How to run an agency tool stack audit

Running our 15-person studio, I could not have told you what we spent on software in a given month. Not approximately. Exactly. The bill was spread across the company card, a partner's personal card with an "I'll expense it" promise, and a couple of freelancer stacks we reimbursed against. Nobody owned the total, so nobody could see it.

A tool stack audit fixes that. It is not a cost-cutting exercise dressed up as good housekeeping. It is the one annual habit that tells you where your stack has gaps, where it has overlap, and where it has dead weight. Done properly it takes a few hours. Done once a year, it compounds, because the bill you can see is the bill you can change.

Why agencies never know their true tool spend

The spend is invisible for three structural reasons, and none of them are about being disorganised. First, the bill is fragmented across payment methods. A typical studio has subscriptions on the company card, on a director's personal card, and inside larger platform invoices as add-on line items. No single statement shows the whole picture, so the running total only exists if someone sits down and builds it.

Second, add-ons hide the headline price. ClickUp Business looks like $12 a seat until you add Brain at $7 a seat and the per-seat figure jumps by more than half. Notion Business at $20 a seat looks like a wiki cost until you remember it is doing the job of three tools badly. The number you remember signing up for is rarely the number you are paying now.

Third, renewals roll silently. Free trials convert. One-off purchases quietly become annual subscriptions. The person who set up a time tracker three years ago has left, and the seat is still being billed against a project that closed in 2024. BetterCloud's 2026 data puts the average company at around 106 SaaS apps, with smaller firms near 42, and a meaningful slice of any stack that size is running on autopilot.

The three-bucket model: gaps, overlap, dead weight

Every tool you pay for falls into one of four states, and three of them cost you money. The audit exists to sort your stack into these buckets so the decisions become obvious.

Gaps are jobs your team does manually that software should be doing. The classic agency gap is real-time margin: the proposal lives in one tool, the time in another, the invoice in a third, and nobody can see mid-project that a phase has burned 70% of its budget at 50% of the deliverable. A gap does not appear on the bill, which is exactly why it goes unnoticed. It shows up as unbilled hours and as projects that looked fine until reconciliation.

Overlap is two or more tools doing the same job. You bought Asana, then a freelancer onboarded the team into ClickUp, and now you pay for both while people quietly split their work across them. Overlap is the most expensive bucket because you are paying twice and the context-switching tax lands on top. Dead weight is the simplest: tools nobody opens that still hit the statement. Trials that converted, seats for people who left, the design tool you replaced but never cancelled.

How to actually run the audit

The mechanics are dull and that is the point. Pull the last twelve months of statements from every card and payment account, not the last month. A month misses anything billed annually, which is usually the most expensive part of the stack. Export to a single spreadsheet so you are working from one surface.

Next, list every recurring software charge as one row each, and write the real annual figure beside it: per-seat price multiplied by active seats multiplied by months active. A 15-seat ClickUp Business stack is $12 x 15 x 12, so $2,160 a year, and Brain on top is another $1,260. Harvest Pro at $13.75 a seat is $2,475. HubSpot Sales Hub Professional at $100 a seat on a five-seat minimum is $6,000. Notion Business at $20 a seat is $3,600 if the whole team is on it, and Slack Pro at $8.75 a seat is $1,575. Doing the multiplication by hand is what makes the total feel real.

Then tag every row by purpose: project management, time, CRM, proposals, invoicing, chat, docs, AI. Sort by that tag and the buckets fall out on their own. Two rows under "project management" is overlap. A purpose with no row, like live margin tracking, is a gap. A row nobody can explain is dead weight. Add a "last used" and an "owner" column while you are there, because a tool with no internal owner is a tool that will quietly renew forever.

The consolidation question: merge or keep best-of-breed?

Once the buckets are clear, the real decision is consolidation. The instinct is to keep the single best tool for every job, the best-of-breed approach, and for some functions that is genuinely right. A best-of-breed tool earns its place when the job is specialist, the integrations are clean, and the data does not need to flow anywhere urgent. Bookkeeping is a fair example: most agencies should keep Xero or QuickBooks rather than fold accounting into an operations tool.

Consolidation wins when the tools are passing the same data between each other and the seams are where your margin leaks. Proposals, project plans, time, and invoicing are the same data viewed at different stages, and when they sit in four tools no single screen ever shows whether a project is profitable. The test is simple: if two tools are constantly being reconciled by a human in a spreadsheet, the integration is the work, and that is the case for merging them. Keep best-of-breed where tools are independent. Consolidate where they should have been one tool all along.

This is the gap we eventually gave up trying to bridge with integrations and built around instead. AgencyFlo runs project management, time, CRM, proposals, contracts and invoicing on one data model, with FloAI inside the loop rather than as a separate per-seat add-on, so live margin is a view rather than a month-end export. Pricing is flat: $50 a month for teams up to 25 people, $100 a month above. The relevant point for the audit is structural, not the price: the consolidation removes the seams where the spend and the margin leakage both hide.

Do this week

You do not need the full audit to start. Pull twelve months of statements and write the total annual software spend on a single line. That number alone changes the next renewal conversation, and most owners I have shared this with come back surprised by how far it sits above their guess.

Then pick the one purpose where two tools overlap and cancel the weaker. One decision, made this week, usually pays for the hour the full audit takes. The stack got expensive one silent renewal at a time, and it gets cheaper the same way, one deliberate decision at a time.