The $1 Test

SOSV/IndieBio’s 72% gross return on 2015 to 2023 LATAM investments is one manager’s portfolio-level number; build the per-dollar comparison bottom-up under a single taxonomy.

The rationale

The region’s headline return figure is four steps away from comparable. SOSV/IndieBio reported a 72% gross average return on LATAM deep tech investments made between 2015 and 2023, reported via the IDB rather than read from the manager. It is gross rather than net, portfolio level rather than fund level, cumulative rather than annualised, and from a single manager. Set against a net IRR benchmark it says nothing, and it is quoted as a headline anyway, including on this site’s own front matter.

The test asks what one dollar committed in a given vintage returned, in LATAM and in matched baskets elsewhere. It has to be built from deal-level records under one taxonomy, the IDB’s twelve verticals, with entry and exit dates, marks and realisations held per company. It cannot be built by differencing published regional totals: the report’s own cumulative bars, US$52bn for the United States, US$14bn Europe, US$13bn Asia and US$2.54bn LATAM, are three vendors stacked in one chart, the first three from Science|Business and the LATAM bar from an IDB total with a Sling Hub annual figure added on top, so differencing them measures the definitions.

The objection is that the data does not exist, and today it does not. European deep tech VC alone was US$20.3bn in FY2025 on Dealroom’s definition, while Tracxn puts US deep tech at roughly US$179bn in 2025 on a much broader one, so the two cannot share a basket at any level of aggregation. That is why the test runs bottom-up on the company records the funders registry and the data commons collect, and why every release reports coverage per basket before it reports a return.

The evidence

First moves

1. 0–6 months

   Write the protocol before the number

   Publish the $1 test protocol, vintage windows, the IDB twelve-vertical taxonomy, inclusion rules, marking convention, and the explicit exclusion of published vendor totals as inputs. Circulate it for comment before any basket is assembled.

   LADP research team + IDB Lab
2. 6–18 months

   Assemble the LATAM basket from company records

   Build the LATAM leg from deal-level records held in the funders registry and the data commons, carrying entry date, amount committed, and current mark or realisation per company, and publish coverage as a share of the region’s mapped deep tech population.

   Coalition secretariat + LAVCA + national development banks
3. 18–36 months

   Match the comparison baskets and report with bands

   Construct US, European and Asian baskets on the identical taxonomy and vintage windows from licensed deal-level data, and publish MOIC per dollar committed with confidence bands and per-basket coverage.

   LADP research team + university research partners

How we would know it worked

• Coverage of the LATAM basket as a share of the mapped regional deep tech population, published per release, above 60%.
• Share of basket value that is realised rather than marked, reported per vintage, rising over time.
• Comparison baskets built on the identical taxonomy and vintage window, counted per release, reaching three.
• Methodology comments received and resolved before each release, logged publicly in the protocol repository.

What could go wrong

Who acts

Rests on

The data behind it