You can use your real world business data to power your process model and create a digital twin of your business process.

For more on using data in HASH, see Datasetsâ€‹

HASH supports CSV and JSON datasets. Upload your dataset by clicking the 'new dataset' option in the editor (*in the top left of the screen, over the file picker)*)* o*r by finding/reusing an existing dataset from the HASH index and adding it to the simulation.

The dataset is stored in the 'data/' folder, and the data will be available in the simulation through context.data()

JavaScript

Python

JavaScript

const behavior = (state, context) => {let cs = context.data()["CustomerServiceEmployeesSheet1.csv"]};â€‹

Python

def behavior(state, context):cs = context.data()["CustomerServiceEmployeesSheet1.csv"]

With the data loaded in the simulation we can use it to power the process model. There are many different ways to use data + process models, but the most common are:

Set the properties of the process model. For example, if a pizza delivery service model has a certain number of drivers available to deliver pizzas, they could use a dataset to set that number.

Create distributions for modeling uncertainty. Now, if the pizza delivery service wants to estimate the number of orders they'll get in a day, they could take a dataset that captures last weeks orders of pizzas, choose a distribution that accurately models the data, and then use that to simulate new orders.

Learn more about modeling uncertainty in Designing with Distributions.â€‹

A common business process simulation is the ticket support queue. Tickets from users come in and arrive in a queue. An available customer service agent takes a ticket, solves it, and then picks up the next ticket.

This process would have three elements:

At the beginning, a

**source**block that generates ticketsAt the end, a

**sink**block that collects finished ticketsIn the middle, a

**service**block where customer service workers take tickets and work on them. When a CS employee is working a ticket, they can't take a different ticket until they finish.

For more on the different blocks and when to use a given block, see Process Model Conceptsâ€‹

Using the Process Model Builder, we can set properties for each of the blocks and update my simulation with it. The resulting code will look like:

const agent = {"agent_name": "process_model","behaviors": ["@hash/age/age.rs","@hash/process/source.js","@hash/process/service.js","@hash/process/sink.js"],"process_labels": ["","start","work_on_tickets","solved_tickets"],"process_parameters": {"start": {"template": {},"frequency": 1,"rate": null,"next_block": "work_on_tickets"},"work_on_tickets": {"time": 1,"resource": "cs_employees","next_block": "solved_tickets"},"solved_tickets": {}},"position": [0,0,0]}

Now we can take a dataset defining different customer service employees, download it as a CSV, load it into the simulation, and use the data to set the number of CS employees we have on hand.

create_process.js"start": {"template": {},"frequency": 1,"rate": null,"next_block": "work_on_tickets"},"work_on_tickets": {"time": 1,"resource": "cs_employees","next_block": "solved_tickets"},"solved_tickets": {}},//added a line for the cs_employees resource"cs_employees": context.data()["ExampleCustomerServiceEmployeesSheet1.csv"].length}

This is a good start, but to make the model more realistic we can add a dataset representing ticket volumes over the past five days. I'll take the mean and standard deviation of the data, and set a triangular distribution for the number of tickets to generate every time step.

create_process.jsconst { triangular } = hstd.stats;let tickets = context.data()["Tickets.csv"];//Get min and max number of tickets received per daylet min = tickets.reduce((min, t) => t.count < min ? t.count : min, data[0].count);let max = tickets.reduce((max, t) => t.count > max ? t.count : max, data[0].count);//Get the mode of the array and average with min and max to find the peaklet peak = (min + max + mode(tickets)) / 3;//in the process parameters ..."process_parameters": {"start": {"template": {},"frequency": 1,"rate": triangular.sample(min, max, peak),"next_block": "work_on_tickets"},"work_on_tickets": {"time": 1,"resource": "cs_employees","next_block": "solved_tickets"},"solved_tickets": {}}

Now when we we run the simulation, we have a good estimate of what my customer service process will look like using real world data.