When should a business move beyond spreadsheets and group chats to a custom tool?
A spreadsheet and a group chat are not, by themselves, reasons to commission software. The decision changes when an important customer or operational outcome depends on a process that neither tool can reliably own: one current record, a named person at each step, a defensible decision history and a way to resolve exceptions. Test a better shared workbook or an existing workflow product first. Build only the critical gap that remains.
Start with the work, not the tool
This question matters to owners and operations leaders managing service requests, appointments, field work or customer follow-up across several people. Spreadsheets remain useful for short-lived analysis and flexible calculations. Group chats remain useful for fast discussion. The issue is whether a message or a row has quietly become the only evidence that a customer request was accepted, completed or approved.
Microsoft's Excel co-authoring guidance documents simultaneous editing and version history. It also explains that, when people change the same item, the last saved change generally wins. Co-authoring solves many file-copy problems; it does not specify who may approve a refund or what “completed” means. Microsoft's Lists approval example illustrates a separate record, approver and workflow. These are examples of product capabilities, not a recommendation to buy a particular vendor's stack.
Four signs that the process needs more structure
- No dependable master record. The same request has different identifiers or status in sales, operations and finance. Try a single identifier, owner and field definition before integrating systems.
- A chat message stands in for a decision. “Done” might mean received, assigned, executed or accepted by the customer. Define the allowed states, responsible role and evidence for each transition.
- Exceptions disappear during handoffs. A cancellation, reschedule or complaint has no durable owner or approval trail. This becomes especially costly when money or customer commitments are involved.
- Repeated work can be measured. People re-enter the same details or reconcile competing versions. Count instances, minutes, corrections and consequences before estimating the value of a new tool.
These signs prompt an investigation; none is a universal numerical threshold. China's Ministry of Industry and Information Technology guidance for small and medium-sized businesses recommends assessing current operations and potential value before setting priorities, and considering subscription and lightweight products to manage cost. Our recommendation to test the smallest viable change follows from those principles; the appropriate product still depends on the company's actual work.
Choose the smallest route that closes the gap
Once a formal record is needed, compare the coverage and long-term cost of existing products with the uncovered workflow. Our custom software versus SaaS guide provides a broader procurement test.
| Route | When it fits | What to test | Ongoing responsibility |
|---|---|---|---|
| Improve the shared workbook | Few people, simple rules, inconsistent copies or fields | One location, identifiers, access, version recovery and a named owner | People still advance and check each case |
| Configure an existing workflow product | Stable forms, assignments, reminders and approvals | Real cases, mobile access, permissions, exports, integrations and full subscription cost | Vendor administration and process change |
| Build a focused custom workflow | A material rule is missing, or work must connect to an existing app, mini program, device or business tool | Available interfaces, data ownership, exception paths, acceptance criteria and maintenance budget | Development, testing, security and continuous support |
For example, a field service company might already have a booking channel while dispatch, arrival evidence and customer sign-off live in separate chats. A sensible first release could connect the existing booking entry point to one service-job record with ownership, status and evidence. Replacing finance and every customer system at the same time would add scope without proving that the handoff problem has been solved. This is a planning example, not a client case or a measured result.
Observe a real process before pricing software
Follow one request from intake to completion. Sample ordinary work as well as changes, cancellations and complaints. For every step, record who enters information, who checks it, where a decision is made and where the result is stored. Use one request ID to count duplicate entry, waiting time, corrections, missed handoffs and outcomes that cannot be verified. Two weeks is a suggested starting window; low-volume work needs a longer sample.
Make cost assumptions explicit. Illustrative calculation: 20 requests a day, three minutes of repeated entry per request and 20 working days produce 20 hours of repeated entry per month. At an assumed internal labour cost of CNY 80 per hour, that is CNY 1,600 of time. It is not a savings forecast. A pilot must establish which time is actually released after review, exceptions, licence fees and maintenance are counted.
Define a first release and its acceptance tests
Pick one business object, a small set of states and the roles that may change them. For a service job, intake, dispatch, field execution and acceptance should each have an owner. Delays, changes and failed visits need explicit exception paths. A first release may need case creation, assignment, status changes, attachments, reminders, an audit trail and only the integration needed to complete that journey. Chat can continue to carry conversation and alerts; the case record holds the formal state.
Test with actual sample cases: duplicate requests do not create duplicate jobs; every handoff has a timestamp and owner; unauthorised staff cannot approve a high-impact change; outstanding work survives an employee departure; cancellation and reopening preserve history; exported records reconcile to the original workbook. Then compare handling time, repeated entry and exceptions with the baseline.
Avoid copying every old column and notification into the new product. Remove unused fields, agree on who owns final data and pilot a complete path. If a configured product handles that path well, stop there or add only a small integration. If a custom pilot does not improve a verifiable business outcome, narrow it or stop expanding it.
Sources
- Microsoft, Excel co-authoring and version history, accessed 19 September 2026; supports the collaboration and conflict-handling description.
- Microsoft, triggering approvals from Lists records, accessed 19 September 2026; illustrates separating a record from an approval workflow.
- Ministry of Industry and Information Technology, interpretation of the SME digital transformation guide, published 9 November 2022 and accessed 19 September 2026; supports assessment, prioritisation and lightweight options.