A conversational, predictive layer that replaced MISAKO’s weekly manual reporting process with real-time access to sales, stock and forecasting data, queried in natural language and audited end to end.

It’s Sunday night at MISAKO. Someone would normally spend the next five hours consolidating sales and stock data into the spreadsheet the entire buying team depends on.

Today, that same person opens a single conversation instead: queries the data in natural language, checks which items are at risk of running out, and acts directly on forecasting, stockouts and store-to-store redistribution, all from the same platform, with every query traceable back to its source data.

Born in Barcelona, 1998

Founded in Barcelona in 1998 by the García and Torres families, also founders of Stradivarius, MISAKO has grown into one of Spain’s leading bag and accessories retailers, with over 220 stores across Spain, Portugal and France.

The company’s analytical and operational core ran on a highly granular Excel of sales and stock, regenerated every Sunday night in a process that took around five hours, feeding buying, replenishment and redistribution decisions across the entire network. No ad-hoc query was possible mid-week without routing through IT, and the decision of what to buy, how much, and when rested on expert judgment applied to an unsystematized process.

MISAKO didn’t partner with Timbal to replace their reference document, but to rebuild it as live infrastructure: a conversational, predictive platform running on the same store-management data the team already used, where any member of the organization queries the data in natural language. The Excel survives as an exportable view, but stops being the only path to information.

A MISAKO tote bag balanced on a glass vase

An Excel file, the company’s analytical core

Five hours every Sunday, static queries, and no real forecasting. The root causes:

Analytical latency. It was generated weekly. No ad-hoc query was possible mid-week without going through IT.
No forecasting. Orders six months out were closed by eye. The hit rate was high thanks to expertise, but the cost of errors was asymmetric, through devalued stock or stockouts before anyone could react.
Manual redistribution. Overstock in one store and stockouts in another coexisted with no shared solution across the network’s 226 stores.
Dependency on IT. Every new business query went through the analytics queue.

Together, these limitations made it impossible to anticipate stockouts, optimize buying for new designs, or react to campaign peaks. The direct consequence was devalued campaign overstock, lost sales from recurring stockouts on continuity products, and zero analytical autonomy for business profiles.

The modules elevating the bottleneck

The central buying document operates today in production as a set of pieces that reinforce each other.

The reference spreadsheet. The master table generated instantly for any season and commercial week, with seven predefined profiles (Management, Sales, Supply, Finance, among others) and full Excel export.
Conversational assistant over the data. A conversational layer over the same data, where any user asks in Spanish or Catalan and gets an answer in text, tables, charts, or a downloadable file, with a visible audit trail of everything it has queried.
Purchase forecasting. An engine that doesn’t give a single number, but a justified range (Prudent, Recommended, Ambitious, and Upside) based on analog articles from previous seasons, with visible trend and price elasticity, validated with continuous backtesting against purchasing decisions.
Network view and store analysis. A network dashboard tracks sales, stock, and a 30-day stockout prediction on the central warehouse, alongside a map with performance clustering (top, mid, low) and a per-store profile flagging products at risk.
Proactive continuity-product tracking. A shared table for continuity products projects weekly stock forward and lets a buyer simulate the impact of a purchase before committing to it.

Transformation in numbers

From ~5 hours every Sunday to instant and on-demand for generating the reference spreadsheet.
From hours or days to under 1 minute, in natural language for a new query response.
−30% of stockouts on continuity products against the current target.
Network-wide view with 30-day stockout prediction on the central warehouse.
Store-to-store redistribution with transfer proposals and net value, to accept or reject.
Purchase forecasting with a recommended range and traceable justification.

In MISAKO’s words

“The central buying document had become a major bottleneck: it took around five hours every Sunday to update, and there was no way to make changes during the week. Now, it has become a platform that responds instantly, making the ordering process faster and more accurate.
Most importantly, we can now react much faster to changes in demand, which is a key competitive advantage in the fashion industry.”
— MISAKO