Software Stack Audit: Find the Real Cost of SaaS Sprawl

A company's software budget looks tidy on a spreadsheet: tool, plan, seats, renewal date. Its operating cost is harder to see. People export spreadsheets, chase approvals across inboxes, update the same customer record twice and reconcile numbers that ought to agree.

A useful software stack audit counts both kinds of cost. You do not need an enterprise procurement programme to begin. Give one operational owner 30 minutes, a list of paid tools and a realistic estimate of the work between them.

What is a software stack audit?

A software stack audit is an inventory of the applications an organisation uses, what each application actually accomplishes, how information moves between them and what they cost to operate. The goal is not simply to minimise the number of subscriptions. It is to remove unnecessary work while retaining the tools that genuinely earn their place.

The five costs worth counting

1. Licences: annual plan fees, seats, upgrades and overlapping subscriptions.
2. Administration: time provisioning accounts, managing access, handling renewals and explaining workflows.
3. Integration maintenance: building, fixing and monitoring imports, webhooks, automations and manual exports.
4. Reconciliation and rework: resolving duplicated or inconsistent records, failed handoffs and out-of-date reports.
5. Switching and delay: time lost finding the right record or waiting for information; quantify it only where you can defend an estimate.

Do not double-count hours. If time maintaining an integration already appears under administration, assign it to one category only. Separate measurable cash spend from estimated labour and from risks that you cannot yet value.

A 30-minute software stack audit worksheet

Minutes 0–5: list the tools

Export recent card or invoice transactions and add free products employees depend on. For each tool, record its owner, business purpose, monthly or annual cost, number of active users and renewal date. Mark unknown information as unknown rather than guessing.

Application,Owner,Business_purpose,Annual_licence_cost,Cost_currency,Active_users,Renewal_date,Workflow_supported,Decision,Notes

Minutes 5–12: map critical handoffs

Choose a real workflow that matters to revenue—for example, new enquiry → qualified contact → booked demo → proposal → invoice → payment. Write down which system owns the record at every stage. Draw an arrow whenever information crosses an application boundary. Note whether that handoff is automatic, manually copied or not reliably connected.

Workflow,From_system,To_system,Handoff_method,Times_per_week,Minutes_per_handoff,Rework_minutes_per_week,Owner,Failure_or_delay

Minutes 12–20: estimate the recurring work

Ask the people doing the work: how many minutes per week are spent on access requests, data copying, report reconciliation and fixing broken handoffs? Capture a low and high estimate when answers vary. Measure a typical week rather than the worst incident anyone remembers.

Minutes 20–25: calculate a working baseline

Illustrative example, not a benchmark: A team pays $1,200 a month across its tools, spends six avoidable hours a week moving and reconciling data at a loaded cost of $30 an hour, and incurs $2,000 a year in additional integration upkeep. Its modelled annual operating cost is $25,760: $14,400 in licences, $9,360 in time and $2,000 in upkeep. The numbers only become useful when the assumptions are checked with the team.

Minutes 25–30: select one change to test

Rank your biggest recurring handoff or duplication by cost, risk and how readily it could be improved. Choose a bounded intervention with a named owner—for example, eliminate one weekly CSV reconciliation or consolidate two overlapping subscriptions. Record the baseline before the change and measure it again after four weeks.

Keep, connect, consolidate or retire?

• Keep a specialist product when its unique value exceeds both its subscription and operating burden.
• Connect systems when a reliable, governed integration removes substantial work without multiplying security or support overhead.
• Consolidate when capabilities overlap and shared identity, permissions or records make the workflow materially simpler.
• Retire tools with no clear owner, measurable use or defensible business outcome—after protecting important data and access.

A lower tool count is not automatically a better operating model. Swapping five applications for one poorly suited product can increase costs and risk. The right comparison is the end-to-end workflow before and after, including data ownership, permissions and failure handling.

Seven questions to ask before buying another SaaS product

1. Which current business outcome will this tool improve, and how will we measure the difference?
2. Does an existing application already provide the required capability?
3. Who will own the tool, its data and its budget?
4. Will it create another copy of customer, financial or project information?
5. How do users and authorised agents get access—and how is that access removed?
6. What happens when its integrations fail or a critical record changes?
7. What would it take to leave this tool without losing business history?

If you want a fuller explanation of why these hidden costs compound, read The integration tax: the work you buy software to remove.

Verify the improvement after four weeks

Record the same workflow and transaction volume before and after your change. Sample at least four representative weeks where practical. Track exceptions, work shifted to another team, and any new maintenance burden. Treat hours released as modelled capacity, not cash saved: a cash saving requires an actual reduction in expenditure. If the volume or workload changed, normalise the comparison before projecting a year.

Workflow,Owner,Baseline_min_per_week,After_min_per_week,Loaded_hourly_cost_USD,Weeks_observed,One_off_change_cost_USD,Annual_hours_change,Modelled_annual_labour_value_USD,Modelled_year_one_value_USD,Evidence
Illustrative weekly reconciliation,Operations owner,180,60,30,4,1000,=(C2-D2)*52/60,=H2*E2,=I2-G2,Illustrative only - replace with observed weekly samples

In the illustrative row, the change from three hours to one hour a week is 104 annualised hours, or $3,120 in modelled labour capacity at $30/hour. After $1,000 one-off implementation cost the simple first-year model is $2,120, not verified cash savings or guaranteed ROI. Compare with actual payroll, licence invoices and exception logs before claiming a financial result. Keep the baseline, source evidence, and decision owner with the record.

What to do after your first audit

Your first audit is a baseline, not a verdict. Retain the worksheet, choose a real workflow, name the next experiment and check the outcome after a month. Where people still reconstruct context across disconnected tools, investigate whether applications built around shared identity and discoverable actions could reduce that work.