Free AI Spend Assessment
Token Waste Score
28% of Northwind's AI spend is recoverable.
That is ~$30,900 a year.
Prepared for
Northwind Labs (example)
Data analysed
1 Apr – 30 Sep 2026
6 months · 412,880 requests
Source
OpenRouter CSV
Issued
16 Oct 2026
Recoverable spend
28%
Of your average monthly AI API outlay
Annualised saving estimate
$30,900
Indicative range $26,300 – $35,500
Average monthly spend
$9,200
Last 3 months · $2,576/mo recoverable

Run rate = average of the most recent months (up to 3) × 12. Five structural detectors passed over every request in the export. Your export is deleted within 24 hours of delivery.

Where the waste sits
Share of your monthly AI spend by status. Two findings are shown on this page; three more are detected and held for the full report.
$1,130
$790
$656
Efficient spend · 72%
Finding 1 · Model-task mismatch Finding 2 · System prompt bloat 3 further findings (full report) No action needed
Your top two findings Monthly recoverable · share of identified waste
$1,130/mo44% of waste
Model-task mismatch
Frontier-tier models are handling routine, short-output tasks. About 61% of requests to your most expensive model return under 150 output tokens and match a classification or extraction pattern, where lower-cost models typically perform equivalently.
Full report: the workflows affected, the recommended substitute model per task, and the engineering steps.
$790/mo31% of waste
System prompt bloat
The same long instruction block is re-sent on most requests. Median input is 6,800 tokens against a median output of 410, and the leading block repeats across 88% of calls without prompt caching applied.
Full report: which prompts to compress or cache, the target token budget, and the expected quality impact.
Held for the full report$656/mo detected
3
Agentic pattern
4
Caching
5
Trajectory risk
Northwind Labs · Token Waste Score · October 2026
1 of 2
Unlock the full report
Page 2 of 2

Your score found ~$30,900 a year. The full report shows how to recover it.

The Waste Score tells you how much and where. The full diagnostic tells you exactly what to change, in what order, and what it is worth, ranked by saving against engineering effort.

What you getWaste Score (free)Full report ($3,500)
Recoverable % and annualised saving estimate✓✓ exact
Top two findings by category✓✓
All findings, ranked by monthly saving vs fix effort—✓
Named workflows and model substitution per task—✓
Engineering implementation notes for every fix—✓
Spend trajectory at 3× and 5× usage and under price changes—✓
Vendor concentration and model lifecycle risk flags—✓
CFO-ready PDF and HTML, plus a 30-minute walkthrough call—✓

Get the full ranked fix list for Northwind

Same export, no new data needed. Delivered within 2 working days of payment. Includes a 30-minute walkthrough with your engineering lead. A follow-up delta report to confirm the saving is available at $1,250.

$3,500
Fixed fee · one report
Commission the full report
Price held until 15 Nov 2026
Every figure shows its working

Every saving in the full diagnostic shows its working: the requests affected, your actual tokens and cost, the recommended alternative and its published price. Your team can check every figure against your own export.

How the assessment and full diagnostic work
Plain English. The full terms of use apply.
  1. Data period: at least one month of usage in one export. Three months or more gives a firmer estimate and better trend analysis.
  2. Best fit: teams spending $5,000 or more a month on LLM APIs. Below that, the full report may not pay for itself, and we will say so.
  3. Supported sources: OpenRouter CSV exports and TypingMind JSON exports, including Anthropic and OpenAI usage run through TypingMind. Other gateways, such as Portkey, on request.
  4. Complete export: the file should cover all production usage for the account it comes from, unfiltered, so the figures reflect your real spend.
  5. How savings are measured: annualised = average monthly identified saving over the most recent months available (up to three) × 12, at the provider prices in force on the report date. Figures are identified opportunity, not implemented savings.
  6. Indicative figures: this assessment is an automated estimate (±15%). The full report replaces it with line-by-line figures, each showing its calculation.
  7. Payment: the full diagnostic is a fixed fee of $3,500, payable when you commission it. It is delivered within 2 working days of payment.
  8. Data handling: your file is deleted within 24 hours of report delivery, and email copies are removed from our inbox on receipt. Where prompt text is present (TypingMind and OpenRouter API exports), up to the first 500 characters of each prompt are processed only to classify the task type.
  9. Validity: the price on this assessment is held for 30 days from the issue date.
TokenomicsIQ · tokenomicsiq.ai · guy@neowise-ventures.com
2 of 2