Roster Monster

Problem

Overworked healthcare professionals, including doctors, spend enormous amounts of time monthly to plan complex rosters, taking precious time away from their core duties. This one-month long process is tedious and time consuming for three reasons:

Roster planning is mostly done manually using Excel, taking at least 5 hours monthly to finally generate the roster.

In addition to creating the roster, roster planners gather leave and shift requests from colleagues through various means such as emails, texts, or verbal communication. These requests are manually compiled as constraints in Excel during roster generation. Any last-minute amendments or requests necessitate a manual regeneration of the roster, which can extend the time required beyond the initial 5 hours.

Roster planners typically spend a minimum of 2 days adjusting system-generated rosters due to existing system constraints, despite the availability of automatic solutions.

We interviewed users from public sector entities, as well as several public health institutions, regarding rostering solutions. These interviews revealed a desire for rostering solutions to have the abilities to be reconfigured easily, change existing constraints and their weightage, and to accommodate colleagues' requests. Some of the interviews also revealed pain points with existing internal processes and the roster solutions users had previously worked with.

Roster planning demands mental agility as planners balance numerous constraints to ensure fair shift allocation.

They must accommodate individual preferences while meeting mandatory shift requirements, such as avoiding intense duties or weekend work. This entails not only creating new rosters but also monitoring past shift allocations to maintain fairness. Doing roster manually aggravates the complexity of rostering in general.

Opportunity

Rostering is an essential task within the public sector.

Currently, we are starting off with the healthcare sector, targeting doctors first. Doctors alone have wasted over 1,000* hours at least across all public acute hospitals in Singapore. This excluded time spent by other healthcare professionals such as nurses, medical social workers, pharmacists and other allied healthcare partners. If we sum it all up, the healthcare sector could have wasted over 4000 hours on manual rostering instead of their core work – patient care.

* This calculation assumes an average of 53 departments based on 10 public acute hospitals as of 2022 (excluding nursing and allied health services) and 2 doctor rosters per department because there is usually more than 1 type of roster to be planned monthly.

Apart from the healthcare sector, there is also possibility to explore the extent of the rostering issue in the following sectors:

Sectors	Possibilities
Healthcare	Hospitals, clinics, and emergency medical services require rostering to ensure round-the-clock coverage of medical staff.
Emergency Services	Police, fire departments, and paramedics need rostering to maintain 24/7 emergency response capabilities.
Transportation	Public transportation systems, including buses, trains, and subways, require rostering to schedule drivers and operators for shifts.
Education	Schools, colleges, and universities use rostering to schedule teachers, professors, and other staff members.
Military	Armed forces use rostering to schedule deployments, training exercises, and operational duties.
Public Utilities	Essential services like water treatment plants, power stations, and waste management facilities rely on rostering to ensure continuous operation.

Solution

Roster Monster (RoMo) is a self-serve platform that allows roster planners to generate their rosters automatically whenever needed, and set constraints flexibly to meet their needs. We have validated the concept with users through interviews with 10 doctors across 4 hospitals, and overall received positive feedback.

In our very first release of the Minimum Viable Product (MVP), we want to address the key pain points highlighted by public acute hospital doctors on rostering. This includes having to:

(For those using Excel) Manually re-generating rosters to accommodate new requests or rules
(For those who are using commercial solutions apart from Excel) Inability to change existing rules
Single view of all the blocked out dates and shift requests from all staff

Automatic roster generation

Re-generate the roster in seconds whenever there are changes to the original rules
Powered by a constraint satisfaction problem (CSP) solver in the backend

Rule management

Bridge the gap between human-readable roster requirements and technical CSP constraints by splitting rules into 4 categories
- Quota rules: Minimum/Maximum shifts staff must do
- Pattern rules: Shifts that must be done in specific orders
- Rolling window rules: Minimum/Maximum shifts over a fixed length of time
- Soft rules: Whether to minimise/maximise number of shifts

Within each category, configuration is done by constructing a sentence that describes what the rule is enforcing

Staff management

See all staff’s blocked out dates for scheduling and preferred shift slots all in one view, during the entire roster period
Group staff members into any number of arbitrary groups that can then be used during rule creation to make the rule target only specific staff groups

To access the prototype, click here.

Traction

As of 20 Feb 2024, we have received 70 sign-ups from interested partners from H4PG and two email blasts.

81% of the sign-ups were in the healthcare sector from both public and private acute hospitals e.g. Singapore General Hospital (SGH), Tan Tock Seng Hospital (TTSH), and Nobel Medical Group while the remaining 19% were from other sectors such as CPF Board, ICA Singapore and People Association (PA) through organic sharing.

In particular, TTSH nursing and Sport Singapore have both reached out to us beyond H4PG and expressed interest to discuss potential pilots within their departments. An existing contact in the SGH Emergency Department (ED) is also willing to run a pilot with us with minimal feature improvements.

The team has just sent out the MVP to all interested parties and is in the process of scheduling user testing sessions with them.

Team

During H4PG, our team comprised of four members:

[Sufyan] 1 Designer: Sufyan was the team’s anchor for user research and design. The problems that Roster Monster are looking to solve are deeply rooted in current work process and user behaviour. Sufyan drove user interviews by diving deep and picking apart how rostering works. He then translated that into an intuitive product for average roster planners. Good design is particularly important in building an intuitive rule-creation experience and bridging the gap between user and constraint solver.
[Latasha & EnYi] 2 Engineers: Latasha was the engineer in charge of the webapp. Her work included setting up the project infrastructure, designing the database schema, as well as writing the core frontend components. EnYi built out the constraint solver, translating commonly requested rules into coded constraints. He designed the API contract between the webapp and the solver by identifying opportunities to group constraints into categories. Both engineers worked closely with Sufyan to design the rule-generation user experience.
[Limin - SNG] 1 ProdManagement/ProdOps: During H4PG, Limin took on both product management and product operations roles. She was the team’s main point of contact with external parties, and helped to secure user interviews with doctors through her outreach efforts. She conducted problem sizing exercises and identified opportunities for RoMo to tackle, working closely with Sufyan to prioritise features for the prototype. She also developed the main RoMo pitch for demo day, which the rest of the team helped to execute.