“Can't we just automate the order desk?”
Before Sandline moved to Beezr, Layla asked that question. Sara had an order open in email. Finance had typed it into accounting. The store had another copy. Everyone had done their job. Nobody knew which copy to trust.
Our view: start business process mapping before choosing software. We'll use four stages: sort, inventory, channels, then processes. You'll finish with a worksheet and one process worth building.
1. Sort your data before drawing the process
A process map shows the steps, people and choices in work. Think of a recipe with names beside each step. Who checks the ingredients? Who says it's ready?
First, sort the information those people use. The three-way sort is our own planning method.
Master data describes the things you use again. Customers, products and suppliers belong here. For Sandline, that's Northbay Hotels, product NG-200 and supplier Lumen Gloves.
Transactions record what happened. An order, purchase order or invoice belongs here. Northbay's order for 40 boxes of NG-200 is a transaction. The customer and product it refers to are master data. Master data changes slowly and is referenced; transactions capture what the business did.
Tasks describe what someone must do. “Check the delivery address” is a task. So is “approve this purchase order.” Give each task an owner and a clear finish.
Aim for one trusted master record per customer or product. Other systems may need copies. Decide which record leads, and how those copies get updates. “One true copy” means clear authority, not a ban on backups.
2. Take inventory: where does each kind live?
Sit with each team and open the files they use. Ask to follow a recent order. Don't settle for a list of software names.
Look for three findings in Sandline's old setup:
- Silos: the store, accounting and Excel price list don't exchange data. IBM defines data silos as "isolated collections of data that prevent data sharing."
- Duplicates: Sara keys the same order into another system. Mark both copies and the shared order reference.
- Leaks: Sara's supplier sheet lives only on Sara's laptop. Here, “leak” means data outside the shared process. It doesn't mean a breach has occurred.
Use this worksheet. Fill the last columns with findings, not guesses.
| Data kind | Where it lives | Owner | Duplicate? | Leak? |
|---|---|---|---|---|
| Master: products and prices | Store and Excel price list | Omar | Yes; compare product codes | Check who can open Excel |
| Master: suppliers | Sara's laptop sheet | Sara | Compare with accounting | Yes; only local copy |
| Transaction: sales order | Store and accounting | Sara | Yes; keyed again | Not found in this check |
| Task: check an order | Email thread | Sara | Check repeated requests | Check shared access |
| Your next record | File or system name | Named owner | Where else? | Who can't reach it? |
An owner answers questions about the record and agrees changes. “Sales” is a department. Sara is someone you can ask.
3. Map the channels: how does data arrive?
A channel is a way a request reaches you. Sandline takes store orders, emailed PDFs, WhatsApp messages and phone calls. Each needs a place on the map.
Ask each team: what do you capture, and where? Who types a phone order? Where does the original PDF stay? Who notices a changed quantity in chat?
If data lives in two places, ask why. Perhaps Sara can't open the main system. Perhaps its form lacks a field. Perhaps nobody named the trusted record.
Treat each answer as a design finding. Blame won't tell you which field is missing.
Write the entry rule beside each channel. For a phone order, Sara records the items and confirms them. For a PDF, keep the source and check unclear product names. “It arrives somehow” isn't a rule you can build.
4. Map the processes as services
Use “service” to mean a request you agree to fulfil. External services include customer orders, returns and questions. Internal services include PO approval and receiving goods. This split helps you find work that has no customer-facing screen.
Each service creates data through a channel. A return request might start by phone. A goods receipt might start with a warehouse entry. Name both the request and its result.
Now draw the steps as they happen today. Then walk through it with the people who do the work. They'll spot the step that only exists on paper.
Sandline's order desk before Beezr
- Sara Receive — a customer emails an order PDF.
- Sara Check — match products against the price list.
- Sara Clarify — ask the customer about any unclear item.
- Sara Enter — type the checked order into accounting.
- Yousef Receive the handoff — get the order details by email.
Beside each step, note the record, owner and next handoff. Mark human checks and routine moves. Add the failure route too. If a product can't be matched, who gets the question? What proves the work is done?
Where should you start with automation?
Choose one bounded process, such as PDF to checked sales order. Keep the wider inventory, but don't build every service at once. Our one-process approach starts there.
| First move | Useful when | The catch |
|---|---|---|
| Fix the shared record | Teams disagree about products | Doesn't move work by itself |
| Connect existing tools | Records and rules are clear | Copies still need owners |
| Build one shared process | Work needs records and approvals | You must agree the rules first |
Choose the next move
- If records disagree → settle ownership first — decide which version leads.
- If the handoff is clear → connect it — test a failed case too.
- If approval rules are unclear → agree them first — then build the steps.
Myth / Reality
Myth: mapping means drawing the ideal process. Start with today's work, including side sheets. Draw the proposed changes separately so people can check both.
Sandline · example
The same desk, built in Beezr
Sandline now runs on Beezr. Customers and Products are modules. Sales Orders hold related records, so Sara can link an order to its customer.
For a PDF order, a Load data step supplies the product list. An AI step reads the PDF and returns matched lines. A Smart Script creates the order lines. Add an approval task before work proceeds. Sara checks the match; the AI answer isn't the approval.
Our take
Our take. The map is the first useful test. If people disagree about a box, resolve that before making it run faster.
Key takeaways
- If you're technical: record the source, shared ID and failure route at each handoff.
- If you're not: ask someone to trace a real order through the worksheet.
- Pick one process with an owner, clear rules and a finish.
Turn the map into a system in Beezr
Start with the records and steps your map names.
- Create master modules — Customers, Products and Suppliers, with relations where needed.
- Add transactions — link order records to the customer and product records.
- Build the workflow — add tasks and approvals for the named owners.
- Connect the channels — register a Beezr webhook URL in the store. Use an AI step for PDFs, with approval after it. Import checked spreadsheet records through the import preview.
- Try one process — walk through a normal request and a missing-product case before expanding.
Your business process mapping questions
Questions readers ask
How do I automate business processes?
Map the current work, agree the rules, then build one process. Test the exceptions as well as the normal route.
What is master data vs transactional data?
Master data describes shared things, like products. Transactions record events, like orders for those products. Tasks say what someone needs to do next.
Do data silos mean we must replace every system?
Use the map to decide. A useful system may only need a clear connection. See the types of system integration for the options and catches.
Where should I start with automation?
Start where the owner and outcome are clear. If the records still clash, work through what a data model is first.
Sandline Supplies is a fictional example company.
Sources
- What are data silos? · IBM · accessed 2026-10-17