CRM vs Excel: when to buy, configure or build software

Keep a spreadsheet when a defined register, manual upkeep and its access model are adequate for the work. Evaluate a configured CRM when people need shared customer history and coordinated follow-up. Consider custom software after a test exposes a material gap that configuration or a bounded integration cannot reasonably close. The company must also be willing to own the resulting service.

Choosing CRM vs Excel also means choosing who will keep the setup working. Somebody will maintain the records, handle exceptions and change the rules after launch. A subscription transfers some of that work to a supplier. A custom application transfers more of it to your company.

Founders and product leaders need to compare the ongoing work alongside the first invoice. This article compares spreadsheets, configured CRM software and custom applications. It is not a vendor ranking or a price estimate. CRM means customer relationship management; a custom application is software commissioned or developed around the company's requirements.

DecisionSpreadsheetConfigured CRMCustom application
What does it fit?A register with acceptable manual upkeepCustomer work that the platform can representA demonstrated gap worth owning
Who changes the rules?Workbook owner and any script maintainerAdministrator within platform constraintsProduct and engineering owners
What remains your job?Record quality, sharing and reconciliationConfiguration, adoption and integration boundariesThe application and its operation
What should funding cover?Upkeep and checksSetup, subscription, training and changesDelivery, operation, maintenance and changes
What proves the choice?An actual working registerA configured scenario, including exceptionsA bounded prototype with acceptance evidence

The columns are not mutually exclusive. A CRM can own customer records while a custom module owns a specific approval. A spreadsheet can remain an analysis tool. For each fact, decide which system is authoritative and how another system receives its updates.

When is a spreadsheet a reasonable customer tool?

A spreadsheet is reasonable when the team can understand and maintain its rules without building an undocumented application around it. Records need stable identifiers, an accountable owner and an acceptable way to check changes. Manual effort may be a sensible trade-off while the work is still changing.

Avoid the weak claim that Excel cannot support collaboration. Microsoft documents workbook co-authoring, subject to supported versions and cloud storage requirements such as OneDrive or SharePoint Online. Test the environment your team actually uses.

Sharing a workbook and separating what each role can see are different requirements. Microsoft states that worksheet protection is not a security feature; its purpose is to restrict changes to locked cells. That does not establish the security of every workbook setup. It means a protected cell should not be accepted as proof of a customer-record permission model.

The question changes when a workbook becomes a dependency for several scripts, copies and unofficial procedures. At that point, identify what people must do to keep it working. You might still retain it, but those tasks belong in the comparison.

If the uncertainty is whether customer work needs a shared system at all, the separate CRM readiness guide addresses that earlier decision. Here, assume there is a defined piece of work to support.

Can configuration close the gap before you build?

Require evidence that a candidate CRM can represent the important records and transitions. Ask for a configured path, including a handover and an exception, rather than a tour of screens.

Additional fields and objects do not automatically require a new application. For example, Bitrix24 documents Smart Process Automation with configurable stages, automation and links to CRM records. This is an example of configuration capability, not a recommendation of a specific plan. Availability and limits still need checking in the intended product configuration.

Use a simple gap record during evaluation:

  • State the action a person must complete.
  • Record what the configured product can already do.
  • Describe the part that fails or requires an unacceptable workaround.
  • Test whether an integration or a small extension addresses that part.
  • Name the owner of the workaround or extension after launch.

"Our process is unique" needs a more specific explanation before it becomes a reason to build. A demonstrated inability to represent a required relationship, enforce a specific transition or support an access rule is something a team can investigate.

Configuration has costs too. An intricate collection of automations may require ongoing specialist attention even if there is little conventional source code. Ask how changes will be documented, tested and handed over.

What do actual Git changes reveal about custom ownership?

A custom customer tool includes details that a feature checklist can hide. Test how an incoming form reaches the working queue: where staff see the record, who accepts it and how they know what to do next. Check what happens if a notification fails, and whether a rule change affects work already underway. These are implementation and acceptance questions, not evidence of revenue improvement, lower ownership cost or current reliability.

When does a custom module deserve investment?

A custom module deserves consideration when a demonstrated gap affects something the business values and the company can support the resulting software. The gap should survive a configuration test. It should also be specific enough to bound.

Consider a hypothetical partner approval process. The CRM already holds the partner relationship and sales opportunities. An internal service must evaluate a particular submission and preserve its decision. The company could build the approval service and return its result to the CRM. Replacing the whole customer system would need a separate justification. This example is illustrative; it is not a reported client engagement.

Keeping the custom boundary around the approval may preserve useful capabilities in the existing CRM. It also creates an integration to own. Decide where the approval is recorded, which system can change it and what happens when an update is delayed or repeated.

A complete custom CRM may still be justified. That needs evidence for the wider scope, an accountable product owner and a credible maintenance arrangement. Recreating standard customer features merely because a prototype is quick to generate does not establish that case.

For a bounded build, the MVP scope guide offers a way to separate the essential path from later features. Apply that discipline to internal software as well as products sold externally.

How should you compare total ownership cost?

Use the same planning period and cost categories for every option. A 12-month period is one illustrative planning choice, not a claim about payback. A different horizon may be more appropriate for your commitment.

Cost categoryEvidence to request
Getting startedData preparation, configuration or development, migration and acceptance work
Running the serviceSubscription or hosting, administration, monitoring and support
Doing the workTraining, reconciliation and error correction by staff
Changing the serviceWho implements a rule change and checks existing work
Leaving the setupExport, relationship reconstruction, history transfer and support handover

Add these categories for each candidate. Keep unknown amounts visible. If you include lost sales or saved employee time, identify the source and calculation; otherwise those figures remain assumptions.

Include implementation work alongside the subscription. Add operating costs to the development quote. An open-source starting point may reduce what you need to build, but you still need to establish who owns deployment, updates and your modifications. The size of that commitment depends on the actual project.

Compare uncertainty as well as estimated totals. If the hardest gap has not been tested, a small investigation may be the next purchase. Use its results to estimate the difficult part before committing to an entire replacement. The MVP budget guide covers other decisions that change the scope of an estimate.

How do you test the CRM vs Excel choice?

Test a complete record journey with representative exceptions and appropriate access roles. Use anonymized examples. A proposed test should include a new request, a handover, a repeated submission and a rule change while work is open.

For each step, record what was saved, what the next person could see and what manual work remained. Follow the record through the journey rather than accepting each screen in isolation. Retain failed or unresolved cases in the decision document.

Test portability separately. HubSpot's export documentation distinguishes current record properties and associations from ways to obtain activities such as calls and notes. A record export is therefore not proof that one download contains everything you need for a move.

Create a small export and check that the intended destination can reconstruct the relationships you care about. A backup file and a tested migration are different deliverables. This requirement applies to a custom system too: owning the code does not establish that another team can recover the data or operate it.

Record the decision on one page:

  1. Identify the authoritative system for each important record.
  2. Name the chosen option and the evidence that supports it.
  3. Document the custom boundary, if there is one.
  4. Assign responsibility for administration, incidents and future changes.
  5. Set a review condition tied to an observed limitation.

The review condition could be a recurring reconciliation task that becomes unacceptable. It is a proposed operating rule, not a universal threshold. If the CRM vs Excel decision cannot be explained at this level, clarify the missing evidence before funding a broader build.

Frequently asked questions

How many customers mean I need to replace Excel with a CRM?

There is no customer-count threshold established here. A small process may need stricter access and handovers than a much larger reference list. Assess how records move through the work, which checks remain manual and whether those limits are acceptable. A customer count alone cannot determine the right architecture.

Is building a custom CRM cheaper than buying one?

That is unknown without comparable scope and an ownership horizon. Include setup, operation, change and exit work for both options. A small internal module and a full CRM replacement are also different purchases. Compare the work each one removes and the responsibilities each leaves with your company.

Should we delete spreadsheets after adopting a CRM?

Keep a spreadsheet if it has a defined job, such as analysis of exported records. Agree which system is authoritative for each fact and how changes move between them. Independently editing the same facts in both places creates a reconciliation obligation that belongs in the operating plan.

Does AI make the build-versus-buy question disappear?

AI can assist with a prototype or code, but that output still needs acceptance evidence. Test access, record handling, exceptions, portability and future changes as you would for another implementation. Assign an owner for the service after launch. A generated interface does not answer those operating questions.

Sources