Anyone buying hazelnut semi-finished products is really buying one thing: the certainty that the product will be the same six months from now. You may like a flavour or not, but a flavour that shifts mid-season forces you to redo recipes on the fly.
This page explains how we guarantee that. Not with adjectives — with what we actually do, which can be verified and, if you like, seen in person.
Before the raw material: who we take it from
We have grown hazelnuts in Avella, Irpinia, since 1940. When we need to supplement our own harvest we turn to selected local growers — but the criterion isn’t that they’re nearby.
We don’t buy hazelnuts from anyone who won’t show us how they work.
Being local isn’t enough: if we don’t know a grower’s harvesting, drying and storage practices, we don’t buy. Because a hazelnut mistreated upstream can’t be recovered downstream, however well you roast it.
Foreign hazelnuts: never. It isn’t a commercial preference, it’s a limit we don’t move even when demand exceeds supply. We’d rather say no.
Calibre: where half the price comes from
We use only hazelnuts of 11/13 and 13/15 mm calibre, never below 11. Never broken ones.
Undersized and broken hazelnuts cost far less, and anyone using them can sell for less. But it isn’t only a matter of cost:
A whole hazelnut is almost spherical, so the centre is equidistant from the inner surface: heat travels through it evenly and the roast reaches everywhere the same way. A broken hazelnut doesn’t have that geometry. Its inner fibres are exposed, and roasting it is like cooking them directly. With more surface in contact with air it also oxidises sooner: part of the aroma is gone before the nut even reaches the roaster.
From there the arithmetic is simple: a lost aroma has to be compensated for, and the quickest way to compensate is to add it. We prefer not to have to, which is why we start from whole, calibrated nuts.
What we check before switching on the roaster
Every batch goes through five checks, and none of them is a formality:
- Visual inspection of the incoming product
- Sorting, optical or manual, to remove defective nuts and foreign bodies
- Moisture of the hazelnut
- Temperature of the hazelnut as it enters
- Ambient temperature and humidity at the moment of roasting
The last three aren’t bureaucracy: they change the parameters. A hazelnut entering warmer needs a different curve from one entering cold. Roasting the same way regardless leads to results that vary from one day to the next — not through carelessness, but because the starting conditions weren’t the same.
We don’t invent the parameters each time
Since 2013, from our very first roast, we have recorded the data of every single batch. Every roast we do adds more.
That archive is where the preset comes from: the starting parameters calculated from cultivar, calibre, hazelnut temperature and moisture, ambient temperature and humidity. It isn’t a fixed setting copied from a manual — it’s thirteen years of roasts accumulating.
The system keeps training itself. Every batch makes it more precise than the last.
The roaster is stabilised before we start
A detail almost nobody talks about, and one that on its own explains a lot of batch-to-batch variation.
The roaster is preheated until it is thermally stable before the first batch. We don’t start as soon as it’s hot: we start when it’s steady.
The reason is simple: if you begin while the oven is still climbing, the first roast of the day isn’t the same as the last one. By stabilising first, every batch in the session works in identical conditions.
And the first three roasts of every session are always tasted before we go on. The data says whether the parameters are right; the tasting confirms it.
Particle size: we declare the filter, not the machine
Fineness of grind is easy to misread. Many suppliers state “under 20 microns”, but that figure is usually the refiner manufacturer’s spec: it says what the machine can do, not what’s in the tub. A paste contains particles of different sizes, and a single number doesn’t describe that spread.
We state below 25 µm because 25 microns is the size of our filter: a threshold nothing coarser gets past. Our pastes are 25 microns and down.
It’s a number that looks higher on paper. It’s the only one we can guarantee.
Why the flavour doesn’t shift mid-season
Every harvest gives different hazelnuts: this is agriculture, not chemistry. Consistency doesn’t come from pretending harvests are identical — it comes from building the blend deliberately.
Each production is composed by balancing calibres and cultivars so as to arrive at the same aromatic profile, even when the starting material has changed. It’s invisible work: the customer doesn’t see it, they only notice that the product is always the same.
For the same reason we don’t declare a single calibre on the label: our pastes are a blend, and one number would be a made-up number.
We roast to order, all year round
We don’t transform the whole harvest at once and then empty the warehouse over the following months. We roast when there’s production to do, throughout the year.
The benefit is direct: what you receive isn’t aged stock. We keep consistently fresh product in store, and if an order exceeds what’s on hand we roast it and ship within a few days.
If a batch doesn’t convince us
With this method it rarely happens — that’s exactly what the method is for. But it happens.
When it does, that batch doesn’t reach the customer. It becomes a household batch: we keep it for ourselves and for the people who work with us.
It isn’t a noble gesture, it’s a consequence: if you’ve built everything on consistency, sending out a sub-par batch undoes thirteen years of work.
Who answers when there’s a problem
If you write with a technical question about a recipe, Marco answers, the same day.
Not a call centre, not a form, not a ticket: the same person who stands at the roaster. It’s the advantage of being small, and we have no intention of losing it as we grow.
What we refuse to do
The quickest way to understand a supplier is to ask what they don’t do.
- ❌ Foreign hazelnuts. Never, at any volume.
- ❌ Growers whose practices we don’t know. Nearby isn’t enough.
- ❌ Added flavourings. Not even on request.
- ❌ Dropping the calibre to hit a price. If costing less is the requirement, the answer is no.
- ❌ Promising deliveries we can’t meet. Better a no than a delay.
And the price?
On price we hold one conviction, and it cuts both ways:
Don’t pay more for something worth less. And don’t pay less only to spend twice as much in work.
A number per kilo, on its own, doesn’t tell you whether a product is worth it. What tells you is what happens next: how many times you have to rebalance a mix, how much product goes unsold because the flavour wasn’t the same as last time, how long you spend chasing an answer that never comes.
Ask for the price list and compare, by all means. We only ask that you compare the same things: hazelnut percentage, calibre, particle size, consistency from batch to batch, and who answers when you have a question at seven in the evening.
If you have a question about a recipe, ask before you order. We’ll answer anyway.