Seculus uses Machine Learning in Kepler to discover new insights in their data

Consumer Goods
Small data team with no prior AI or ML experience.

The Business

Seculus is a watch manufacturer that prides itself on contemporary design, versatility, and the attention to detail one expects from a high-end watch company — but at a mid-level price. The company has been in business since 1960 and began a portfolio diversification process in 2005 to bring its offerings to a larger audience who value both form and function when it comes to their timepieces.

The Problem

Seculus manufactures watches based on expected demand for each product. 

The in-house data team is expected to model nine months in advance which watch models would have the highest sales volume in order to optimize the supply chain processes that directly impact the company’s gross margins.

But Seculus had an extensive catalog with over 1,400 SKUs from more than 50 watch styles that all require different parts. Their traditional framework, based on statistical demand forecasting models, wasn’t providing the level of predictive accuracy the company needed to anticipate inventory volume. And given their large catalog of designs, it was functionally impossible to scrub and analyze that volume of data in a reasonable amount of time.

And, lacking any internal AI or ML resources, Seculus didn’t have the tools or resources to process the data to gather insights or make predictions in time to have any impact on the next retail season.

The Solution

Seculus quickly honed in on Stradigi AI and Kepler AI software. Teams on both sides worked together to develop prediction models based on previous sales and watch feature data. Stradigi AI’s data analyst worked closely with the product manager at Seculus, who had subject matter expertise and could validate the results of the AI models.

Here’s how Stradigi AI helped Seculus use its data to the fullest to uncover outcomes for more accurate predictions of future product demand:

  • Kepler AutoML with multiple AI technologies and pipelines ready to use off the shelf: used a tabular regression model to identify which factors drive demand for Seculus’ products, including trends in watch features and audience segments that correlated with higher sales volume.
  • Feature engineering with AutoML:  Seculus worked with the Stradigi AI team to perform data pre-processing to cleanse leakage issues and data values and combined inventory and retail sales data.
  • Developed and iterated prediction models in parallel: Up to ten models were run in parallel, and each model was evaluated with ease to get recommendations on improvements to model performance


The speed of model iteration enabled improved data processes across Seculus, including capturing better data from sales and factory teams, which led to better models and higher data quality for operations.

Feature insights from the models and individual predictions provided by explainability tools helped uncover correlations within the dataset.

  • Introduced website and brick-and-mortar retail store data: available data was upgraded for better accuracy and included in training new models to enhance SKU demand predictability.

The Results

With the help of Stradigi AI and Kepler AI software, Seculus has built a methodology and practice going forward that promises to provide a continued positive impact on the accuracy of their forecasting. Seculus didn’t just get a point-in-time improvement – the benefits will last a lifetime.

The first prediction model was ready within a week, and hundreds of models were built, analyzed, and learned from within six months. The data team at Seculus was able to build its own machine learning models — allowing them to undertake additional use cases to optimize processes over a larger part of their organization.

The speed at which Seculus learned from each model iteration helped the company make real and impactful operational changes that led to improved efficiency and improved data overall, paving the way for even better decision-making in the future.

Seculus discovered previously unknown insights from its data that identified which watch features were more popular than others and was able to predict which models were going to sell with greater accuracy.

With the new information, Seculus fine-tuned its inventory requirements to know which parts it needed to order in which quantity and was able to make data-driven decisions to reduce its catalog to a more manageable size – and the company was able to reduce its total SKUs by reconciling duplicates and similar products, allowing Seculus to more clearly understand its inventory requirements with less of a time investment.

Kepler helped us build our first prediction models without prior AI experience. Once we became familiar with how it works, we were able to quickly expand AI insights to additional use cases.

Oswaldo Moreira Neto
Director of Marketing & Innovation

days to produce their first AI model


DATA POINTS analyzed



Kepler helped us build our first prediction models without prior AI experience. Once we became familiar with how it works, we were able to quickly expand AI insights to additional use cases.

Oswaldo Moreira Neto
Director of Marketing & Innovation
This site is registered on as a development site.