Aquaculture suppliers are being asked to answer harder questions, faster. Where does this product actually work best? Under which conditions does it underperform? How confident are we in the claims we put in front of farms and regulators? Traditionally, the only way to get those answers was a new trial, a new grant, or another multi-year project that competed with everything else in the R&D pipeline.
Meanwhile, farms are quietly generating exactly the kind of data those questions need: daily records of treatments, mortality, feed, growth, welfare outcomes, and harvest quality. Most of it never makes it into structured analysis.
Observational studies are how you close that gap. Instead of waiting years for new experiments, you use the data that already exists across farms to understand how products and biology behave in the real world.
Academic and internal field trials are still essential. They’re the places where mechanisms are tested, doses are refined, and regulatory dossiers are built. When you need to prove that a vaccine triggers a specific immune response, or that an in-feed treatment reaches a target concentration, you need controlled conditions and a clean experimental design.
The problem isn’t that trials exist; it’s that they’ve been asked to carry the entire weight of product understanding.
Trials take time. A multi-site program can easily run across several production cycles and 2–5 years of calendar time. They’re narrow. Even the best-designed study usually involves a handful of farms, cages, or cohorts, often under conditions that don’t reflect the full spread of environments your customers operate in. And they’re expensive to repeat. Each new question often means another round of negotiations, logistics, and capacity constraints on the farm side.
Most importantly, these types of research tend to answer a specific question under specific conditions: Can this work here, like this?
What they don’t easily tell you is: How does this behave across hundreds of sites, over a decade, with different management strategies, lice pressures, or temperature profiles? That’s not a failure of trials. It’s a mismatch between what they’re built to do and what suppliers increasingly need.
Observational studies fill in that missing layer. Instead of assigning treatments and watching what happens, you reconstruct what already happened across farms and use structured analysis to tease out patterns: where products work, where they don’t, and what conditions matter most.
In practice, an observational study is simple to describe:
You take real-world farm data (production histories, treatments, feed deliveries, environmental records, harvest results) and you compare groups of fish that experienced different products or strategies under natural farm conditions. No new cages are set up. No fish are assigned by a researcher. You’re analyzing decisions farms already made.
If you’re used to running common garden experiments, you can think of this as rebuilding many “gardens” after the fact: identifying cohorts that shared similar environments but differed in the product, dose, or strategy they received, and then comparing how they performed.
That doesn’t mean it’s just “looking at a spreadsheet”. The value comes from how those histories are reconstructed and compared.
The first step is grouping fish into coherent populations. A population might be a smolt batch through its entire seawater phase, a generation on a specific site, or a more complex lineage traced through multiple movements. What matters is that you’re following real cohorts through time, not just averaging a site’s numbers for a year.
The second step is defining exposures: which populations experienced which products, doses, timing, or combinations. One group might have used an in-feed lice treatment as a primary tool, another might have relied more on mechanical delousing, another on bath treatments. One set of farms might be using higher EPA/DHA levels in feed, others closer to conventional formulations.
The third step is comparing outcomes across those exposure groups while controlling for the obvious confounders: geography, temperature, lice pressure, smolt type, stocking density, and so on. You’re not pretending this is a randomized trial. You’re acknowledging the messiness and using statistics and data science to reduce bias rather than ignore it.
The result isn’t a laboratory verdict. It’s something closer to a field map: a view of how products behave across a large portion of the industry, under the conditions that actually exist.
You can run observational studies on small datasets, but their real power appears when the data becomes dense enough. Most controlled EPA/DHA feeding trials, for instance, have historically involved hundreds to a few thousand fish, sometimes a single commercial farm with hundreds of thousands of fish. Those studies are valuable and often necessary. But they are, by design, narrow windows.
A high-density observational study is a bit like taking a common garden experiment and spreading it out across an entire country. Instead of a single site where multiple products are tested side by side, you’re effectively comparing thousands of “gardens” at once – farms and cohorts that share similar profiles but use different products or strategies – and reading the signal across all of them.
Manolin’s work with Veramaris is a good example of what changes when you scale observational research up.
In Phase 1 of the Veramaris–Manolin program, we analyzed around 232.6 million fish across 99 Norwegian farms over a ten-year period. In Phase 2, that expanded to roughly 430 million fish across 166 farms and nine production zones. Instead of asking “What happened on one farm in one trial?”, the question became “What happens across a significant share of Norway’s salmon production when EPA/DHA levels change?”
At that scale, patterns start to stabilize. You can see not only whether higher EPA/DHA diets are associated with better outcomes, but how consistent those outcomes are across regions, seasons, and management styles. You can start to ask second-order questions: under which conditions does the benefit grow or shrink? Where does the product clearly shine, and where is it marginal?
High-density data doesn’t mean ignoring data quality. It means you have enough observations to model noise, handle missingness, and still get robust signals. You don’t need perfect coverage. You need enough structured data, across enough populations, to ask the right questions honestly.
In our full whitepaper on high-density observational research, we walk through how this scale translates into real project advantages, compressing timelines from years to months, cutting study costs, and giving suppliers industry-wide evidence instead of single-farm anecdotes. The paper also breaks down a full Veramaris case study, including study design, data volume, and the impact on survival, eFCR, and quality. If you’re debating whether your next big research trial should even exist, read this before you commit the budget.
For most suppliers, the practical question isn’t “Do observational studies exist?” It’s “When is this actually relevant to us?”
A simple way to frame it is through the product lifecycle:
Early on, academic and internal R&D work establish the scientific basis of a product: what it does, how it should behave, what mechanisms are involved.
Internal trials and pilots then refine formulations, dosing, and basic field performance on a small number of sites.
After that, the product enters the real world. It’s used by different farms, in different regions, under different regulatory and biological pressures.
This is the point where many suppliers end up flying partially blind. A product may have solid internal data and positive anecdotes, but little structured evidence of how it performs across hundreds of farms and cycles.
This is exactly where Manolin and Harpoon fit. Once internal R&D has defined or piloted a product and you either want to pressure-test it in the field before committing to scale, understand performance gaps and usage patterns to guide future R&D, or sharpen how the product is positioned with customers, a high-density observational study becomes the right tool.
By running those studies on top of existing farm data, Manolin turns this part of the lifecycle into a continuous learning loop: proving performance at industry scale, identifying where and how products work best (or don’t), and feeding those insights back into product strategy, R&D roadmaps, marketing, and ongoing customer support.
Every study looks slightly different, but the pattern is consistent.
It starts with questions. A feed company might want to understand how a functional diet affects eFCR and survival in high-lice-pressure regions compared to lower-pressure ones. A health company might want to see how an in-feed treatment performs relative to mechanical delousing across temperature bands. A genetics supplier might ask under what farm conditions certain families deliver the most robust performance.
From there, the relevant data pool is defined: which farms, which years, which zones, which populations. Manolin pulls the necessary production, treatment, feed, environmental, and harvest records from its network and partner integrations, and then traces populations through time so that comparisons are made between like-for-like cohorts.
The analysis layer sits on top of that structured base. Benchmarking quantifies performance differences. Descriptive statistics and significance testing reveal whether those differences are stable or just noise. Cluster profiling highlights profiles of success – the combinations of conditions under which a product is consistently associated with better outcomes. Forecasting extends those insights into the future, helping teams understand how products are likely to behave under upcoming cycles or in new areas.
The final step is translation. R&D needs to see the full technical story. Commercial teams need a clear narrative and a set of honest, defensible messages for customers. Leadership needs to understand where to invest, where to pull back, and where to double down.
The outcome isn’t just “a study”. It’s product intelligence: a shared, data-grounded understanding of how a product behaves in the field and how that should guide decisions.
Not every question needs a high-density study. There are still problems best solved in tanks, on a few carefully selected sites, or through small pilots.
But if you recognize yourself in any of these situations, you’re in observational territory:
You’re preparing to scale a product and want field evidence beyond your initial trials.
You have a mature product and suspect performance varies by region, farm type, or strategy, but don’t have a clear map of where and why.
Your commercial and marketing teams need stronger, data-backed claims to stand on in conversations with farms and regulators.
R&D is planning the next generation of a product line and needs to understand how current usage patterns and outcomes look across the industry.
In those cases, the data your customers generate every day is no longer just exhaust. It can be the foundation of a research program that matches the scale of the industry you’re operating in. If you’re nodding at any of these, the next step is usually a feasibility conversation: what data exists, what questions you want answered, and whether a 3–4 month observational project can replace or complement a planned trial.
Manolin’s role is to make that possible by removing data barriers on the front end. We bring the data network, the tracing and modeling infrastructure, and the study design experience. You bring the products, the hypotheses, and the decisions that need to be made.
Put simply: observational studies are how you stop guessing how products behave in the field and start knowing.