FOYI

For a used-car sales dealership, growth came with one major problem. The more experienced sales staff were able to sell at 12% better price than their internal pricing matrix. This meant lost opportunity to make higher profits.
The internal pricing matrix used by the sales staff was built by the pricing team at the headquarters and major revisions were made just once a year. While the branches had some discretion for the range of the prices, it was observed that about 8% of the transactions were grossly undervalued.

FOYI was involved to craft an on-demand car valuation process so that the pricing team can revise the prices more often.

Challenge
The used-car sales customer proposed a project scope with the task of building a car price prediction application. The idea was to use this application to generate on-demand pricing based on the recent changes in the market pricing and therefore allow the pricing team to revise the pricing matrix more frequently so that the prices reflect the current market conditions.

To this extent , FOYI proposed a 2 day discovery workshop to understand the current state of the data and estimate what the project timelines, dependencies and the cost. As part of the 2 day discovery workshop, the following challenges were uncovered.

1. Vintage and limited edition cars were valued very high.
Age of the car is key attribute of car valuation i.e older the car, lower the value. However, a 30 year old car is valued more than a 5 year old car on account of a special value for vintage and other reasons. Such data cannot be used to train a machine learning model.

2. The value at auctions had a wider range compared to buy-now.
Broadly speaking, there were two channels of sale at this customer namely, auction and buy now. The customer prefers to put the cars in auction first and when they are not sold for a few months, they put a sticker price and move it to buy-now. Since the price in auction varies drastically, sometimes as much as 80% higher than the ask price, this data cannot be used for training the machine learning model.

3. The value at 2 biggest cities were at least 8% more than the rest.
The stock and the prices were different between the branches across the country. Therefore, the influence of the branches cannot be ignored while training the machine learning model.

Approach
The challenges identified as part of the 2 day discovery workshop helped FOYI propose the scope of the project and the approach to solving the problem as follows.

1.  The value of the vintage and special/limited edition cars was more subjective and therefore such transactions (around 3% of the total transactions) was excluded from this project.

2. The value at the auction was also dependent on the bidding audience and was more subjective given the emotional buying behavior at the auctions. Therefore, auction related transactions(around 27% of total transactions) were excluded from the project.

3. Some branches were more similar to each other than the others based on the valuations of similar cars. Therefore, a branch segmentation model needs to be built and fed into the price prediction model.

Outcome
The 3 key deliverables of this approach and their benefits are as follows.

1. Insights dashboard on the historical sales data from 2 day discovery workshop.
Benefit: An on-demand report that provided visibility on the changes in the market valuation and helped in identifying the right time to revise their pricing matrix 

2. Grouping branches into meaningful segments from the segmentation model.
Benefit: The segmentation of the branches helped the sales teams to get insights into sales targets for the branches.

3. An on-demand price prediction process. 
Benefit: A well documented price prediction process that the internal IT team was trained on enabled the business to revise the pricing matrix faster and more frequently.

A tertiary education institution aimed to increase pass rates for their courses. They observed that while some students needed help from tutors along the way, some others did fine by themselves.
They wanted to make the most of their tutor’s time by helping the students to whom it matters the most. FOYI was involved in solving the problem of identifying the students who have the most likelihood to pass if the tutor would help them during the course work.

Challenge
The tertiary education customer proposed that a preliminary analysis be conducted to understand the nature of the problem and the scope of the project. As a result of this initial analysis, the following challenges were uncovered.

1. The course structure and grading changed over the years.
The nature of scoring and the EFTS have changed for more than 10% of the courses over the years. This meant that historical data cannot be used as-is for training machine learning models.

2. Some key attributes of the students were not permitted to be used on account of the privacy policy.
From the existing reports, it was well known that some student attributes such as pass rate for previous courses are good indicators of current course pass rates. However, such information was not permitted to be used in this project.

3. The board of the institution wanted to have the knowledge of this model retained within their team.
The team had a couple of recent graduate hires for data science and they needed to be taken along the journey of this project i.e. on the job training.

Approach
Given the nature of the challenges, FOYI was able to call out at the very beginning that the project duration would run into a few months and not just a few weeks. In order to ensure that the project was within the budget, FOYI proposed a more advisory approach to this project. This meant that FOYI would provide the detailed set of steps to be taken at each stage of the project and then facilitate 2 day workshops with their data science teams to analyse the outcome and propose next steps.

The documentation of the proposed approach would be shared by FOYI while the heavy lifting in terms of the deployment and the final as-built documentation will be handled by the internal data scientists.

Outcome
The 3 key deliverables of this project and their benefits are as follows.

1.  A heuristic engagement model.
Benefit: A simple and logic based formulation was crafted to ensure there is an explainable score right from the first week of the course. This helped the tutors to engage via email with the students having a low score.

2. A probabilistic pass likelihood model.
Benefit: A predictive model based on historical data of comparable courses and predicting the pass probability each week. This helped the tutors to call the students for following up on assignments on a weekly basis.

3. Academic style documentation with literature reviews and empirical methodology. 
Benefit: This documentation that followed the academic rigor helped the data team showcase their scientific approach to the executive leadership and sustain their buy-in during the course of the project.

The operations team at a nurse staffing firm during the 2020 pandemic had a major challenge. The demand across different wards at multiple hospitals has changed fundamentally.
This fundamental shift in demand affected the downstream work of staffing the nurses appropriately. They had to revise their entire demand forecasting. FOYI was involved for this work.

Challenge
The nurse staffing customer initially proposed a project scope with the task of building a rostering application to enable them to roster the nurses automatically. Over the course of initial discussion, FOYI was able to highlight the key challenges with that project scope.

1. The pre-pandemic historical data is less relevant.
The rostering process is dependent on the demand for a staff per hospital, ward and shift. Since this demand is now different during pandemic, the earlier estimates do not work.

2. The pandemic data is only 6 months old and therefore lacks seasonality.
The operations team are used to observing seasonality in the data. For example, the winter months had a few wards requiring more staff than the other. However, since the pandemic started 6 months ago, there was not enough data to see how these known estimates have changed.

3. Some nurses were staffed between in-patient and out-patient wards.
The number of in-patients were dramatically low for some wards and the number of staff needed for the out patients dramatically increased on account of longer and more complex process of segregating COVID symptoms and the rest. This increased the unpredictability of required staffing for in-patient vs out-patients.

Approach
As a first step, FOYI proposed an exploratory data analysis project be undertaken first. The objective was to identify the insights from the 6 months of pandemic data in comparison to the pre-pandemic data. This project can then help in defining the scope of demand forecasting project.

Once the demand forecasting model is in place, this could then be the key input to the customer’s actual request i.e. staff rostering application.

This proposed approach led the customer to start with two foundational projects before the rostering application project.

Outcome
The 3 key deliverables of this approach and their benefits are as follows.

1. Understanding the parts of the business that changed more than the rest
Benefit: Identifying the outliers meant more efficient and effective use of the project budget.

2. User driven dynamic application to generate the staffing demand per hospital, ward and shift
Benefit: An on-demand training and forecasting application meant that the application lends itself to any further changes to the demand.

3. Detailed documentation to maintain the model and the application.
Benefit: No dependency on external consultants for retraining the model.

A telecommunications business was on a major digital transformation journey. They were moving away from the on-premise legacy infrastructure to cloud.
This transformational journey entailed a set of data migration and data science model migration projects. The data was migrated from on-premise SQL databases to the cloud based Snowflake data sources. At the same time, the on-premise manually refreshed data science models needed to be re-written in Azure ML pipelines.

FOYI was involved in migrating the data science models to Azure ML on the cloud.

Challenge

The telecom customer initially proposed a project scope with the task to migrate 6 on-premise production models to the cloud in 6 months. Furthermore, it was required that these models be enhanced by performing feature engineering and where needed have new features built.

There were 3 key challenges in this objective identified by the business and FOYI.

 The cloud architecture i.e. Azure ML was evolving rapidly.

• This meant that what worked in Week 1 of the project will change significantly by Week 8. FOYI demonstrated that the ability to upload and use a model object as a pickle file at the time of Request for Proposal(RFP) was not available by the time the project scope is being discussed.

The underlying data sources were also getting migrated at the same time.

• This meant that training data for the models was on-premise and prediction data was on the cloud. FOYI brought out the challenge of deploying trained models as the new features were not available in the same format as the old ones in the training data.

Four out of the six models in question were used as weekly inputs to marketing teams and hence cannot be stopped.

• This meant that migrating these models must be done only when the entire end-to-end process of migration was well known.

Approach

As a first step, FOYI proposed that a pilot project be undertaken with the intention of just one project be migrated as a lift and shift project. The proposed objective was to relocate the on-premise model to the cloud without any enhancement but with changes where the new features be replaced with existing ones where they are not available in the exact format. This will help in estimating the effort for the project more accurately.

This proposal led to the customer revising the objective to a pilot project with 2 models that needed monthly refreshes.

Outcome

The 3 key deliverables of this project and their benefits are as follows.

1. Migration of 2 production models to the cloud
Benefit: Automation of the monthly refresh process means more productivity for the team

2. Identify & mitigate the impact of the new data sources on the model features.
Benefit: Feature stores were updated with the substitute features thus reducing the time to find them for other models using them.

3. Detailed documentation of all the challenges of model migration and the solutions/workarounds thereof.
Benefit: No dependency on external consultants for migrating the rest of the models and therefore low cost of migration project.