, adds that 20 to 25% of funded projects are now in collective insolvency proceedings and that "problem files are piling up faster than they are being resolved" [10]. On the developer side, Altares counted 214 insolvency procedures in property development in the first quarter of 2026, up 50.7% year on year, according to JeChange [9]. These measures do not use the same definitions as the AMF's: they cannot be added together, only read as pointing in the same direction. We deliberately quote no definitive loss rate: the available estimates contradict each other and the barometer's primary document could not be consulted.
A delay is not a loss, but it locks up capital. The AMF notes that a procedure ending in a definitive loss "takes several years" [1], and Homunity's FAQ, as quoted by the law firm Kohen Avocats, puts developer procedures at "often between 12 and 36 months" [11]. With a single line, one incident is enough to freeze most of your allocation for years.
For a more detailed picture, see our article on delay rates by platform.
Layer 1: the number of projects
The first rule is arithmetic. In 2017, Nicolas Lesur, then president of the French crowdfunding federation, put it this way in Le Figaro: "If you finance a hundred projects on a platform's advice, you will necessarily have one or more defaults, but there is little chance they will eat into your capital" [13]. Diversification does not prevent defaults, it makes them bearable.
The European ECSPR regulation sets a useful marker. According to a summary by the law firm Jeantet, a non-sophisticated investor goes through a knowledge test and a simulation of their ability to bear losses, and any investment above the higher of 1,000 euros or 5% of their net worth in a single offer triggers a risk warning and requires explicit consent, with a reflection period of four calendar days [19]. That threshold is not a ceiling, since an investor can go above it after the warning, and still less a target: in our reading, an individual line should stay well below it.
The minimum ticket determines in practice how many lines you can hold. In our data, it ranges from 1 euro (La Première Brique, Debitum) or 10 euros (Afranga, Enerfip, Bricks) up to 1,000 euros (Homunity, ClubFunding, Lendosphere, Fundimmo).
| Minimum ticket | Platforms (CrowdPickr data) | Lines possible with 5,000 euros |
|---|---|---|
| 1 to 10 euros | La Première Brique, Debitum, Afranga, Enerfip, Bricks | As many as there are eligible projects |
| 50 to 100 euros | Maclear, Baltis, Les Entreprêteurs, Lita, Miimosa, WiSEED | 50 to 100 |
| 500 euros | FF Forest | 10 |
| 1,000 euros | Homunity, ClubFunding, Lendosphere, Fundimmo | 5 |
With 5,000 euros and 1,000-euro tickets, you hold five lines: a single total default wipes out 20% of the portfolio. With 250-euro tickets, you hold twenty, and the same incident weighs 5%. Our practical recommendation: tickets of roughly equal size, so that no line dominates, and a number of lines decided before you start, rather than larger amounts on the projects you "like" most.
Layer 2: borrowers, the hidden concentration
Counting projects is not enough. What matters is the number of distinct borrowers, because a developer in trouble usually drags all of its programmes down with it.
The textbook case dates from 2017, when UFC-Que Choisir, the French consumer association, reported a "premier défaut de paiement de grande ampleur" (first large-scale payment default) [12]. Groupe Terlat, placed in court-supervised recovery on 6 January 2017, had six programmes financed on two platforms at once: two on WiSEED, where it had raised about 997,000 euros from some 260 investors, and four on Anaxago, for about 1.7 million euros raised from 250 investors [12]. An investor who thought they had diversified by using two platforms was in fact carrying a single risk. The magazine noted that it would "not be a total loss", as the developer had already repaid half of the sums raised [12].
The same pattern exists outside real estate. According to The Crowd Space, a specialist trade publication relaying Bignewsnetwork, Mintos investors had more than 61 million euros exposed to a single partner lender, Nera Capital, whose payments have been delayed since late March 2026 [17]. The publication sums it up: "Concentrated exposure turns delays into freezes" [17].
Our own data shows this concentration on two foreign platforms:
- Debitum: of 94 projects scored, 52 are issued by a single borrower, the Latvian Forest Development Fund, and 5 more by a fund with a similar name, Latvian Forest Development Fund II. The first alone therefore accounts for more than half of the offering we analysed, even though each issue is a separate security. Our review of FF Forest looks at the questions raised by forestry assets.
- Afranga: of 159 loans scored, 69 come from a single originator, Tiberus, and 43 each from Stikcredit and Lendivo. Three originators thus account for 155 of the 159 loans.
On French platforms, projects are generally named after the programme rather than the developer, so our query does not catch repeat developers there. It is up to you to check them in each project's documentation. Our guide on how to read a crowdlending project file shows where to look.
The practical rule: keep a table by borrower, not by project, and set a cap per developer or originator. In our reasoning, a single borrower should not exceed twice the size of a standard line, across all platforms combined.
Layer 3: platforms, and what counts as a separate platform
A platform going bust does not erase your investments, but it can freeze them. When Tudigo entered court-supervised recovery on 5 August 2026, the platform stated, as reported by MoneyVox, that investors' funds, held in separate accounts, "are therefore not affected by the recovery procedure", while specifying that "any payout is subject to the prior authorisation of the court-appointed administrator" [14]. After the liquidation ordered on 12 October 2026, Tourmag described "26,000 investors and nearly 700 shareholders in complete uncertainty" and 116 investment vehicles left "without a pilot", while noting that the liquidation "does not, in itself, cause the sums already invested to disappear" [15].
The law firm Kohen Avocats explains that a platform's liquidation "does not, by itself, bring your securities into its estate" [16]. The deadline that can cost you a claim mainly concerns the borrower: if the developer or issuer enters collective insolvency proceedings, claims must be filed within two months of the judgment's publication in the BODACC, France's official gazette of commercial notices, and JeChange points out that "after that deadline, the claim is lost" [9]. For bonds, the bondholders' representative (représentant de la masse) is empowered to file claims on behalf of all holders [16]. For the detailed procedure, see our guide on platform failures and filing a claim and our analysis of the Tudigo liquidation.
Platform risk is not limited to bankruptcies. The AMF's list of withdrawn crowdfunding service providers records seven licence withdrawals between June 2025 and July 2026: Koregraf (20 June 2025), Lendopolis under its number FP-2023-26 (16 December 2025), Lymo Finance (8 January 2026), Zaleo (1 April 2026), Lumo (2 April 2026), Tylia Invest (2 June 2026) and Crowdybee (31 July 2026) [2]. The register gives no firm-specific reasons, only the list of general grounds for withdrawal, among them renouncing the licence, nine consecutive months without activity or a serious breach of the rules [2]. A withdrawal is therefore not necessarily a sanction: the Lendopolis brand, for example, still operates under another licence [3].
The Lendosphere and Lendopolis trap
Two brands do not always make two platforms. In November 2024, MoneyVox reported that Lendopolis, a La Banque Postale subsidiary, "va finalement être rachetée par Lendosphere" (will ultimately be bought by Lendosphere), adding that the move changed nothing for investors on either platform [18]. Today, the AMF white list attaches the trading names "Lendosphere, Lendopolis" to a single licensed provider, Lendocompany, under number FP-2023-22 [3]. In other words, Lendopolis now operates under the same licence as Lendosphere.
Our data confirms this on the supply side: at least eight programmes appear on both platforms, including the Chevenon, Ségrie, Le Pal 2, Charentonnay and Venaco solar plants, the Puichéric agrivoltaic project, the Photosol group and VALREA. Holding projects on Lendosphere and on Lendopolis is therefore not platform diversification, and can even double your exposure to the same borrower.
A good platform does not replace project selection
Our scores show that the platform is only part of the equation. Averages differ sharply, but within a single platform, projects spread over a wide range.
| Platform | Projects scored | Average score | Min / max | Share ≥ 65 | Share < 35 |
|---|---|---|---|---|---|
| Lendopolis | 25 | 70.0 | 53 / 82 | 84% | 0% |
| Lendosphere | 38 | 65.3 | 46 / 80 | 61% | 0% |
| Enerfip | 82 | 61.1 | 40 / 84 | 35% | 0% |
| Homunity | 20 | 54.5 | 17 / 77 | 35% | 15% |
| Debitum | 94 | 48.0 | 17 / 70 | 20% | 35% |
| Afranga | 159 | 41.9 | 20.3 / 63.3 | 0% | 19% |
| BienPrêter | 40 | 27.9 | 15 / 69.9 | 5% | 70% |
| FF Forest | 62 | 16.4 | 1 / 26 | 0% | 100% |
At Homunity, scores run from 17 to 77; at BienPrêter, from 15 to 69.9; at Debitum, from 17 to 70. Choosing a "good platform" lowers the average risk, but it does not replace project-by-project selection. These scores are an assessment of project quality at the time of fundraising, not observed default rates, and they only cover projects scored since May 2026.
Layer 4: asset classes and yield levels
Yield, sector and platform are not independent bets: the highest yields cluster in the sectors and on the platforms with the lowest scores.
| Advertised yield | Projects | Average score | Score ≥ 65 | Score < 35 ("Avoid") |
|---|---|---|---|---|
| Under 7% | 66 | 63.2 | 41 | 3 |
| 7 to 9% | 110 | 56.9 | 31 | 3 |
| 9 to 11% | 143 | 52.1 | 21 | 12 |
| 11 to 13% | 220 | 50.1 | 36 | 39 |
| 13% and above | 186 | 35.9 | 9 | 92 |
The average score falls with every yield bracket. At 13% and above, almost one project in two (92 of 186) is rated "Avoid" and only 9 reach 65. The share of "Avoid" projects rises from 8% between 9 and 11% to 18% between 11 and 13%, then to 49% at 13% and above. This table covers the 725 projects for which the yield is recorded.
By asset class, the gap is even wider:
| Asset class | Projects | Average score | Average yield | Average duration |
|---|---|---|---|---|
| Renewable energy | 147 | 62.9 | 7.49% | 41.3 months |
| Real estate, development | 41 | 56.6 | 10.76% | 21.3 months |
| Real estate, buy-renovate-resell | 108 | 55.7 | 10.51% | 16.8 months |
| Commercial real estate | 34 | 55.1 | 10.04% | 32.8 months |
| SMEs and originator loans | 319 | 43.9 | 11.74% | 20.5 months |
| Invoice finance | 35 | 26.5 | 13.14% | 18.5 months |
| Agriculture and forestry | 74 | 24.9 | 15.89% | 21.4 months |
Renewable energy has the strongest profile (73 of 147 projects at 65 or above), agriculture the weakest (60 of 74 below 35). That last category is dominated by FF Forest (59 of 74 projects), which partly explains the result. The SME and originator loan category comes mostly from Afranga and Debitum (251 of 319 projects).
The opposite trap: betting everything on the top scores
Following only the highest scores creates a different concentration. Of the 112 projects open or upcoming when we ran our queries, only 18 score 65 or more, and 13 of those 18 are energy projects on Enerfip, Lendopolis or Lendosphere. Two of those three brands belong to the same operator, and the average duration of energy projects exceeds 41 months.
A portfolio built on that criterion alone would therefore be concentrated in one sector, on two operators and on long maturities. Our reading: combining long energy projects with shorter buy-renovate-resell real estate (around 17 months on average) staggers repayments and lets you reinvest gradually. For more on the energy layer, see our report on renewable energy crowdlending.
The tax angle: diversify so you can use your losses
Tax gives another reason to diversify within the same regime. According to the BOFiP, the French tax administration's official doctrine, losses on crowdfunding loans (prêts participatifs) and minibons (for minibons subscribed up to 10 November 2023) can only be offset against interest from other crowdfunding loans or minibons, in the year the claim becomes definitively irrecoverable and the five following years, up to 8,000 euros per year [6]. The same text specifies that the offset reduces income tax, but not the base for social contributions [6].
The practical consequence: a loss is only useful for tax purposes if you also receive interest of the same kind. A sufficiently broad portfolio of crowdfunding loans creates that offset base.
One point remains open. The BOFiP paragraphs we consulted cover crowdfunding loans and minibons, and do not mention bonds [6]. We found no source confirming that losses on simple bonds can be offset against crowdlending interest. If your portfolio is mainly made of bonds, check this with your adviser. For the rest, the French flat tax (PFU) is 31.4% in 2026, made up of 12.8% income tax and 18.6% social contributions [20]. Amendments in the 2027 social security finance bill that aimed to raise the CSG on capital income were rejected in committee on 13 October 2026 [21]. On 23 October, the National Assembly rejected the whole bill at first reading, and the Senate is due to start examining it from 9 November [22]. Nothing is settled at this stage: one to watch. Details in our crowdlending tax guide.
An example method for 10,000 euros
The example below shows how to apply the four layers. It reflects CrowdPickr's own reasoning, not market consensus or personalised advice: we could not verify any numeric diversification rule from an established press source.
- Line size: 20 lines of about 500 euros.
- Borrowers: at least 15 distinct borrowers, and never more than 1,000 euros with the same developer or originator, across all platforms combined.
- Operators: three to five separately licensed operators, checked on the AMF white list. Lendosphere and Lendopolis count as one.
- Asset classes: a mix of energy (long maturities, high scores), buy-renovate-resell or development real estate (shorter maturities) and, if you wish, a limited share of business loans.
- Yield: projects at 13% and above must justify their premium with a decent score, not the other way round. Based on our data, they should not form the core of the portfolio.
- Staggering: invest over several months rather than all at once, to spread vintages. The AMF shows that delay rates vary sharply from one vintage to the next [1].
Finally, the prior question is still the size of the envelope. The AMF recommends diversifying your savings and investing "only sums you will not need in the near future", pointing out that these securities can be hard to sell quickly [4].
What our data does not tell you
Our 776 scores cover projects scored since 6 May 2026, not each platform's full history. The sample is uneven: 159 projects for Afranga, a single one for Anaxago. Above all, a CrowdPickr score is an assessment of project quality before investment, not an observed default rate. Scores help you avoid piling into weak projects; they guarantee nothing.
The CrowdPickr verdict
Diversify in this order: by borrower first, by platform operator second, by asset class third. Adding lines achieves nothing if they share the same developer, sit on brands that share one licence such as Lendosphere and Lendopolis, or all carry yields of 13% and above. Across our 776 scored projects, the average score falls with every yield bracket, and 92 of the 186 projects at 13% or more are rated "Avoid". Conversely, following only the top scores leads mostly into energy, with average durations of more than three years. A sound portfolio combines equal, modest tickets, at least fifteen or so borrowers, several operators checked on the AMF white list and staggered maturities. And it stays sized for money you will not need for several years, because a delay usually lasts longer than expected.
Sources
- L'Autorité des marchés financiers publie une analyse des performances affichées du marché du financement participatif immobilier, AMF, 20 January 2025.
- Liste des PSFP ayant fait l'objet d'un retrait d'agrément par l'AMF, AMF, updated 14 October 2026.
- White list entry: Lendocompany, AMF, consulted 3 November 2026.
- Investir via le crowdfunding par souscription de titres, AMF, savers' section, 15 November 2023.
- Financement participatif : l'AMF appelle les investisseurs à la plus grande vigilance sur les risques, AMF, 18 June 2025.
- BOI-RPPM-RCM-20-10-20-30, paragraphs 102 to 106, DGFiP, BOFiP, version in force since 15 May 2023.
- En crowdfunding immobilier, 1 projet sur 4 a plus de 6 mois de retard, mais retard ne vaut pas perte, JeChange, 18 October 2026.
- Crowdfunding immobilier : taux de défaut, retards... le bilan 2025, Fortuny Conseil (adviser blog), updated 10 April 2026.
- Homunity: projects in definitive loss, Cabinet Kohen Avocats, 4 October 2026.
- Crowdfunding immobilier : premier défaut de paiement de grande ampleur, UFC-Que Choisir, 19 January 2017.
- Crowdfunding immobilier : les pièges à éviter pour bien investir, Boursorama (Le Figaro content), 19 January 2017.
- Crowdfunding : la plateforme de financement participatif Tudigo placée en redressement judiciaire, MoneyVox (with AFP), 17 August 2026.
- Liquidation de Tudigo : un coup dur pour le financement des start-up ?, TourMaG, 16 October 2026.
- Crowdfunding, PEA-PME, issuer or platform liquidation, Cabinet Kohen Avocats, 3 October 2026.
- Crowdfunding Weekly Digest: October 26 to November 1, 2026, The Crowd Space, 2 November 2026.
- La Banque Postale a tranché sur l'avenir de sa filiale dédiée au financement participatif, MoneyVox, 5 November 2024.
- New European framework for crowdfunding, Jeantet, November 2020.
- Flat tax (PFU) : champ d'application et taux applicables 2026, LégiFiscal, updated 26 August 2026.
- Flat tax 2027, Club Patrimoine, situation as of 21 October 2026.
- PLFSS 2027 : rejet de l'Assemblée nationale, LégiSocial, 30 October 2026.
CrowdPickr data as of 3 November 2026. This article is an independent analysis and does not constitute investment advice.