Two different jobs
Reporting is operational. It exists so that the person managing a channel can see yesterday's spend, this week's cost per lead and whether the new campaign is delivering, and act on it the same morning. The data comes from the platform that produced it, the metrics are the platform's own, and the latency is whatever the platform offers. Reporting is also what every tool gives away: Google Ads, Meta, GA4Google's event-based analytics platform. and the CRM all ship with it, and a connector into Looker Studio stretches it a little further without changing its nature.
Business intelligence is analytical and cross-system. It exists so that the business can ask questions whose answer lives in more than one place: which acquisition channel produces customers who are still paying a year later, what the blended cost of a lead is once agency fees and tooling are included, whether the rise in organic traffic is cannibalising paid. Those questions need the data copied out of the platforms, joined on shared keys, kept for as long as the question needs and computed from one agreed definition. None of that is free; it is built, and then it is maintained.
The two are not stages in a maturity model where BI replaces reporting. A team with a warehouseA database designed for analysis that stores data loaded from many systems, such as analytics, ad platforms and the CRM, so it can be joined and queried. still needs its channel managers looking at platform reports every morning. The BI hub describes the stack; this page is about choosing how much of it to build, and when.
Reporting and BI compared
Marketing reporting vs Marketing business intelligence
| Row label | Marketing reporting | Marketing business intelligence |
|---|---|---|
| Question answered | What did this platform or channel do, in its own terms | What happened across channels, systems and time, and why |
| Data sources | One platform at a time, read through its interface or a connector | Many sources copied into a warehouse the team owns |
| Joins | None, or shallow blends on a date | Spend to sessions to leads to revenue on shared keys |
| History | Whatever each platform retains, in its own aggregation | Every row kept for as long as the business decides |
| Latency and freshness | Near real time for most platforms | Scheduled loads; hours to a day behind, by design |
| Metric definitions | The platform's own; they differ between tools | Defined once in a metrics layer; the same everywhere |
| Ownership | The channel manager who uses it | A data owner, with named owners per dataset and dashboard |
| Tooling | Platform reports, Looker Studio on connectors, spreadsheets | Warehouse, transformation, metrics layer, BI tool |
| Skills | Platform literacy; chart building | SQL, data modelling, some engineering, governance |
| Cost | Mostly time | Licences, warehouse usage, connectors and people |
| Change speed | Immediate; anyone can add a chart | Deliberate; changes go through review and versioning |
| Typical failure | Numbers that cannot be reconciled across tools | Over-engineering before anyone has a question that needs it |
What reporting does well
Reporting is fast, cheap and close to the action. A channel manager who sees spend pacing ahead of budget at nine in the morning can fix it by ten; no warehouse load, no model run, no ticket. The platform's own report is also the authoritative record of what the platform charged for and what it attributed to itself, which is exactly what is needed to run the channel. For a single-channel business, or a team of one, it is usually all that is needed, and the SEO KPIs and reporting guide shows how far a disciplined single-channel report can go.
Reporting also sets the questions that BI later answers. The first time two platform reports disagree, or a manager asks for revenue by first-touch channel and nobody can produce it, the gap has been named. Good reporting practice, meaning one dashboard per audience, a cadence, an owner and a short list of each audience's KPIA metric tied directly to an objective. figures, is also the foundation BI sits on. A team that cannot run clean reporting will not run clean BI either; it will have the same arguments at a higher cost. The reporting and BI overview covers the audience and cadence side of that discipline.
Reporting's limits are structural, not a matter of effort. It cannot join what it does not hold, cannot remember what the platform has aggregated away, and cannot apply a definition the platform does not offer. A blended view built by connecting five platform connectors into one Looker Studio page is still reporting: five truths side by side, not one.
What BI adds
Three things, and each requires the warehouse.
Joined data. Spend from the ad platforms, sessions and events from analytics, leads and deals from the CRM, orders and refunds from the commerce platform, all keyed so that a row in one can find its counterpart in another: click ids, order ids, hashed emails, campaign names that follow a convention. Once joined, questions that were impossible become routine. Cost per qualified lead by landing page. Payback period by acquisition channel. Retention by first campaign. The SQL for marketers guide teaches the query shapes those questions take.
History under the team's control. Platforms change their interfaces, their attribution defaults and sometimes their counting; the warehouse keeps what was loaded, as it was loaded. Year-on-year comparisons stop depending on whether the platform still shows that period. Cohorts can be followed for as long as the business wants. Restating a metric after a definition change is possible because the raw rows still exist.
One definition of each metric. The metrics layer holds the formula, grain, filters and owner for every metric once, and every dashboard reads from it. When finance and marketing use the word "conversion", they mean the same rows. Disagreements become a question about which definition to use, rather than an argument about whose export is right.
What BI does not add is speed, simplicity or cheapness. The data arrives on a schedule, the model has to be maintained when a platform changes its export, and the dashboards still need governance or they sprawl exactly as spreadsheets did.
The transition
Teams rarely move from reporting to BI in one project, and the ones that try tend to build a warehouse nobody queries. The pattern that works is to let each named gap pull in one layer of the stack.
- Fix the reporting first. One dashboard per audience, an owner, a cadence, a written definition of each metric even if it is only a document. Much of what looked like a BI problem turns out to be a definitions problem.
- Land the first raw table. For most teams this is the GA4 export into BigQuery, because it is documented, free to enable and gives the richest event data. Nothing reads from it yet; it accumulates history from the day it is enabled, which is the point.
- Bring spend beside it. A connector or a scheduled export for each ad platform into daily spend tables with a shared channel and campaign naming.
- Build the first join. One modelled table, spend to outcomes by day and channel, queried by one dashboard. Reconcile its totals to the platforms and document every difference. The tutorial on moving a Looker Studio report onto a warehouse shows that reconciliation for an existing dashboard.
- Move definitions into the layer. Once the second dashboard reads from the same model, the formulas move out of the dashboards and into one place.
- Govern what now exists. Inventory, certification, retirement. The Looker Studio section covers what changes when dashboards read from a warehouse rather than from connectors.
The third step hides most of the surprises, and the guide to loading ad platform data into a warehouse covers the refresh windows for late conversions, attribution settings, currencies and renamed campaigns it involves.
A software company's twelve-month transition
A business-to-business software company with a marketing team of six runs paid search, LinkedIn, webinars and content. In January the monthly report is a 30-slide deck built from five exports; cost per marketing-qualified lead is reported as £212 by marketing and £340 by finance, who include agency fees and the webinar platform. By March the team has one Looker Studio dashboard per audience and a two-page metric dictionary that settles the fee question. In April they enable the GA4 BigQuery export and, in June, add daily spend tables from both ad platforms through a managed connector costing roughly £90 a month. In August the first modelled table joins spend to CRM leads on the click id for 78 percent of leads and on campaign name for the rest, and a single dashboard reads from it. By December the dictionary lives as code in the metrics layer, the deck is generated from the certified dashboard in under an hour, and the finance and marketing figures agree to the pound because they are the same query.
Common mistakes
- Building the warehouse before naming a question. Tables nobody queries are a cost, not an asset.
- Calling a blended Looker Studio page "BI". Five connectors side by side are still five truths.
- Skipping the definitions step. Most reconciliation arguments are definition arguments and need no warehouse to settle.
- Expecting BI to be as fresh as a platform report. Scheduled loads lag by design; channel managers still need the platform.
- Letting the model fall behind the platforms. Export schemas and attribution defaults change; an unmaintained model quietly drifts.
- Treating BI as a replacement for reporting. It is an addition. The morning check of the platform report does not go away.
- Starting without an owner. A warehouse with no named owner becomes the spreadsheet problem with a bigger bill.
