Micolet scales its business and reduces its costs thanks to Sciling

The rise of e-commerce and the growing consumer sentiment towards sustainability have caused secondhand clothing sales to boom. Micolet, a Spanish startup company founded in 2015, has profited from this situation year over year since its inception. However, they recently experienced bottlenecks in their production process that threatened to slow down this growth


Micolet’s value proposition for sellers is convenience. Unlike other secondhand eCommerce sites, they take care of the whole selling process, from the moment products are collected at seller’s homes up to the moment they are delivered to buyers. With a permanent stock of 180,000 unique SKU’s, this is quite an operational challenge.

To get an idea of the impact of this number has in the operations, we can compare it with the number of unique SKU’s handle by one of the biggest retailers in the world: Inditex. In his 2018 annual report the Spanish fashion giant reported having designed 75,000 SKU over a whole year for their entire universe of brands. “The number of different articles we have to keep track of and handle is so vast that any improvement we introduce in the product lifecycle has a real impact for the business”, mentions José María del Moral, one of the two founding partners.

About the client:

Micolet is a Spanish website for buying and selling second-hand women’s clothing born in 2015 and currently employs 50 people and “saves” 60,000 clothes and accessories per month and 8 tons of CO2 per year.

The challenge

Co-founder Aritza Loroño realized that photo retouching processes outsourced to freelancers around the world were weighing down the entire production chain. They never knew for sure when the pictures would be ready, and that unpredictability made the process non-scalable. This had serious implications over their growth prospects. Controlling this process was vital for the continuity of the company. They should be able to scale their business or they could die one day from too much success.

“The problem with leaving this process in the hands of external personnel was the lack of control. More often than not regional holidays, personal trips and other absences caused freelancers to be late in their deliveries and they even didn’t let us know in advance so we could take corrective action. This generated delays in product uploading on the website what, in turn, translated into sales”- Loroño commented.

Not only this, but also costs generated by photo retouching services increased as business grew. “In this kind of business profit margin for most products is as low as cents and large sales volumes are needed to cover fixed costs, so each cost must be controlled”.

Loroño and his partner, José María del Moral, began to consider options. Seeing the advances taking place in the field of Artificial Intelligence, they decided to opt for this option. After talking with many suppliers, they chose Sciling as a strategic partner. “Artificial Intelligence allows us to grow profitably while assuming little risk. The time has arrived to go faster and grow as much as possible,” mentioned Loroño.

Strategic approach:

After conducting an in-depth analysis of the situation together with Micolet’s management team, it was concluded that the photo retouching process needed to be automated in order for the business to continue to grow.

Our solution

The solution developed by Sciling, which applies a mask to the original image to center it and completely eliminate the background, achieves even better results than those obtained manually. With a background removal success rate of 99% and the possibility of executing it in real time so as to make any necessary adjustment as photos are taken, the algorithm has led to significant improvements in Micolet’s daily operations.

“With 160,000 pics a month and counting, it has been a very profitable investment. We have recovered the investment in just a couple of months. What’s more, we paid 6 cents per pic, so in a year we’ll have our operative expenses reduced by 115.000 €. This money can produce much more value in other areas of the company. And much more important than this, the new process is totally scalable and easy to manage. A workload increase with the previous system involved adding administrative costs from our side. A workload increase with the new system is completely transparent for us”, comment the company partners.

They couldn’t have known it at that time, but some months later this scalability proved to be critical. The COVID-19 pandemic has caused Micolet to grow exponentially, reaching a 60% increase in sales. “Since the beginning of the quarantine we have been receiving 10,000 garments a day, twice the quantity we received before. And people are also starting to warm up to the idea of buying secondhand”, says Oratz Elezcano, Head of Communications.

We can affirm, without fear of making a mistake, that the improvements introduced by Sciling in the production process have allowed Micolet to take advantage of the boom experienced by the sector. In addition, they have also allowed them to achieve their goal of selling clothing and accessories in the shortest possible time and at the most competitive price, positioning themselves as a global reference in circular fashion at the same time.


Successful retouching


Months to recover the investment


Months for concept testing

Applying AI to our processes allows us to grow profitably and with low risk

Aritza LoroñoFounder

"It was a pleasure working with Sciling and their team. They are clearly experts in their field. Their transparency and working philosophy gave us confidence. We knew the project would be successful from the very start."

José María del MoralFounder Micolet

Communications with Sciling were fluid throughout the entire project, but especially during the deployment phase, when we had his help at all time

Raúl RodríguezChief Technology Officer

The implementation process

A first approach to solving the problem was made using traditional techniques of Computer Vision but this didn’t produce the expected level of precision. After this, Sciling’s engineers relied on state-of-the-art Deep Learning techniques to create a tailored solution to meet Micolet’s needs. 

“We started this project with a very delimited proof of concept – says Germán Sanchis, CEO of Sciling. After that initial phase both Micolet and us realized we’ll had to resort to very new techniques in order to obtain desired results. Techniques so recent that there’s still much research involved. If Aritza and José Mari hadn’t been so open to trying new things, we couldn’t have gotten so far”.

“Sciling worked with us in a way that gave us confidence. Even when this proof of concept showed us there was still a lot of job to be done from both sides, we could see the shoreline beyond the rough waters ahead. We knew the project was going to be a success”, says del Moral.

Technologies used

Computer Vision

For a first approach to the problem

Deep Learning

To improve results

Keras & OpenCV

To develop the model

Why us?

  • We use state-of-the-art techniques to solve operational challenges.
  • We have extensive experience in Machine Vision technologies
  • Our three-step work model -analysis, proof of concept, implementation- minimizes risks.
  • We have a team with more than 15 years of experience in Machine Learning.