Sales data across all channels

Technically there's no hard limit — but in practice a staggered approach pays off.

Phase 1 (typically 3–5 sources): the most important sales channels plus one CRM source. With that you usually cover 80–90% of revenue and can base the biggest decisions on it. Setup effort: 6–10 weeks.

Phase 2 (expansion to 8–15 sources): marketing data (Meta, Google Ads), email platforms, warehouse and logistics systems. Here the focus shifts from “measuring sales” to “steering the business”. Per source 1–2 weeks setup effort.

What determines the limit: not the tool choice, but the data maintenance. More than 20 active sources rarely make sense without a dedicated data team.

The answer depends on which decisions you make on the data — real-time is more expensive and complex and not sensible for every use case.

Cost difference: real-time pipelines typically cost 3–5× more than daily batches — both infrastructure and setup.

Practical advice: start with daily aggregation. If concrete real-time use cases show up after 6 months, switch over individual streams selectively — not everything at once.

• Daily aggregation is enough: operational reports (CFO, marketing, branch managers), weekly sales reviews, monthly reporting routines, strategic cohort analyses
• Real-time is worth it: operational out-of-stock warnings, real-time marketplace pricing, live dashboards for promotions, fraud monitoring

BigQuery has a pay-per-use logic with two main cost drivers: storage and query costs.

Comparison to alternatives: a classic BI SaaS tool for comparable functionality often costs €500–2,000 monthly per seat — at 5 users that's quickly €2,500–10,000. BigQuery is almost always cheaper over total cost of ownership.

• Small (1–5M sales events/month, 3–5 sources): typically €30–80 storage + query
• Medium (5–30M events/month, 5–10 sources): typically €80–250
• Large (50M+ events/month, 10+ sources, many dashboards): €300–800
• Enterprise (real-time + many concurrent users): from €1,000 — reserved-capacity contracts pay off here

One of the most important decisions — and it depends more on your team than on the data volume.

Hybrid is often the best solution: standard sources via cloud tools (e.g. Airbyte for Shopify, Stripe), custom sources via self-build (e.g. Lightspeed API, marketplace reports). dbt as the transformation layer is the same in both cases.

Rule of thumb for the cost trade-off: cloud tools often cost €100–500 monthly per source — at 10 sources that's €1,000–5,000 per month. Self-build has setup costs of 4–8 hours per source, then ongoing costs of €0. From 6+ sources, break-even is usually reached after 12 months.

• Cloud tools are worth it: no internal data engineers, mostly standard APIs, fast time-to-value more important than optimized costs
• Self-build is worth it: proprietary sources, data volume exponential for SaaS, special transformation logic, compliance requires full control

Yes — and that's one of the most important insights you internalize after the first multi-channel projects. Building a single view for all stakeholders is almost always a mistake.

Important: all views are based on the same raw data in the data warehouse. Only the aggregation and visualization layer is tailored per audience. Otherwise the data conflicts between departments — which we actually wanted to avoid — arise again.

• CFO view: gross revenue, net revenue, tax per channel & location, returns-adjusted figures, forecast comparison, margin/contribution
• Marketing view: performance per campaign/channel/creative, cohort analyses, CLV, repeat rate, channel attribution, real-time during promotions
• Branch/sales view: daily revenue per location and employee, top/flop products, comparison to the previous period and to other branches

Yes, in most cases historical data can be built up retroactively — with a few concrete limitations per source.

Typical approach: an initial backfill of the last 12–24 months during setup, after which the pipeline continues on ongoing data. For larger data volumes, the backfill is done in weekly chunks.

Important: reconstruction is a one-time investment with clearly estimable effort. Ongoing data quality is more important long-term than historical depth — if you had to choose between 5 years retroactively and a clean pipeline for the future, choose the pipeline.

• Reconstructable well: Shopify (2 years retroactive via API), marketplaces like Amazon/Otto/Zalando (1–2 years), POS systems (mostly locally available), CRM data (complete)
• Not, or only partly, reconstructable: data from shut-down sources, manual Excel corrections without an audit trail, aggregated reports from tools without raw-data export

A typical intro call for multi-channel sales data lasts about 30–45 minutes and is deliberately designed to classify your case — not to sell a finished package.

What it isn't: a sales call with pressure. If your case is too small, too large or topically unsuitable for me, I say so openly — and name suitable alternatives if needed.

• What we discuss: which sales sources exist today and which are missing, where the biggest data pain sits, what data knowledge exists in the team, which tools are already used
• What you have afterwards: an assessment of whether a tracking audit or a data-architecture project makes more sense, a rough order of magnitude (effort, duration, cost), concrete recommendations for action even without further collaboration