SFR3 Fund is a tech-enabled real estate fund that acquires, renovates, and rents affordable homes. They specialize in renovating distressed homes, using software-driven operations to grow their market presence in a large number of smaller markets concurrently. Their software automates work orders, payments, and other back office drudgery which speeds their deal velocity and drives growth.
They have grown to over 10,000 homes in over two dozen metros, and are building/repairing more every month.
SFR3 works with vast amounts of accounting data from various property management companies (PMCs). Each PMC records and manages their data using a different system, making it challenging for SFR3 to reconcile. Currently, the data transfer and reconciliation is done manually from each property management company’s accounts to SFR3’s systems, with each transaction taking an experienced user anywhere between 30 and 40 seconds to record. As a result, the data management process is time-consuming and error-prone, leading to inefficiencies and potential losses.
Some of associated challenges are:
Eliminating time- consuming manual work: Multiple data sources from different PMCs and a lack of a uniform data recording process for each PMC caused manual reconciliations while being transferred to SFR3’s QuickBooks
Avoiding data loss and fidelity: Varied account ledger management techniques – including split amounts – by each PMC and different interpretations of transaction information by each user while entering data into QuickBooks resulted in low data quality impacting the ability to automate processes
SFR3's objective was to streamline accounting processes and management, leverage data science and AI/ML to reduce cost and human errors, and future proof the overall business process, and particularly the FP&A function with a modern data stack. The current project serves as foundational work to enable SFR3 to track all activities related to each property, serving as the basis for future advanced analytics and AI.
RapidCanvas collaborated closely with SFR3 to develop the early version of an accounting bot to solve many of the challenges associated with manual data migration, reconciliation and processing.
STEP 1: Exploratory Data Analysis
STEP 2: Data mapping/matching
STEP 3: Automation
STEP 1: Exploratory Data Analysis (EDA)
The RapidCanvas team studied the SFR3 data, which is received from individual property management companies, to better understand the different datasets of each company. The datasets contained details of properties, account information, rental amounts, and dates of payments, among other fields. This was compared with how this information had been translated by humans into the historical accounting system records.
While one company recorded rent as ‘rental income’, another recorded it as ‘rent income’. Utility expenses like water were also recorded differently by different companies. These are examples of the differences in data capture and the critical need for reconciliation.
This step was crucial in understanding the ways in which different property management companies record their data in their ledgers and where the data mismatch happens during the manual matching process.
STEP 2: Data mapping/matching
Once the EDA was complete, the next step was to clean the data and filter it to extract the fields required for SFR3’s requirements. In the example of the rent expense, the data was uniformly mapped as ‘rent income’, to ensure consistency.
The next crucial function carried out was to map the input data from the PMCs to the way in which it is entered in SFR3’s QuickBook ledger. This mapping was carried out through a series of transformations carried out in the RapidCanvas platform. The matching process used fuzzy logic and proprietary natural language processing (NLP) related algorithms for text classification, with both unsupervised and supervised machine learning techniques.
This created a shared data convention between SFR3 and each respective PMC without having to modify the PMC-generated data, to make the flow of data into SFR3’s Quickbooks from the PMC between the two more seamless and less error-prone.
The differences in dataset naming conventions and organization are reconciled through this process. This ensures that the data from each PMC property management company is properly and uniformly captured in QuickBooks, SFR3’s accounting system. The confidence in the data matching is also measured for each step, ensuring high accuracy for the data automation in the next step.
STEP 3: Automation
Once the final matching framework was available for each property management company, data from each continues to flow into the platform and the matching process is automatically performed. This gives SFR3 a clean and comprehensive view of their accounts with each of these companies on QuickBooks.
All of these efficiencies translate to savings of upwards of tens of thousands of dollars that is spent on manual processes. The company’s time to value using the platform was realized within a matter of weeks.
With a clear and automated data reconciliation pipeline, compliance improved and investment opportunities were identified faster. Now, expert time can be reapplied to higher value processes, improving business outcomes, and producing monetary value for SFR3 Fund as a whole.