Vendor and SaaS spend is where money leaks quietly: two tools that do the same thing, a subscription that auto-renewed at triple last year's price, twenty seats for a team of eight, a contract that renews next month and nobody's watching the date. It never shows up as a crisis — it shows up as a burn rate that's higher than it should be and no one can say why. This is the recurring ops loop that catches it: consolidate every vendor into one view, surface the duplicates and price creep, and build a renewals calendar so the next auto-renew is a decision you made, not a charge you discover.
- Your spend data with vendor and amount —
transactions.csvcleaned via Make a messy export trustworthy and categorized with yourcategorization-rules.md, plus the trailing 12 months so you can see renewals and price changes over a full year. - Your contracts or subscription list if you have one — renewal dates, seat counts, and committed terms. If you don't, this playbook helps you reconstruct it from the charges.
- Headcount or usage context — how many people actually use each tool — so 'unused seats' means something real, not a guess.
-
Consolidate every vendor into one view
You can't manage spend you can't see. In Claude Desktop, open the folder with your trailing-12-months data and ask in the chat — no terminal needed. Have Claude roll all charges up by vendor first — total spend, charge frequency, and trend — so scattered line items become one ranked list.
you askRead the last 12 months of transactions. Roll everything up by vendor: total annual spend, how often each charges (monthly, annual, irregular), the category, and whether spend is rising, flat, or falling over the year. Sort by total spend, biggest first. Don't flag anything yet — I just want the full vendor picture in one table.what you get back A ranked vendor view: "Payroll Co 165,000/yr (monthly, flat). Cloud Co 52,000/yr (monthly, rising 8%/mo). Design Tool 14,400/yr (annual, jumped from 9,600). 38 vendors total." The scattered charges are now one list you can actually reason about.
Just seeing every vendor ranked by annual spend is half the value — most teams have never looked at their tools as a single sorted list, and the duplicates and surprises jump out immediately.
-
Surface the duplicates, price hikes, and dead weight
Name the three patterns that leak money — overlapping tools, silent price increases, and unused capacity — so Claude checks each deliberately instead of giving a vague 'looks reasonable.'
you askNow flag three things across these vendors: (1) likely duplicate tools — two or more vendors that do the same job (two project trackers, two e-sign tools), (2) price hikes — any vendor whose per-charge amount rose more than 15% over the year, and (3) probable dead weight — annual tools with one charge and no sign of ongoing use. For each, show the evidence and mark it as a candidate to review, not a confirmed cut.what you get back A flag list with evidence: "Duplicate: Asana 12,000 and Monday 9,600 — two project trackers, consolidating one saves ~10k. Price hike: Design Tool 9,600 → 14,400 (+50%) at renewal. Dead weight: Webinar Co 4,800 annual, single charge, no usage in the data." Each tagged as something to check, not auto-cancel.
-
Build the renewals calendar
The charge you can't stop is the one you didn't see coming. Turn the annual and committed charges into a forward calendar so every renewal has a date and an owner before it hits.
you askFrom the annual and recurring charges, build a renewals calendar for the next 12 months: each vendor, its expected renewal month, the amount, and how many days' notice you'd need to cancel or renegotiate. Flag any renewal in the next 60 days as urgent. Note where you're inferring the renewal date from the charge pattern rather than a real contract date, so I know which to confirm.what you get back A forward calendar: "July: Design Tool 14,400 (URGENT — renews in 3 weeks, was the +50% one). September: CRM 28,000 (90-day cancellation notice — decide by June). 6 renewals inferred from charge dates — confirm against contracts." Now no renewal is a surprise.
The 90-day-notice renewals are the dangerous ones — miss the window and you're locked in for another year. Surfacing the notice period, not just the renewal date, is what gives you room to renegotiate.
-
Turn it into an action list and keep it re-runnable
End with a prioritized worklist and the saved view, so this becomes a quarterly habit, not a one-time spring clean.
you askPull it together into vendor-spend-review.md: the ranked vendor view, the flagged duplicates/hikes/dead-weight with estimated annual savings each, and the renewals calendar. Then give me a prioritized action list — the 5 highest-savings moves, biggest first, each with the saving and what to verify before acting. Save it so I can re-run this every quarter.what you get back A
vendor-spend-review.mdplus a ranked action list: "1. Consolidate Asana/Monday (~10k/yr) — confirm both teams agree. 2. Renegotiate Design Tool before July renewal (~5k). 3. Cancel Webinar Co (4.8k) — confirm truly unused." Evidence-backed cuts, not vibes — and a re-runnable review.
- Feeds the cash-flow model and the budget: the consolidated recurring total is the real burn number for Build a cash-flow model you understand, and the cuts you confirm flow into next quarter's budget vs. actuals — run this first and your burn rate is grounded, not guessed.
- Per-team or per-owner cut: add a column for which team owns each tool, then ask Claude to summarize spend per team — so the renewal conversation goes to the person who can actually decide to keep or cut it.
- Multi-currency vendors: if vendors bill in different currencies, have Claude normalize to your reporting currency (state which rate and date) before ranking — an un-normalized list ranks vendors by exchange noise, not real spend.
- Renewal watch on a schedule (Power Track): once the calendar exists, a scheduled agent (see the Features tab) can check it monthly and surface any renewal entering its notice window — so the alert finds you instead of you remembering to look.
- A flag is a candidate, not a verdict. A 'duplicate' tool might serve two genuinely different teams; 'dead weight' might be a critical once-a-year service. Claude surfaces the candidate and the saving; a human confirms before anything is cancelled.
- Confirm inferred renewal dates against real contracts. A renewal date guessed from the charge pattern can be wrong by weeks — and missing a notice window costs you a full year. Use the calendar to know where to look, then verify the dangerous ones against the actual contract.
- Cutting a tool is a business decision, not a line-item one. The 5k saving is real; so is the workflow that breaks if you cancel the wrong thing. Route every proposed cut past the team that uses it before acting.
- Vendor contracts and pricing can be confidential. Keep them in your approved workspace, swap any commercially sensitive terms for a
[placeholder]before sharing, and remember the cut/keep/renegotiate call is yours to own.
you'll end up with A consolidated vendor and SaaS spend review — every tool ranked, duplicates and price hikes and dead weight flagged with estimated savings, and a renewals calendar with notice windows — so you cut spend with evidence and no contract ever auto-renews by surprise again.
Questions people ask
- What does this find that a normal spend audit doesn't?
- The spend audit looks at *one month* for anomalies — outliers, duplicates, new vendors. This playbook takes the *full year* per vendor to surface the slow leaks a single month can't show: overlapping tools you're paying for twice, prices that crept up 50% at renewal, and annual subscriptions nobody uses. Run the monthly audit for anomalies and this quarterly review for structural vendor savings.
- How does the renewals calendar help if I don't have my contracts handy?
- Claude reconstructs a likely calendar from the charge pattern — an annual charge in September implies a September renewal — and flags which dates are inferred so you know to confirm them. It won't be perfect without the contracts, but it turns 'I have no idea when things renew' into a dated list you can verify, with the urgent next-60-day ones surfaced first.
- Should I cancel everything Claude flags as a duplicate or dead weight?
- No — every flag is a candidate to review, not a confirmed cut. A 'duplicate' might serve two different teams and 'dead weight' might be a critical annual service that just charges once. Claude shows the evidence and the estimated saving; a human confirms with the team that uses the tool before anything is cancelled.
- How often should I run this?
- Quarterly is the sweet spot — often enough to catch renewals before their notice windows close, rare enough that the trailing-12-months view has new signal each time. Save the `vendor-spend-review.md` so each run is an update, and pair it with the cash-flow model so your burn rate always reflects the real recurring spend.