The seven-day cycle. What Ukraine teaches manufacturers, not fighters

Three years ago Ukraine had about seven drone makers; today it has about five hundred. The number that matters most, though, is another one: the redesign cycle of a model, from front-line feedback to the new version in production, fell to seven days. War on the Rocks describes the mechanism: the operator who uses the drone writes to the engineer who built it, over Signal, and within days the change is in production. The cost per target hit went from about $60,000 to about $1,000.

On production totals, millions of drones a year, only Ukrainian government claims exist, not independently verified, and this piece does not use them. The seven-day cycle, on the other hand, is documented by several sources, and it concerns something any Italian factory can copy without a euro of investment: the distance between users and builders.

It is easy to read the Ukrainian case as a story about quantity. It is a story about speed. Five hundred small makers, each close to a unit, beat a few large makers far from the front, because the information about what does not work reaches whoever can change it in a day, and in a form that can be used: a photo, a message, a batch. Scale came afterwards, as a consequence; the short cycle came first.

In the first piece of this series we wrote that the cycle shortens first with data and only then with machines. Here the point is more precise: the cycle is made of four steps, who reports, who receives, who decides, who changes, and its length depends on how many desks the information crosses and how many times it is rewritten by hand along the way.

Five hundred makers close to the people who use the product beat a few makers far away. The edge lies in the cycle, and the cycle is made of desks.

Take the most common cycle in a supply-chain factory: a non-conformity. An operator finds it, writes it on a form or in an email, quality records it in a free-text field, the monthly meeting discusses it together with the other twenty, and a corrective action is decided once the cause is clear, which is often never. In the supply-chain companies we have met, the typical time between report and decision is measured in months, and nobody measures it.

The people are the same ones who close the cycle in seven days in Ukraine: attentive operators, competent technicians, managers who want to decide. What is missing is the form of the information. Free text cannot be counted, so it cannot be sorted by frequency, so nobody decides what to act on first. Oliver Wyman said it about supply-chain departments, “most do reactive problem-solving”, and it holds just as much for quality: you chase the latest report instead of the most frequent one.

Vertesa is the fictional company we use in this series, precision mechanics, 120 people, because no real client data can appear here. Its non-conformity log for the last quarter, which we publish as an open case with synthetic data, contains seventeen different wordings for six real causes. “Part out of spec,” “bent raw part,” and “arrived already out of size” are the same thing: a raw part out of tolerance, twenty-six times in three months out of ninety-six reports, the most frequent cause. In the log, though, it is almost invisible, because everyone wrote it their own way and the single most frequent wording is a different one, “paint runs,” which appears ten times.

The AI use case is obvious: group the wordings and propose the cause on new reports. But the value comes first, in the taxonomy. Quality decides the six causes in one afternoon, just as management decides the four information classes; without that decision the model groups at random, and with that decision even a spreadsheet could do the grouping. AI is there to do in an hour, on the history, what nobody would ever have done by hand, and then to keep it in order every day.

Cutting the cycle from months to weeks does not require software: it requires four written answers. Who reports, and through which channel it reaches the decision maker. Who decides, by name, with the mandate to stop the line. In how many days, with a measured target. With which data, that is a cause picked from a closed list instead of free text. The same four questions apply to the cycle with the customer: who reports a problem to us, who answers, in how many days, with which data.

Then the cycle gets measured, and the measurement goes into the one place where the company’s know-how and rules live, next to the supplier map and the six data points to answer within an hour. That is the point where the seven-day cycle stops being a Ukrainian story and becomes a number on the dashboard of an Italian factory.

1. [1]KSE Institute and Brave1, first comprehensive study of Ukraine’s drone industry. kse.ua
2. [2]Exponential View, “Ukraine’s seven-day drone advantage”. www.exponentialview.co
3. [3]War on the Rocks, “Inside Ukraine’s battlefield innovation loop”. warontherocks.com
4. [4]Forbes, Oliver Wyman, “European defense buildup may cause supply chain delays and shortages”, June 25, 2025. www.forbes.com