Kvistlund Fund I closed in March 2021 at $35 million. It is now fully deployed. We've backed seven companies across the logistics stack — from predictive port intelligence to freight procurement to distributed fulfillment — and we've had three years of close partnership with founding teams across what has been, to put it diplomatically, a volatile period in logistics technology.
Writing a retrospective while portfolio companies are still operating and raising feels premature. But the lessons from Fund I have already shaped our Fund II thesis in concrete ways, and I think transparency about that thinking is useful — both for founders evaluating whether we're the right partner, and for our own LPs who deserve to understand how we're incorporating experience into judgment.
What we got right
The core thesis held. We invested in companies where ML is the product — where the predictive model or the optimization engine is what the customer is paying for, not a workflow that happens to run some machine learning in the background. That distinction turned out to matter more than we initially anticipated. The companies in our portfolio where ML is genuinely central to the value proposition have been more defensible through the market correction of 2022–2023 than their workflow-software counterparts. Customers who are using an ML-based product for the quality of its predictions are stickier than customers who are using a workflow tool that could be replaced by a cheaper alternative.
We also got right the conviction that European logistics AI was underfunded relative to the US. When we were deploying Fund I in 2021-2022, many of the best logistics AI founders in Europe were being overlooked by US-centric investors who didn't have the network or the domain familiarity to evaluate them. That created a real advantage for us. Three of our Fund I investments came to us through founder networks in Sweden, Finland, and the Netherlands — relationships that we had because of our operational backgrounds, not because we had an office in every market.
The operator-to-operator dynamic in diligence worked as designed. Founding CTOs who are building ML models for freight arrival time prediction need to have conversations with people who understand why ETA data from carriers is structurally unreliable, how port authority reporting lags introduce systematic biases, and what a real shipper actually does with a 4-hour early arrival prediction. We could have those conversations at the technical level, and it helped us diligence more accurately and build better relationships with founding teams post-investment.
What we underestimated
We underestimated how long enterprise logistics sales cycles would extend in a post-2021 market. Several of our Fund I companies were growing rapidly through 2021 into early 2022, with enterprise pilots converting to contracts and ARR compounding quarter over quarter. When freight markets corrected sharply in mid-2022, logistics operators pulled back on software spending. Enterprise procurement cycles that were running at 3–4 months stretched to 9–12 months. Pilots that were progressing toward contracts stalled as buyer organizations reorganized or froze budgets.
This wasn't unique to our portfolio — it hit the entire logistics tech sector. But we had deployed capital with assumptions about sales cycle velocity that were calibrated on the 2021 market. When that market evaporated, companies that had sized their teams and burn for faster growth found themselves needing to extend runway into a slower environment. We worked closely with affected portfolio companies on this, and the founders navigated it well. But the experience updated our default assumptions about enterprise sales timelines in logistics, particularly for products in categories where the buyer is an operations director rather than a CTO.
We also underestimated the integration complexity for some product categories. Predictive logistics tools that rely on carrier data integrations for their model inputs face a real operational challenge: carrier APIs are inconsistent, data formats vary, and integration maintenance is ongoing rather than one-time. Companies that estimated their data ingestion infrastructure costs at one level found that maintaining data pipelines across 30+ carrier integrations consumed more engineering time than expected. This is a solvable engineering problem, but it requires more investment in infrastructure than an early-stage seed valuation typically anticipates.
How Fund I shaped Fund II thesis
The most direct lesson from Fund I is a sharper filter on data pipeline maturity as a founding team capability. We now diligence explicitly on whether the technical team has experience building and maintaining production data pipelines at scale — not just building ML models in clean data environments. The best ML architecture in the world doesn't generate business value if the data flowing into it is unreliable or incomplete. This sounds obvious in retrospect. It wasn't always obvious in the due diligence conversations we were having in 2021.
We've also sharpened our view on go-to-market founder expertise. Several of Fund I's companies had technically excellent founding teams who underestimated the complexity of selling into logistics operator organizations. Logistics operators are skeptical buyers — they've been through multiple software implementations that underdelivered on promises, and they evaluate new software with a healthy distrust of demos. Founding teams that had operated in logistics understood this instinctively and knew how to run a proof of concept that earned trust. Teams that came from software backgrounds without logistics operations experience sometimes ran POCs in ways that didn't generate the kind of outcome data that logistics operators need to make a buy decision.
Fund II prioritizes founders who have a combination: ML or software engineering depth plus direct logistics operations experience. We've tightened this criterion. We've also added explicit evaluation of how founders think about the proof-of-concept motion — what outcome they're trying to demonstrate, over what time period, and how they plan to close from POC to contract.
On the vintage and what comes next
Fund I was deployed across a period that included the extraordinary volume and market conditions of 2021 followed by the sharp correction of 2022-2023. That vintage will produce interesting learning regardless of how the individual portfolio outcomes develop. The companies that survive the correction and emerge with durable customer relationships and efficient unit economics will be better businesses for having been tested in a difficult market. The ones that don't survive will have taught us something real about which assumptions were fragile.
We started deploying Fund II in early 2024, into a market that is more rational in its pricing than 2021 but more skeptical in its enterprise buying behavior. The thesis is tighter, the diligence is more operationally focused, and the founding team criteria are more specific. We're backing fewer companies than we might have in 2021, with more conviction on each one. That feels like the right response to what we've learned.