Kvistlund's mandate is AI-native logistics software in Europe. Within that, we make specific bets about which problem areas are ripe for seed-stage investment right now versus which are earlier or later than our fund's stage focus. This piece is an attempt to be public about how we're thinking about freight technology specifically — the movement of physical goods by road, rail, ocean, and air, and the software intelligence layer that sits on top of that movement.
Freight tech is a sprawling category that encompasses freight brokerage platforms, TMS vendors, visibility networks, carrier management software, customs and compliance automation, freight procurement tools, and port operations software. Not all of these are interesting to us at seed stage in 2022. Writing down the ones that are — and the ones that aren't — is useful for founders who are evaluating whether to bring us into a process.
Where we are actively sourcing
The area we are most actively sourcing in is predictive freight visibility. The existing freight visibility market — multimodal shipment tracking aggregated from carrier APIs — has been served since the mid-2010s by first-generation visibility platforms whose core product is event aggregation: getting status updates from 100 carriers into a single dashboard. That problem is largely solved. The next generation is predictive visibility: using the aggregated event stream as training data to build models that predict exceptions before they manifest.
Predictive ETA accuracy on ocean freight is a genuinely hard ML problem. Ocean vessel tracking data from AIS is high-quality and publicly accessible, but port dwell time prediction requires integrating port authority data, berth scheduling information, and congestion models that are neither standardized nor consistently available across the ports that matter most for European trade lanes. The companies that are building proprietary data pipelines for port-level prediction — not just vessel-level ETA from AIS — are doing something defensible. We have looked at five companies in this space since Fund I closed and are in active diligence on two of them.
The second area where we are actively sourcing is freight procurement intelligence. European shippers who are procuring road freight through broker networks or direct carrier relationships spend considerable time benchmarking rate environments. Most of that benchmarking is still done through manual rate requests and broker relationships, producing rate information that is days or weeks stale by the time it influences procurement decisions. Companies that are building spot rate intelligence using real transaction data — not surveys or broker estimates — are addressing a problem where the information asymmetry between large shipper procurement teams and the carriers they are buying from is large and measurable.
We are not interested in general freight brokerage platforms. That market in Europe has several well-funded incumbents and the differentiation story for a new entrant is difficult to tell at seed stage. The specific wedge of rate intelligence using ML on transaction data is interesting; the broader freight brokerage TAM story is not.
Where we are watching but not yet investing
Customs and trade compliance automation is a category we think will be important but where the timing for seed investment in Europe is complicated. Post-Brexit customs processing for UK-EU freight has created genuine pain and a real demand signal for better customs documentation automation. The problem is that the regulatory environment is still settling. Software that automates customs workflows under one regulatory regime may need significant rework as that regime evolves. We want to see more stability in the post-Brexit customs framework before we commit seed capital to a company whose product is tightly coupled to its current state.
Carrier capacity management and dynamic pricing is another area we find interesting in principle but difficult in practice at seed stage. The carriers who would most benefit from dynamic pricing infrastructure — small and mid-size road carriers in fragmented European markets — have the least appetite for software procurement. Selling dynamic pricing software to a 20-truck haulier in Poland or a 50-truck regional carrier in the Nordics requires a sales and implementation motion that is expensive relative to the contract sizes those customers generate. The category will be important, but the go-to-market constraints favor well-capitalized companies over seed-stage entrants.
Where we are not investing
We are not investing in hardware-adjacent freight tech at seed stage. Autonomous long-haul trucking, robotics for freight handling at distribution centers, and sensor hardware for cargo condition monitoring are all areas where the CapEx requirements and regulatory pathways are incompatible with seed fund economics. We are happy to discuss these categories at length because our backgrounds make us informed about them, but we will not write a check.
We are also not investing in freight marketplace platforms whose primary differentiation is aggregated carrier capacity without a prediction layer. The freight marketplace as a category has been well-funded globally, and the incumbents in European markets have sufficient network effects that a new entrant without a genuinely differentiated AI-based matching or pricing capability is unlikely to build a defensible position at seed stage.
We are not saying the non-ML freight marketplace opportunity is zero. Some of those businesses have built substantial revenue. We are saying that the specific edge Kvistlund brings as an investor — operator backgrounds in logistics AI, technical diligence capability, and portfolio company relationships — is most useful for founders who are building ML-native products, not for founders who are primarily building marketplace liquidity.
A note on ocean versus road freight
Our portfolio to date has been weighted toward road freight, which reflects our team backgrounds — Erik comes from Nordic 3PL road operations, and my background is in warehouse automation and ground-level distribution. Ocean freight is a category we're spending more time in heading into 2022, partly because the supply chain disruption of the past two years has made ocean visibility and port intelligence a higher priority for European importers, and partly because the data infrastructure for building ML-based ocean freight products has improved meaningfully.
Ocean freight presents different data challenges from road freight. Road freight data comes primarily from telematics units on trucks, mobile carrier apps, and TMS event streams — sources that are variable in quality but increasingly standardized in format. Ocean freight data comes from AIS vessel tracking, port authority systems, liner carrier APIs, and customs filing feeds — sources that are higher quality at the vessel level but harder to integrate at the port and terminal level. The companies building proprietary data pipelines for port-level prediction and dwell time modeling are doing the harder technical work, and in our experience, harder technical work with good data creates more durable product advantages.
Founding team signals we respond to in freight tech
In freight tech specifically, the founding team signals that make us want to move quickly in a diligence process are fairly specific. We respond strongly to founders who have built data pipelines in production against real carrier or port data systems — not founders who have designed a pipeline architecture, but founders who have actually integrated with carrier EDI systems or port authority APIs and understand where the data quality problems are. That operational data experience is much harder to acquire than general ML engineering skill, and it is the thing most directly predictive of whether a freight visibility or freight intelligence product will work in production.
We also respond to founders who have strong opinions about their initial customer segment and can explain why that segment is the right beachhead for their product — why a specific category of shipper or a specific trade corridor creates the right early customer profile for building a ML model that generalizes. Founders who are vague about early customer selection and plan to sell to anyone who will buy are less interesting to us than founders who have a specific thesis about which customer's data problem is the right starting point.
If you are building in European freight tech and you think we should talk, reach out directly. We move quickly when we find something that fits this thesis.