KPI dashboards.

What “KPI dashboard” actually means#

A small set of executive-grade views — typically three or four — that show the metrics the leadership team would defend in a board meeting. Pipeline, qualified leads by source, cost per qualified lead, retention, expansion, NRR. The metrics that map to revenue. Not bounce rate, not impressions, not engagement-rate-of-LinkedIn-posts. The non-commodity test for content (per SEO) applies to dashboards too: would this dashboard exist if we hadn’t built it?

What’s wrong with the default#

The default agency dashboard is a Looker/Studio template hooked up to Google Analytics, with the agency’s logo in the corner. It shows sessions, average session duration, top pages, top sources, and a goal-conversion number that no one in the leadership team trusts. It does not show pipeline. It does not show CAC. It does not connect a tracked-by-channel inbound to a closed deal in the CRM. Six weeks in, no one opens it.

How we build them#

On your real data#

Not on GA’s sampled data. We pull from the CRM (HubSpot, Salesforce, Pipedrive, Attio, whatever the firm uses), the billing system (Stripe, Chargebee, internal), the support system, and only then from the marketing-channel sources — and we model the joins so a tracked-by-channel inbound shows up as a closed deal six weeks later in the same view.

Modeled, not assembled#

The right tool is a small data warehouse (BigQuery, Snowflake, or Postgres with dbt) and a thin BI tool on top (Hex, Metabase, Lightdash). We model the data once, then surface it through whatever interface the team prefers. We do not assemble dashboards by dragging fields onto a canvas in Looker Studio.

Few views, each load-bearing#

Three or four views, each one tied to a decision the leadership team makes. A “pipeline this quarter” view that maps every closed deal to the channel that originated it. A “cost per qualified lead by source” view that shows where the next dollar should go. A “retention and expansion” view. A “channel-mix-over-time” view. That’s it. We don’t build the fifteenth tab.

What we don’t sell#

We don’t sell read-only Looker Studio templates with the agency’s logo on them. We don’t sell a “dashboard service” billed monthly to maintain a static report. We build the system, document it, hand it off so the firm’s analyst (or future analyst) can extend it, and operate the parts that genuinely require ongoing operation (data pipeline reliability, schema changes, source migrations).

Frequently asked questions#

Can you build this on top of Google Analytics?

We can include GA4 as one source, but we won’t make it the primary one. GA’s sampling, attribution modeling, and AI-referral misclassification (~70% of AI traffic mis-bucketed as direct, per published research) make it structurally unfit to drive executive decisions on its own.

Which BI tool do you prefer?

Hex for analyst-flavored work; Metabase for “executives need to self-serve”; Lightdash if the team is already on dbt. We don’t have a religious preference. The choice depends on the team’s analyst depth.

Do we get to take it with us if we leave?

Yes. Every dashboard we build runs on infrastructure the client owns — the warehouse is yours, the BI tool seat is yours, the dbt repo is yours. There is no lock-in by design.

Can you maintain it as our data changes?

Yes — that’s part of operating the function. Source migrations, schema changes, new channels, new metrics. The dashboard is alive, not a snapshot.