QuantUp

BACKGROUND

Please describe your company and your position there.

I work for a company operating in the field of automation and new technologies for elevators. I hold the position of the President of the Management Board.

OPPORTUNITY / CHALLENGE

For what projects/services did your company hire QuantUp, and what were your goals?

Our vision is to provide non-contact, automatic devices. We are determined to change the way lifts are used. We hired QuantUp to develop algorithms that will help us achieve our business and social goals.

SOLUTION

How did you select this vendor and what were the deciding factors?

We have been preparing for the project for 2 years. We spoke with several service providers. We chose QuantUp due to the excellent understanding of our needs. The project was a challenge but only their team took care of every detail to find out what we really need.

Describe the scope of work in detail, including the project steps, key deliverables, and technologies used.

Our project consisted in building a database on the basis of which advanced prediction algorithms were developed for us. During the development phase of the project, they advised us on the best solutions. We verify the software test took place during the implementation in the device. In this phase, they showed the value in the form of a smooth start-up of the system and making quick necessary adjustments in the software!

How many people from the vendor's team worked with you, and what were their positions?

High-class Data Scientists developed algorithms. The project was coordinated by PhD Engineering. The installation was done by IT specialists.

RESULTS & FEEDBACK

Can you share any measurable outcomes of the project or general feedback about the deliverables?

The team completed the works in the assumed period of time. The deployment was quick and all corrections were made immediately. It was important for us to complete the project at the MVP stage and, most importantly, to stay within the assumed budget.

QuantUp met our expectations and even exceeded them. We received added value in the form of support at the implementation stage.

Describe their project management style, including communication tools and timeliness.

Cyclical meetings coordinating the project allowed for the exchange of information and checking the progress of work. Communication was clear and specific. QuantUp was looking for answers to all the doubts that arose during the work. We have never been left with understatements.

What did you find most impressive or unique about this company?

Involvement in the project. Understanding our needs. Delivering what we expected. Receiving added value from them. Perfect communication.

Are there any areas for improvement or something they could have done differently?

We did not see any areas for improvement.

BACKGROUND

Introduce your business and what you do there.

Schneider Electric is a Fortune 500 company. We are leaders in energy management and industrial automation and we are a €25 billion business. Our global teams are spread all over the world and I myself am located in Warsaw, Poland where we have the core AI development team.

I am the former VP of Analytics and AI in the company. I lead the AI organization as part of the end-to-end global digital customer relationship team, where we basically deal with sales and marketing topics. My hierarchy level is N-2, N being the global executive committee.

OPPORTUNITY / CHALLENGE

What challenge were you trying to address with QuantUp?

We hired QuantUp to help us with custom software development.

SOLUTION

What was the scope of their involvement?

QuantUp helped us with a recommendations engine, which we use for our sales and marketing function. This engine solves the specific business problem of figuring out what next best product a given customer might like.

Though we could resort to Amazon, Netflix, and many other websites that have this function, our business deals in engineered solutions and not products so the logic of product selection is fundamentally different. That's why QuantUp customized the recommender engine’s algorithm and came up with actionable recommendations that could help us grow our business.

Though they leveraged a couple of existing machine learning frameworks, QuantUp built everything from scratch. We provided documents to introduce them to the project and conducted meetings to outline the problem we were trying to solve. We worked together on the application and their team was specifically responsible for data processing and writing the algorithm.

What is the team composition?

We worked with three people from QuantUp: Artur (CTO & Founder) and two data scientists.

How did you come to work with QuantUp?

We had previously worked with the team in 2014. During that time, we did market research to try to find a company that could train us on data science around a specific software. We chose to work with QuantUp and they were excellent in this regard.

We’ve been in touch with them since then, and when the opportunity to invite them for another project came, we double-checked if they would be fitting for it. 

We found that they had expertise in end-to-end development, a strong background in sales, retail, and government-related projects, and experience in making sophisticated machine learning software. They also offered great value for their cost, so we decided to work with them again.

How much have you invested with them?

We have spent between €50,000–€100,000 (approximately $58,000–$116,000 USD).

What is the status of this engagement?

Our first partnership was in 2014, but this specific project was started at the end of 2018 and completed in mid-2019. We still keep in touch with QuantUp for future projects.

RESULTS & FEEDBACK

What evidence can you share that demonstrates the impact of the engagement?

There were two things that we used to measure the success of the project: tangible business value and user satisfaction. The user satisfaction was great, and we would rate QuantUp highly on this front. They also positively impacted our business by helping increase our website views by 15%–20%. 

We also noticed a 3% increase in the use of QuantUp’s solution; our salespeople are beginning to turn to it for actionable insight.

How did QuantUp perform from a project management standpoint?

QuantUp was very straightforward during the data preparation phase. They knew the pitfalls for cases and companies like ours, so they were able to bring invaluable advice. They performed very well in terms of project management as well. We didn't really have any problems with scheduling and meeting deadlines

Our organization is very complex to work with, but their team showed a lot of flexibility with adjusting the deliverable timelines.

What did you find most impressive about them?

I loved their practicality. These guys were like a part of our own company. The questions they asked and the way they worked made us feel like we had our fellow data scientists close to us as opposed to having vendors that just wanted to sell us their services.

Are there any areas they could improve?

We learned that we should’ve aligned certain stakeholders upfront and fixed some foundations for data availability. This is more on our side, however, because QuantUp couldn’t have performed any better. They were excellent.

Do you have any advice for potential customers?

Don't treat QuantUp just as technical executors; involve them from end-to-end or from design to maintenance, because they bring a lot of value along the way.

Also, make sure to integrate them as actual team members and empower them to perform. Ensure that they have a voice around the table because they are excellent in what they do. Some of them will be shy at first, but once you get to talking, you will see just how brilliant they are. 

BACKGROUND

Please describe your company and your position there.

I'm CIO in TOP three debt collection companies in Poland. We service over 4 million clients.

OPPORTUNITY / CHALLENGE

For what projects/services did your company hire QuantUp?

Valuation of debt portfolios prior to their purchase is one of the key areas of a debt collection company's operations. It is this key area that the solution provided by QuantUp addressed.

Valuation models developed by QuantUp allowed not only for more accurate valuation of individual portfolios but also provided more precise information allowing pricing analysts to make more effective decisions in the process of debt purchase.

SOLUTION

How did you select this vendor and what were the deciding factors?

We chose QuantUp because of our successful consulting and training partnerships in the past. The second important factor was their experience in applying mathematical methods to bad debt management.

Describe the project in detail and walk through the stages of the project.

The goal of the project was to develop an alternative to the current method of valuing debt portfolios prior to purchase. Its key element was to predict recovery from a purchased debt. The first stage of the project was a series of workshops to define its scope (e.g. kinds of debt portfolios to be priced), the functionality of the solution, and model validation criteria (data and measures). They delivered the first results and then iteratively improved the models.

Next, software to integrate the algorithms with our pricing system was provided. We received support for the integration. After initial testing, it became apparent that revisions were needed to the assumptions made at the beginning of the project. In the next iteration of modeling, the changes were incorporated. An important element was the validation of models.

A fairly complex validation scheme was proposed to reflect the conditions under which the model would be used. Custom methods based on historical data and multidimensional statistics (with optimization) were applied. They are heavily parameterized to achieve both flexibility and accuracy of recovery forecast.

How many resources from the vendor's team worked with you, and what were their positions?

A team of four people was involved:

(senior data scientist, data scientist, junior data scientist, software developer)

Proof: 

Supervised by an expert data scientist.

RESULTS & FEEDBACK

Can you share any outcomes from the project that demonstrate progress or success?

The project was successful and the solution is used for the pricing of portfolios. The reports designed for pricing experts help them to understand the reasons behind the pricing delivered by the algorithms and facilitate manual corrections (if needed e.g. because of not having enough historical data).

We are in the process of switching our product pricing processes to use this solution. Unfortunately, we can’t share specific business results.

How effective was the workflow between your team and theirs?

Close collaboration and good communication were important to us. The high responsiveness of the team was an advantage.

They also suggested how to present the results and additional analysis in the report for analysts valuing debt portfolios. This proved to be a very big help in the pricing process.

What did you find most impressive or unique about this company?

We were impressed with the development of dedicated algorithms for our data and needs. This is a rare ability.

Are there any areas for improvement or something they could have done differently?

The project was professionally done. We are pleased with the quality of the analytical work and the support for integration with our pricing system. Delivery time could be shorter.

BACKGROUND

Please describe your company and your position there.

In transcash we care for the financial safety of companies in the transport industry. Our services include investigating the credibility of contractors, financing invoices and monitoring payments, amicable debt recovery and judicial recovery. My position in the company is the manager of the risk management team. The team is responsible for the risk management processes in the financing services.

OPPORTUNITY / CHALLENGE

For what projects/services did your company hire QuantUp?

QuantUp developed a solution automating the assessment of credibility of customers and their counterparties, parties of the invoices which are the subject of factoring by Transcash. This enabled the automation of the credit risk assessment process and decision making in factoring - the key area of the company's activity.

The role of QuantUp's team was not only limited to the creation of appropriate Machine Learning models, but was also related to the process analysis and development of a new shape of the credit risk assessment process.

SOLUTION

How did you select this vendor and what were the deciding factors?

The deciding factors were joint and extensive experience of QuantUp in fintech projects (for factoring companies, collection agencies, banks) and projects for transportation companies. A long-term experience in applying many kinds of mathematical models in solving diverse problems has been a great asset.

Describe the project in detail and walk through the stages of the project.

The project started with a workshop in order to analyze business needs, the underwriting process, and the data being collected during the process. The next stage was focus on data preparation and their preliminary analysis. It was concentrated on understanding the data characteristics, relationship between different data sets and assessing (and improving) data quality.

To implement the solutions we decided to apply black box Machine Learning models equipped with an explaining mechanism. Next, the models were gradually improved. The final models were thoroughly validated, including out-of-sample and out-of-time approaches.

At the end of the process the models were additionally calibrated and integrated with our IT system.

How many resources from the vendor's team worked with you, and what were their positions?

The team consisted of 5 people (senior data scientists, data scientists, software developers) supervised by a domain expert.

RESULTS & FEEDBACK

Can you share any outcomes from the project that demonstrate progress or success?

Unfortunately we can’t, but I would like to emphasize that the project's goals have been reached.

How effective was the workflow between your team and theirs?

It was a very close collaboration that resulted in a very efficient workflow. The key factor was the communication within the team - very open and focused on the goal of the project. QuantUp managed to intellectually discuss challenging topics in an approachable way.

What did you find most impressive or unique about this company?

They guided us through the project anticipating all the risks to appear and proposed alternative methods to solve. What was important, they were fully transparent in their activities and took care of us to understand all the steps of the projects and decisions taken in the area of predictive modelling.

It was very important that QuantUp’s team analyzed deeply our underwriting processess, data accessible, and costs of gathering new data or changing the current process. Based on results of this analysis they proposed an optimal approach. Last but not least, they are experts in applications of Machine Learning.

Are there any areas for improvement or something they could have done differently?

I can’t see it right now. The project was really well done.

BACKGROUND

Introduce your business and what you do there.

I’m the managing director at Goldschmidt. We manufacture visual inspection systems and automate the processes of railway businesses.

OPPORTUNITY / CHALLENGE

What challenge were you trying to address with QuantUp?

We needed their team to aid our software development efforts.

SOLUTION

What was the scope of their involvement?

QuantUp initially conducted software analysis to determine the weak points of our platform and proceeded with outlining possible solutions we could implement. We gave them a project brief and indicated our goal of identifying specific damages among railway sleepers, so they performed a stability study and presented it to us. Once QuantUP secured our approval for their initial concept, they then developed the final software package to address our problems, which also included a data processing library.

They developed a software package based on the images we provided at the start of our engagement, which we collated from our measurement system. The package processed each of those images by marking areas with potential defects, including sleeper cracks or chippings. Upon completing that package, they reported the results and submitted the source code to us. Once we were satisfied with the entire package, we implemented QuantUp's solutions into our core software system and delivered the product to our client.

What is the team composition?

We worked with four people from QuantUp in total, including their managing director, a data analyst, and a developer.

How did you come to work with QuantUp?

We found QuantUp on Google while looking for potential agencies that we could hire to tackle our project. We liked their portfolio, so we reached out to them and asked them to collaborate with us. 

What is the status of this engagement?

The project ran from February–August 2020.

RESULTS & FEEDBACK

What evidence can you share that demonstrates the impact of the engagement?

QuantUp built a system that was well-received by our end-user, which in this case was the National Railway. In addition, we published their findings and presented them at a conference hosted by the Machine Vision Week, and the audience was [leased with the accurate results and efficient quality of our software library.

How did QuantUp perform from a project management standpoint?

We had weekly meetings with QuantUp's experts and developers. Their team had reliable project management skills — the flow of the development process was smooth and well-monitored. While we initially encountered challenges in terms of translating the domain language and knowledge, we eventually overcame that with their team's assistance and quick solutions. They crafted on-point yet casual reports, which we highly appreciated because they were more digestible. Additionally, QuantUp was a team of experts who not only knew their craft well but were also willing to impart their knowledge for us to get a better grasp of the system.

What did you find most impressive about them?

QuantUp's expertise in image processing was impressive — they were always ready to integrate this level of knowledge into our package, which allowed us to complete everything in a short span of time.

Are there any areas they could improve?

There’s nothing I’d have them improve. We were satisfied with the results that QuantUp delivered.

Do you have any advice for potential customers?

Allot a good amount of time in scoping your project and estimating timelines. You'll also need to define your goals clearly so QuantUp can come up with more tailored solutions for your needs.

BACKGROUND

Introduce your business and what you do there.

I’m the CEO of Bioavlee SA, a company that develops and sells microbiology lab devices, which are used for bacteria identification and quantification. 

OPPORTUNITY / CHALLENGE

What challenge were you trying to address with QuantUp?

We worked with QuantUp on a difficult project involving algorithms that identified bacteria. Our company came up with this technology, and it involved capturing photos of diffraction patterns from lasers placed on bacteria colonies. To do that, we needed an algorithm that differentiated and identified bacteria species through the spectrum analysis. 

SOLUTION

What was the scope of their involvement?

QuantUp was responsible for developing our statistical algorithm. Initially, we gave them a wide range of photos of the bacterial species’ diffraction spectrum. We also gave them raw data about the species. In essence, our team asked QuantUp to use whatever solution they had to provide the most accurate results.

On top of that, we wanted end users to access our analytical device and set new tests, so QuantUp developed a web interface connecting the device to our servers. Testers could place new bacteria samples on said device, which would then send spectrum photos to our servers. After that, the servers would receive identification results from the algorithm, and then they would send the results back to the device. Moreover, the algorithm was connected to our database, which QuantUp also created. This system should be able to simultaneously connect hundreds of devices to the database. The tests must run smoothly and shouldn’t last more than five minutes per user. 

For tools and technologies, QuantUp used R, a statistical programming language for developing algorithms. Their work went on for a year since it was complex, involving plenty of development and testing phases. A third-party microbiology laboratory in the UK tested the algorithms as part of the process. 

What is the team composition?

QuantUp assigned 3–5 people to this project. Two of them were in charge of communicating with us, while the rest of the team worked directly on the project. 

How did you come to work with QuantUp?

I knew one person from the QuantUp team from previous projects. We also asked two other companies to make an offer, but QuantUp’s proposal was ultimately the best fit for us. On top of that, it was better for us to work with a provider that we already knew could do the job right. As a result, we decided to go with them. 

How much have you invested with them?

We spent around zł350,000 PLN (approximately $92,000 USD).

What is the status of this engagement?

The collaboration lasted from October 2018–November 2019. 

RESULTS & FEEDBACK

What evidence can you share that demonstrates the impact of the engagement? 

QuantUp did a great job since the algorithm had 98% accuracy in identifying six species of bacteria. We also published a paper from the UK-based microbiology laboratory and a feasibility study that took four months to create. Overall, the software that QuantUp developed was instrumental in making bacterial tests possible, and that was proof that they did a good job.

Unfortunately, we had to close the project because we didn’t sell the devices. However, if we started selling them, we’d work with QuantUp again to make sure that the program worked well for large-scale usage. 

How did QuantUp perform from a project management standpoint?

Our collaboration was well-managed. We used project management software that made communication between us easier. The software also allowed our teams to have a clearer view of our tasks and project timeframe. On top of that, we had face-to-face meetings with the QuantUp team twice a month since we were in the same city. We also used phone calls and emails to communicate. 

What did you find most impressive about them?

QuantUp had a vast amount of knowledge about the solutions they used in the project, which was important for us. In a nutshell, they didn’t have to spend much time learning what they should do; they simply knew what needed to be done.

Are there any areas they could improve?

At the time of our engagement, their team wasn’t large enough. They had three senior resources and two junior ones. As a result, those members didn’t work at the same pace, and some things in our project took a bit longer than expected. 

That being said, their team has grown since then, so they’ve probably resolved this already.

Do you have any advice for potential customers?

Set your expectations properly from the very beginning. Sit together and make a plan to ensure nothing escapes your mind. Apart from that, use tools to monitor the project so you can react quickly to things that aren’t considered in the beginning. 

BACKGROUND

Please describe your company and your position there.

I am Head of Software Development at Wattstor, company providing solutions to Optimize energy consumption and generation.

OPPORTUNITY / CHALLENGE

For what projects/services did your company hire QuantUp, and what were your goals?

For project to help us with creating prediction models and analyzing historical data.

SOLUTION

How did you select this vendor and what were the deciding factors?

Based on references and experience in energy sector.

Describe the scope of work in detail, including the project steps, key deliverables, and technologies used.

Analyze historical data, propose solution to create prediction models for energy consumption and generation, build and tests prediction models. During project were considered classical models but also ML models.

How many people from the vendor's team worked with you, and what were their positions?

Total four people, data scientist and AI/ML experts.

RESULTS & FEEDBACK

Can you share any measurable outcomes of the project or general feedback about the deliverables?

Project was delivered in time and in budget. Accuracy of created models is matching our expectations.

Describe their project management style, including communication tools and timeliness.

Close communication with client (us), project was managed as Agile based on time/material basis, focus on speed of delivery and set goals. They were able very quickly understand our business goals.

What did you find most impressive or unique about this company?

They are real experts in ML / AI & optimisation, they delivered the project on time and within budget despite ambitious targets and tight deadlines.

Are there any areas for improvement or something they could have done differently?

Extend project timelines to have more time for additional iterations. But this was our decision.