Per-row pricing looks great on paper for small datasets. Then you get that first invoice after a few analytical workloads and realize you just paid $500 to scan a 10TB table.
It's a trap because it decouples cost from actual value. You're penalized for exploratory queries, full table scans during development, and any inefficient joins. The vendor has zero incentive to help you optimize your queries because their revenue goes up when your queries are expensive.
Real-world example from a migration I reviewed last month:
- Client moved from BigQuery (on-demand, per-TB scanned) to a vendor charging per-row processed.
- Their monthly dashboard query: joins a 50M row fact table with a 10M row dimension table.
- Under per-TB pricing: ~$5 per run (scanning only necessary columns).
- Under per-row pricing: 60M rows processed * $0.10 per 10k rows = $600 per run.
The killer? That's just *one* dashboard refresh. Add in ad-hoc analysis and data science work, and you're looking at five-figure monthly bills for what was previously a few hundred dollars.
If you're considering a per-row model, you must:
* **Instrument everything** before signing. Log all query patterns, row counts, and frequencies.
* **Demand detailed, query-level billing** from the vendor. If they can't provide it, walk away.
* **Calculate the worst-case scenario**: (Total Rows in All Tables) x (Query Frequency). That's your potential maximum.
* **Negotiate a hard cap or tiered pricing** where costs plateau after a certain volume.
The core problem is misaligned incentives. Your goal is to derive insights from data. Their revenue increases when you inefficiently access that data. That's a dangerous partnership.
cost optimization, not cost cutting
You've nailed the operational risk, but the procurement angle is just as dangerous. Vendors love to pitch per-row models with "predictable growth," but that predictability only works one way. Your costs scale linearly with usage, but your budget approvals don't.
I've seen teams get locked in because their initial POC cost was trivial. By the time finance sees the real bill, you're already dependent on the platform, and re-architecting to leave is a six-month project. The negotiation leverage evaporates after you sign.
The real trap isn't the pricing model itself, it's signing up without a hard usage cap or a firm commitment to fixed-cost tiers in the contract.