Step-by-step: negotiating a SaaS renewal with data

Just went through a renewal with our team's primary AI code assistant SaaS. We were staring down a 35% auto-renewal increase on the dashboard with a 30-day notice. 😬 Felt like a trap. Instead of just accepting it, I built a quick data-driven counter-proposal and wanted to share the exact steps. It worked—we ended up with a 15% increase (on more seats) and a 2-year lock-in.

The core idea: replace "this is too expensive" with "here's the data on our usage and market alternatives." You need three pieces of data:

1. **Your own usage metrics.** Export everything. For us, that meant:
* Monthly active users (seats actually used)
* Feature usage breakdown (e.g., % of commits using the advanced autocomplete vs. basic)
* API call volume (if applicable)

2. **Competitive benchmark data.** Don't just list competitors; find their *public* pricing. I scraped pricing pages and used archive.org for historical data. Structured it in a simple table.

3. **Your counter-offer.** Frame it as a "growth-based" proposal, not a discount demand.

Here's a snippet of the Python script I used to collate our usage logs and generate a simple summary. The key is to show you're a data-informed customer, not just negotiating blind.

```python
import pandas as pd
import matplotlib.pyplot as plt

# Simulated data load
usage_df = pd.read_csv('monthly_usage_export.csv')
# Calculate seat utilization
total_seats = 25
active_seats = usage_df['user_id'].nunique()
utilization_pct = (active_seats / total_seats) * 100

print(f"Seat Utilization: {utilization_pct:.1f}%")
print(f"Avg. Monthly Active Users: {active_seats}")
```

This output went into a one-page PDF alongside the competitive pricing table. We sent it to our account rep with a clear ask: **"Can we align the contract to our actual usage pattern and the current market? Here's our proposed structure based on active users."**

The result? A much more productive call. They came down on the per-seat price when we agreed to a slightly higher commit (based on projected growth). The 2-year lock gives us predictability.

Has anyone else tried a similar data-driven approach? What metrics did you find most persuasive? I'm thinking of building a more generic script for this if there's interest.

-- Weave