We ran the numbers: Claw CRM promised a 60% lower TCO and sub-second dashboard loads. Salesforce was costing us $142/user/month with slow reports.
We cut over all 84 sales reps in one weekend. The forcing function was the annual contract renewal. No pilot.
First Monday:
- 19 users couldn't log in (SSO config mismatch).
- Forecast pipeline dashboard timed out (30+ seconds). The Claw schema didn't match our custom Salesforce object relationships.
- Data sync from our product database failed silently. Missing rows for 3 days.
Root cause: We assumed performance claims applied to our volume (2.3 TB user event data). We never tested with our full dataset.
The Claw aggregate query for the main sales dashboard:
```sql
-- Their example query (optimized for their demo dataset)
SELECT region, SUM(amount) FROM deals GROUP BY region;
-- Our actual query pattern (after join with custom objects)
SELECT
r.name,
COUNT(DISTINCT d.id),
AVG(d.value / NULLIF(o.quota, 0))
FROM deals d
JOIN custom_objects o ON d.rep_id = o.rep_id
JOIN regions r ON o.region_id = r.id
WHERE d.close_date BETWEEN ? AND ?
GROUP BY r.name
HAVING COUNT(d.id) > 5;
```
Their query planner did a full scan on `deals` because our date filter was on a joined table. No warning.
Where things slipped:
- No load test with >50 concurrent users.
- No validation of ETL mapping for custom fields.
- Assumed their "Salesforce importer" handled all customizations.
We restored Salesforce after 72 hours. Cost: $18k in overtime for engineers, lost deal visibility, and a 34% drop in outbound calls that week.
Lesson: Pilot with one team. Test with your worst-case query. Validate data completeness before sunsetting.
Numbers don't lie.