Building A Multi-tenant Price-monitoring Saas Data Layer For Scalable Retail Intelligence

By Author: Actowiz Solution
Total Articles: 37
Comment this article

Client Snapshot
A B2B SaaS startup building a price-monitoring product for mid-market retailers. They owned the application and billing; they needed the data engine underneath it.
The Challenge

The founders had built a working single-tenant prototype that scraped [VERIFY: 3] sites. It broke roughly [VERIFY: weekly]. Scaling it to a paying multi-tenant product surfaced problems the prototype had never faced:
Each tenant wanted different sites, different SKUs, different check frequencies
Anti-bot systems began blocking as volume rose
One tenant's heavy schedule could starve another's
Every site layout change meant a silent data gap nobody noticed until a customer complained
Objectives
A data layer supporting [VERIFY: 7–10] retail sources with per-tenant configuration
Scheduled collection with per-tenant isolation and fair scheduling
Anti-bot resilience with graceful degradation, not silent failure
Historical price storage and change-based alerting
Health monitoring that catches breakage before customers do
The Actowiz Approach
We separated ...
... the system into four layers: collection (per-source workers), scheduling (queue with per-tenant quotas so no tenant can starve another), storage (append-only time series, never overwriting history), and monitoring (per-source health checks with expected-yield thresholds).
That last layer is the one prototypes always omit and production always needs. If a source that normally returns 5,000 records suddenly returns 40, that's a layout change, not a market event — the system flags it and alerts rather than writing garbage into a customer's dashboard.
Anti-bot handling used a rotating session pool with per-source fingerprint profiles, and a fallback ladder: if the light method fails, escalate; if escalation fails, alert rather than retry infinitely.
Data Delivered
Tenant ID, source, SKU, price, availability, seller, captured_at, run_id, source_health_status.
Format: API + Postgres · Cadence: Per-tenant configurable
Results
[VERIFY: 10] sources in production
Uptime improved from [VERIFY: weekly breakage] to [VERIFY: 99.2%] successful scheduled runs
Source breakage detected automatically in [VERIFY: under 30 minutes], before customer-visible impact
Client onboarded [VERIFY: their first paying tenants] on the new layer
Compliance Note
Public product and pricing data only. Per-source rate limits configured conservatively.