Assemble your data stack by stage — Shift

Not a list of tools, a connected system

You now know the bricks: analytics, behavior, centralization, dashboards, KPIs, tracking, decision. The trap would be to install everything at once. The right approach is the opposite: build the minimum that answers your current questions, cleanly connected, then add a brick only when a real gap appears. A data stack is judged by its fluidity, not its number of tools.

Stage 1 — Zero budget, getting started

At launch, the goal is simple: know where people come from, what they do, and how many buy — without spending a cent.

Analytics: Plausible (if a small budget is fine, ~€9) or free GA4; free Microsoft Clarity for behavior.
Centralization: a Google Sheet kept by hand, updated every week.
Dashboard: a chart tab in that same Sheet, or free Looker Studio.
Tracking: UTMs on all links, a one-page tracking plan.
KPIs tracked: visitors, conversion rate, sales, revenue.

Cost: €0 to €9/month. Setup time: half a day. This stage is plenty until your first few hundred visitors and first few dozen sales.

Stage 2 — It's taking off, automate

When volume grows and manual entry becomes painful, you automate collection and dig deeper into behavior.

Analytics: keep the same, configure conversions seriously.
Behavior: Clarity or Hotjar to understand friction points.
Centralization: Make or Zapier automatically feeds the Sheet (Stripe → sales, form → prospects).
Dashboard: Looker Studio connected, emailed every Monday.
KPIs: add CAC, LTV, 30-day retention.

Indicative cost: €20 to €60/month. You start measuring profitability, not just activity.

Stage 3 — Scaling, structuring

At growth, data comes from everywhere and calculations exceed the spreadsheet.

Product analytics: PostHog or Mixpanel for journeys and cohorts.
Warehouse: BigQuery + an ETL (Airbyte, Fivetran) replicating sources.
Dashboard: Metabase or Looker Studio on the warehouse.
Decision: systematic A/B tests, cohort analyses, a tracked North Star Metric.

Cost: variable, from a few dozen to a few hundred euros depending on volumes. You arrive here only when the previous stages show their limits — never in anticipation.

The five-step method

graph TD
    A[1. List my 3-5 key questions] --> B[2. Pick the minimal tool per question]
    B --> C[3. Lay down a tracking plan + UTMs]
    C --> D[4. Centralize into ONE dashboard]
    D --> E[5. Set the weekly reading ritual]

List the questions your data must answer (not the tools — the questions).
Choose the minimal tool that answers each, favoring those that connect.
Lay down the tracking: naming plan, UTMs, GDPR compliance.
Centralize into a single dashboard readable in one minute.
Set the ritual: a weekly slot where you read and decide.

The mistakes to avoid

Over-equipping: ten tools installed, none looked at. Better a Sheet that's read than a PostHog that's ignored.
Measurement without ritual: data piling up without ever triggering a decision.
Sloppy tracking: no UTMs, no plan, so uninterpretable data — the most costly mistake.
Vanity metrics on the wall: a reassuring dashboard that says nothing actionable.
Premature over-investment: building a data warehouse for ten sales a month.

The final test of your stack

A successful data stack passes a simple test: on a Monday morning, in under fifteen minutes, you know whether your week was good, why, and what to act on. If you manage that with a Google Sheet, your stack is excellent. If you have ten tools and can't, your stack needs redoing. Sophistication is never the goal; fast, sound decision-making is.