Guaranteeing the quality of data, a key issue at Homiwoo

To provide reliable, accurate estimates from a massive amount of property data, Homiwoo has to check the quality of the information. This is one of the tasks carried out by Muneeb ul Hassan, a data scientist with the company since 2019.

Homiwoo's core business is to help property professionals with their decision-making. Success in that aim relies on the quality of the data being provided to customers, so they can make the right choices – particularly when it comes to pricing a property. This is a crucial issue for professionals: a company that advertises a property at a below-market price would lose money, while over-pricing would increase the time needed to find a buyer. 

At Homiwoo, Muneeb ul Hassan has a key role to play in ensuring that the information is reliable. "We have huge quantities of data, so we need a department that makes sure the data we provide to our customers is of high quality, so that their trust in our data is guaranteed", explains this data scientist who joined the company in late 2019.

His background is decidedly international. After a bachelor's degree in Pakistan in computer systems engineering, he spent a year in Melbourne, Australia, followed by an internship in Germany, and then a Master of Research degree in artificial intelligence and data science in Grenoble.

Correcting human errors

The checks carried out by Muneeb enable Homiwoo to give its clients the estimated value of a property and the state of the market at any given moment. "We also tell them how confident they can be in this data on a scale of 1 to 5, with 5 being the best score. Sometimes we can say that the data is 100% reliable", says the data scientist, who knows perfectly well that it’s better for data to be unavailable rather than unreliable!

Among the data sources he uses to estimate market trends is the French government’s official database of sale prices, the DVF (Demandes de valeurs foncières). Published by the public finance department (Direction Générale des Finances Publiques), it provides information on property transactions for the previous five years in France. 

From time to time, however, human error can undermine the data being collected. The surface area of sold properties might have been incorrectly entered manually, for example, while transactions that are conducted at well below market prices can also alter the data. Such inaccuracies are then reflected in the listed sale prices and can distort the subsequent analysis. 

Data quality, a business issue

Muneeb's work involves detecting possible anomalies and correcting them with algorithms to prevent poor quality data finding its way into Homiwoo's results. "It's a reputational issue for us, and therefore a business issue", says this data scientist, who has developed an in-house tool, coded in Python, that automatically checks the reliability of the data. Comments from users are also essential. Muneeb believes their feedback is constructive and helps to foster a process of continuous improvement.

Apart from discussions with end-users, what really enables him to flourish in his work is talking to colleagues. "Teamwork helps me to progress", says Muneeb, who chose to join Homiwoo after his Master of Research degree because he wanted to "be part of the adventure of a start-up that’s growing”. An enthusiast of Machine Learning who wants to bring constant innovation to Homiwoo’s activities, his work requires more than just technical skills. "You really have to pay attention to details”, he says. Rigour is required to guarantee the quality of data, a task with undeniable importance for the business.