A new South African online platform uses machine learning to help agents and landlords pre-emptively identify well-behaving tenants.
Averly is a web-based service which lets rental applicants submit themselves to an online screening questionnaire that generates a personal score.
This score can then be used by landlords and agencies to assess an applicant’s eligibility for rental.
Averly CEO Zabeth Venter explained one of the challenges in leasing is attempting to predict how a tenant would behave or treat a property.
She said the tool was created as a method for addressing relationship management in the property space.
“I think if you want to describe the rental industry, there are actually tenuous relationships all around,” Venter said.
“You have the classic stories where the tenant made a bonfire on the apartment floor or actually used the carpet to put out their cigarettes and they left more than 200 burn marks,” Venter said.
On the other side, there are also badly behaving landlords and agents.
“We’ve heard stories about landlords and agents not paying utilities, for instance, so the poor tenant had to sit without electricity and water,” Venter said.
One tenant even went as far as to buy a property to fix a problem with a blocked drain.
Pure financial focus is a problem
Venter said the platform is not about giving one party an upper hand over the other, but about enabling tenants to show their value in an application, apart from just their financial standing.
She explained that most rental screening processes are focused primarily on the ability of applicants to pay their rent.
For the most part, the agent would then have to use their “gut feel” to try and make an assessment of a person’s potential behaviour.
Venter said this process was flawed, as good financial standing does not always translate into good behaviour.
“What we’ve heard is that a tenant paid every single penny that was owed in terms of rent, but when he moved out, the property owner actually had to spend a year and a half’s rent on repairs and replacements,” Venter said.
On the flip side, certain candidates who may not be able to afford what the landlord or agency is asking, may make for dependable tenants if they could negotiate for a reduced monthly fee.
The science and tech behind Averly
Averly’s questions extract a prospective tenant’s convictions or opinions about renting in order to model what the person would behave like as a tenant, Venter said.
“Science tells us that if you are very convinced about what you are saying, there is an 80% chance that you will act on that,” she noted.
The screening system uses an algorithm/model which is based on three building blocks:
- Neuroscience from established scientific publications was used to define how a person’s emotions and convictions can be measured.
- Experts determined the 22 questions to extract the right behaviour.
- Machine learning takes the user responses, measures them, and converts the information into a score.
As part of the machine learning process, the algorithm was trained with real objective evidence in the industry.
“We went to agencies and surveyed tenants and benchmarked the model with true observations in the market,” Venter said.
The model is not being left to operate completely on its own from here, however.
It supports a component of human intervention which lets the company measure outcomes to see if bias came into play.
“That’s your biggest risk with machine learning – over time it incorporates bias all over the place and it comes up with answers that make no sense,” Venter said.
“We built a moat around our model that we can continuously moderate it and make sure that we understand the answers that come out of it,” Venter said.
Testing it out
We decided to test the platform to see what types of questions are asked and the features which are supported.
After creating a profile and confirming our email address, we were directed to the screening questionnaire, which started with a few test questions.
Following this, we were presented with a few simple statements to which we could either respond with a “Yes” or “No”.
Examples of these include:
- I value good relationships with my neighbours.
- I often disagree with my landlord or agent.
- I will pay my rent in advance.
Once we finished the questionnaire, we were asked to complete our profile, which included providing personal information such as name, ID number, address, and contact details.
We also had to upload a profile picture and copy of our ID and were provided with the option to add a reference letter.
In addition, we had to fill out employment information, before we were presented with a score of 3.9 out of 5.
The Averly dashboard allowed us to keep track of applications and manage our information.
Pricing and future plans
Averly is free to use for tenants, while landlords and agencies are charged starting at R30 per month per active lease.
This price goes down as the number of leases is increased, Venter explained.
“A portfolio of 160 properties would cost about R4,300 per month,” she said.
This includes the application, inspection, and screening processes, while integration with credit agencies can also be done.
While the initial functionality will be focused on the tenant’s eligibility, Averly plans to introduce scoring of landlords and agents in future.