Section B · Data systems

BI — Looker, Power BI, Tableau, Hex, Mode

Where business users actually look at the data. The customer's BI tool determines how your output reaches their decision-makers. Match what they have.

Looker

What it is

Google-owned BI tool with a strong semantic layer (LookML). Common at large enterprises that adopted the modern data stack pattern.

How to integrate

Deliver data to the customer's warehouse; build LookML models on top. Some platforms generate LookML programmatically and let customers extend.

Strengths

  • Semantic layer (LookML) enforces metric definitions across dashboards. Single source of truth.
  • Explore feature lets non-SQL users self-serve.
  • Enterprise governance and access controls.

Weaknesses

  • Expensive.
  • LookML is its own language; lock-in.
  • Iteration cycle is slow compared to notebook-style tools.

Power BI

What it is

Microsoft's BI tool. Ubiquitous in Microsoft-shop enterprises; tightly integrated with Office, Azure, Teams.

How to integrate

Deliver to the warehouse or Azure Synapse; Power BI connects via native connectors. Datasets (Power BI's semantic layer) are built on top.

Strengths

  • Excellent visualization library.
  • Embedded analytics options (Power BI Embedded).
  • Strong integration with Office tooling.

Weaknesses

  • Desktop-first workflow can feel awkward for cloud-native teams.
  • DAX (Power BI's expression language) has a steep learning curve.
  • Less self-serve than Looker.

Tableau

What it is

Long-time leader in data visualization. Common in finance, operations, and traditional enterprises. Salesforce-owned since 2019.

How to integrate

Deliver to the warehouse; Tableau connects via native connectors. Tableau extracts (.hyper files) for offline / accelerated query.

Strengths

  • Sophisticated visualization capabilities.
  • Strong workflow for analyst-led exploration.
  • Tableau Server / Cloud for enterprise governance.

Weaknesses

  • Weak semantic layer compared to Looker.
  • Workbook proliferation — definitions drift across dashboards.
  • Vendor lock-in.

Hex

What it is

Modern notebook + dashboard combo. Strong with technical analyst teams. Fast iteration cycle.

How to integrate

Deliver to the warehouse; Hex notebooks query directly. dbt models can be the semantic layer.

Strengths

  • Notebook UX for analysts.
  • App-mode for non-technical consumers.
  • Fast to ship prototypes.

Weaknesses

  • Smaller customer base; less mature governance.
  • Newer; less proven at very large enterprise scale.

Mode

What it is

Analyst-friendly notebook-and-dashboard tool. SQL + Python + R; reports as documents. Common at SaaS and tech-forward companies.

How to integrate

Connect to the warehouse via Mode's native connectors; SQL + Python in the notebook.

Strengths

  • Fast iteration for analysts.
  • Notebook-style reports.
  • Strong for data-team-internal use.

Weaknesses

  • Weak semantic layer.
  • Less polished for non-technical end users.

Delivery patterns for contract intelligence

1. Match the customer's tool

Do not introduce a new BI tool. The customer's analysts and stakeholders are already trained on what they have. Resentment from the existing BI team kills deployments quietly.

2. Ship a semantic layer they can extend

Build dbt models (or LookML for Looker, datasets for Power BI) that the customer's analysts can extend. Document the joins, the SLAs on underlying data, the known gotchas.

3. Three starter dashboards

Don't ship 30 dashboards. Ship 3, calibrated to the customer's named priorities. Common starter set:

  • Renewal pipeline: upcoming renewals with notice deadlines, color-coded by urgency.
  • Off-contract spend: spend per supplier with no active agreement, ranked by amount.
  • Supplier consolidation opportunities: suppliers with near-duplicate canonical names.

4. Pair-session with their analysts

In the last 2 weeks of deployment, pair with the customer's BI analysts to walk them through the semantic layer. They extend; you watch; you correct. By end of pairing, they should be building dashboards on top without your help.

5. Document the SLA in BI context

Every dashboard footer should reference data freshness, accuracy SLA, and the link to the underlying eval dashboard. The customer's analysts will field questions about why a number looks off; the docs help them answer without paging you.

The "BI is the customer's surface" rule

The customer's business users see the BI tool, not your platform's UI. Investment in semantic layer + starter dashboards + analyst training is investment in customer success. Skip it and the deployment may technically work but never get traction.

Takeaway

  • Match the customer's BI tool. A new tool invites resentment from the existing BI team, and that quietly kills deployments.
  • Ship a semantic layer the customer's analysts can extend, in their own idiom: dbt models, LookML or Power BI datasets.
  • Ship three starter dashboards tied to the customer's named priorities, not thirty.
  • Pair with their analysts in the last two weeks, until they build on top of the layer without you.
  • Put data freshness, the accuracy SLA and a link to the eval dashboard on every dashboard.
  • Business users see the BI tool and never your platform, so the semantic layer and the analyst training decide whether the deployment gets traction.

Many contracts never reach a CLM, so the next question is which file system holds them. → Document sources

Further reading

  • What is LookML? — Google Cloud docs, accessed 2026. How Looker's semantic layer defines metrics once for every dashboard.
  • Semantic models in the Power BI service — Microsoft Learn, accessed 2026. What a Power BI dataset is and how reports reuse it.
  • Tableau Hyper API — Tableau, accessed 2026. The .hyper extract format the chapter mentions, and how to write it.