Internal Rate of Return
A LATAM benchmark, starting with what can actually be measured
Deep tech funds returned 17% net IRR against 10% for traditional tech across Europe and North America; LATAM has no comparable series and almost no exits, so build the benchmark on gross MOIC first.
The rationale
The comparison the region is measured against is precise, and the region’s side of it is empty. McKinsey found an average net IRR of 17% for deep tech funds against 10% for traditional tech funds, seven percentage points, or roughly 70% better in relative terms, across 115 deep tech funds and 1,572 traditional funds with vintages from 2003 to 2020. LATAM has nothing on that basis. The one figure in circulation, SOSV/IndieBio’s 72% gross return on investments made between 2015 and 2023, is gross, portfolio level and from a single manager.
Net IRR needs distributions, and LATAM deep tech has almost none: Dealroom recorded no Series B or later round closing in the region in 2024. A first-pass benchmark therefore reports gross MOIC and TVPI on marked positions by vintage and stage, with the marking convention published, and defers net IRR until a realisation history exists. It is one leg of the KPI framework’s standardised core, alongside graduation rates and CVC participation, so it should be built inside that framework rather than beside it.
The objection is that a gross, largely unrealised benchmark is not a benchmark, and it is partly right. Marks in an illiquid market carry the last round’s optimism as much as value. The containment is disclosure: report the marking basis, the share of value still unrealised, and the dispersion across managers instead of one regional headline. Two LATAM outcomes are already public and priceable rather than marked, Establishment Labs at roughly US$2.59bn of market capitalisation on 4 July 2026, up from the US$1.8bn the report carried, and Satellogic on about US$249m raised across seven rounds. A benchmark that cannot yet compute a net IRR can still anchor itself on outcomes a public market prices every day.
The evidence
What it unlocks
A defined fund universe and a gross return series LATAM can actually report, and a net IRR benchmark the day the region has exits to compute it on.
It has been done before
France, the Tibi initiative
Tibi shows that a public sponsor can build a fund-level dataset where none existed. By homologating funds against published criteria, France reached 160 homologated funds and €14bn invested by February 2026, and raised its target from €13bn to €15bn as it opened a European phase. The homologated list is a defined, named universe measured over time, which is precisely what a LATAM return benchmark lacks.
First moves
- 0–6 months
Define the fund universe before the metric
Publish inclusion criteria for a LATAM deep tech fund universe, vintage, mandate and minimum deep tech share of portfolio, and name the funds that meet them, the way Tibi defined its universe through homologation.
LAVCA + IDB Lab + coalition secretariat - 6–18 months
Collect gross MOIC and TVPI by vintage
Run a standardised annual reporting template collecting gross MOIC, TVPI, marking basis and unrealised share by vintage and stage, published anonymously in aggregate with dispersion rather than as a single regional average.
LAVCA + national development banks as LPs - 18–36 months
Publish the realisation series, then switch on net IRR
Track exits, secondaries and write-offs as they occur, and publish net IRR only once a stated minimum of realised value exists, using McKinsey’s definition so the comparison is like for like.
LADP research team + LAVCA
How we would know it worked
- Number of funds in the defined LATAM deep tech universe filing the annual template, rising above 25.
- Share of reported portfolio value that is realised rather than marked, published per release, rising from near zero.
- Series B or later rounds closing in LATAM deep tech per year, from Dealroom, rising above zero and reported annually.
- Interquartile range of gross MOIC across reporting funds, published each release, so the region is never represented by one average.
What could go wrong
Only the winners report
A voluntary template collects returns from funds that like their numbers, and the benchmark then flatters the region into a claim nobody can repeat. Contain it by publishing response rates and coverage as a share of the defined universe, and by having development-bank LPs make template reporting a condition of their commitments.
The benchmark is used before it means anything
A gross, largely unrealised MOIC will be set against McKinsey’s net 17% by someone raising a fund. Contain it by labelling every published figure with its basis, refusing to publish a net IRR before the realisation threshold is met, and printing the non-comparability on the same surface as the number.
Who acts
Rests on
02 Metrics & Success Stories A deep tech KPI framework
14 Research & Visibility An open data commons and a funding catalogue
16 Extended Funders Mapping A versioned registry the community can audit
The data behind it
Hello Tomorrow’s European dataset finds deep tech ventures fail no more often and exit no slower than regular tech. Whether their exits are larger is still inconclusive.
SourceHello Tomorrow, Deep Tech Investor Mapping (LATAM) & The 2025 European Deep Tech Report
| Category | Value |
|---|---|
| Do deep tech companies fail more often? | No |
| Do they need more time to exit? | No |
| Do they have larger exits? | Inconclusive |