Brandon Cantrell.

Brandon Cantrell · Sr. Manager, BI (Analytics Lead) · LERMA/ · Dallas, TX

I build analytics systems people actually use.

I also build the teams and platforms behind them. I'm a player-coach, and I'm happiest when I ship something that makes a real difference in how people work.

I'll be honest up front. A dashboard nobody uses is just a screensaver with a database behind it. I've built a few of those in my career, and I don't build them anymore. I was the first BI hire at LERMA/, so I got to build its data platform from scratch: governed metrics, a semantic layer, and AI-assisted self-service that executives actually run the business on. I lead a team of 3 BI engineers and I still ship models myself, because the day I stop building is the day I stop understanding the work. The goal has never changed. Analytics that gets used and creates actual intelligence. Clear, consistent numbers that people trust and tie to a decision.

1st
BI hire at LERMA/
Team of 3
BI engineers led
From scratch
Platform, metrics & CRM builds
B.S. Cognitive Science
UT Dallas
THE STAR SCHEMA · NAVIGATION

The whole career, modeled.

I'm a data modeler, so yes, I modeled my own homepage. The work itself sits in the middle, in the fact table. Five dimensions sit around it, and each one is a link into an area of work. Click a table to preview it, then dive in.

dim_toolingDIMENSION
  • PKtooling_id int
  • ··modeling · dbt · SQL · layered modeling · canonical metrics
  • ··warehouses_bi · BigQuery · Looker · Metabase · SSRS
  • ··engineering · Python · REST APIs · Cloud Run · Fivetran · Streamlit · Supabase
Broad modern stack. It serves the model, never the headline. Dive in →

HOW TO READ IT: the fact table holds only what can be measured: decisions enabled, adoption, trust. Everything else is a dimension that explains them.

SELECT * FROM dim_leadership;1 dimension · 3 proof points

Leadership: the founding story

FIRST BI HIRE → TEAM OF 3 → EXECUTIVE TRUST

Most portfolios show dashboards. Mine left this part out for a long time, which was a mistake. I didn't inherit a function at LERMA/. I founded it. I set its standards, I hired and coach the team that runs it, and I present its work to executives.

    Open dim_leadership ↓
    HOW I WORK

    From “can we see…” to a decision.

    01

    Ambiguous question

    Usually starts with “can we see…” or “why is this happening.”

    02

    Clarify decision + metric

    Figure out what decision this is actually tied to and what metric should drive it.

    03

    Model the data

    Lock in grain, definitions, and logic so it doesn't break later.

    04

    Build the system

    Pipelines, models, and whatever interface people will actually use.

    05

    Decision-ready output

    Something a team can look at and know what to do next.

    01 · dim_platform_builds

    Analytics platforms, built and made better.

    Sometimes the work starts from zero. Sometimes there's already a modern stack, but the next useful thing still needs to get built. I work across ingestion, models, metrics, and interfaces to ship things people actually use.

    Timesheet automation and workforce planning at LERMA/

    PYTHON · REST APIs · METABASE · GCP · DBT
    Problem
    Several people spent real time every week pulling timesheets by hand out of Webvantage/AQUA and pasting them into messy Excel. And the allocation plans built on top of that mess could only be adjusted once a year.
    System built
    Automated API ingestion of timesheets, feeding a live Metabase dashboard self-hosted on GCP with RBAC for the CFO and 1–5 department heads. I built most of it solo. A dev on my team finished the dbt modeling, which is how a team is supposed to work.
    Impact
    Agency-wide time logging for workforce planning. The CFO and I added quarterly metrics that line up logged hours against planned hours, which replaced a once-a-year spreadsheet exercise everybody dreaded.

    Paid media modeling: dbt + BigQuery

    DBT · BIGQUERY · FIVETRAN
    Problem
    Paid media data was split across platforms, and every platform had its own idea of what a conversion was and what spend meant. Ask three platforms, get three answers.
    System built
    dbt models in BigQuery with explicit day-platform-campaign grain, platform staging, and governed metric marts.
    Impact
    Executive reporting got standardized, and adding a new channel stopped breaking everything downstream of it.
    View Data Model ↗

    ProductDW: a mini ELT system

    PYTHON · SQL · STREAMLIT
    Problem
    Raw GA4 event data is not analytics-ready. It just isn't. Product KPI reporting needed more than querying raw events and hoping for the best.
    System built
    Python-orchestrated BigQuery pipeline with staging, core, and mart tables plus a Streamlit consumption layer.
    Impact
    Turned raw events into reusable product analytics tables and a practical KPI interface.
    GitHub ↗

    API feed to BigQuery

    PYTHON · REST APIS · CLOUD RUN
    Problem
    Event data needed to get into the warehouse reliably. Manual exports are how you end up with duplicate loads and numbers nobody trusts.
    System built
    Python ingestion job deployed on Cloud Run with API auth, pagination, schema mapping, and BigQuery watermarking.
    Impact
    A scheduled, low-maintenance pipeline feeding the warehouse for downstream reporting.
    02 · dim_leadership

    First BI hire. Then a team. Then a function.

    Player-coach is not a slogan for me, and it is not a line I put on a resume to sound balanced. I set the modeling standards. I review the work. I coach the engineers. And I still take on builds myself. The proof is not the title. The proof is that executives ask for the team's work by name. The CEO at LERMA/ asked me to personally present how the team built a client data warehouse from scratch so quickly. The CDO called our BI platform recommendation an “Amazing job.”

    Founding, not inheriting

    I was the first BI hire at LERMA/. The partner model, the governance, the single source of truth, the trust from executives. None of it existed when I got there. We built it from zero, in that order.

    Team of 3 BI engineers

    Hiring, standards, and review habits that let a small team ship platform work, including finishing each other's builds, like the dbt modeling on the workforce dashboard.

    Visible to the top

    I work directly with the CFO on workforce metrics and the CGO on the growth pipeline, and I present the platform work to the CEO and client executives myself. No game of telephone.

    “The founding story is my favorite part of the work: done the hard way, from scratch.”
    03 · dim_governance

    Governed metrics are the trust layer.

    If the model is messy, everything downstream is harder than it should be.

    I keep models boring on purpose. Clear grain, consistent naming, and logic defined once instead of copied into twelve places. Definitions belong in governed models, not scattered across dashboard calculations where nobody can find them. Because the minute two dashboards show two different numbers, the meeting stops being about the business and starts being about whose number is right. Nobody wins that meeting.

    Raw→ Staging→ Intermediate→ Mart→ BI

    BeforeExcel reporting: manual refreshes, version drift, hidden logic.

    AfterAutomated BI: scheduled pipelines, governed metrics, shared definitions.

    BeforeSSRS extracts: static outputs that need an analyst to touch them.

    AfterLooker workflows: reusable models, self-serve paths, decision dashboards.

    BeforeSiloed metrics: teams debate definitions instead of outcomes.

    AfterUnified layer: shared marts and canonical measures across teams.

    04 · dim_partnership

    Dashboards are products. Stakeholders are customers.

    I learned this one the hard way. Early in my career I built dashboards I was proud of that nobody opened. A dashboard is not useful just because it exists. If people don't use it in their actual workflow, it's dead, no matter how clean the SQL is. So now I think about whether people will actually use it before I build it. And sometimes the right product isn't a dashboard at all.

    Lite CRM for the CGO at LERMA/

    SUPABASE POSTGRES · AUTHENTIK · OUTLOOK SYNC
    Problem
    The CGO's leads and growth pipeline lived in an Excel doc. No shared view of 50+ historical and current leads and opportunities, and nothing he could check on the go.
    System built
    A lightweight browser CRM on Supabase Postgres with auth via Authentik, Outlook email sync, and a mobile-friendly view for leads and notes on the go. Built solo.
    Impact
    The CGO uses it regularly for growth updates and pipeline management. A spreadsheet nobody wanted to open became a product somebody uses every day.

    What adoption actually looks like

    SIGNALS FROM SHIPPED WORK
    Adoption signal
    Increased dashboard usage by focusing on clearer structure and aligning views to real decisions.
    Workflow fit
    Replaced manual Excel reporting with automated dashboards so teams didn't have to rebuild the same numbers every month.
    Trust layer
    Reduced back-and-forth questions by standardizing metric definitions upstream.
    01

    Awareness

    Teams know the source exists.

    02

    Usage

    It fits recurring workflows.

    03

    Trust

    Definitions are stable and explainable.

    04

    Decision

    The output changes what people do.

    THE ADOPTION CHECKLIST: what has to be true before a dashboard becomes part of the way a team works.

    05 · dim_tooling

    A broad modern stack, kept in its place.

    Tools are the part of this page I care about least, which is exactly why they live down here in a dimension instead of up in the fact table. The stack is broad because the work demands it: ingestion, modeling, warehousing, BI, and the occasional product build. It's here as evidence, not as the headline.

    Modeling

    dbtSQLLayered modelingCanonical metrics

    Warehouses + BI

    BigQueryLookerMetabaseSSRSExcelSnowflakePower BI

    Engineering

    PythonREST APIsCloud RunStreamlitFivetran

    Platforms shipped on

    GCPSupabase PostgresAuthentikGitHub

    Before LERMA/, I did BI and analytics at Everbridge, Old Republic Title, Central Insurance, and DFW Airport. Plus math tutoring, which is where the “explain it until it lands” habit started.

    PRINCIPLES

    How the work holds together.

    Start with the decision

    If it's not tied to a decision, it's probably not the right problem yet.

    Make grain explicit

    Every model should say what one row means. This prevents hidden joins and metric drift.

    Centralize metric logic

    Definitions belong in governed models, not scattered across dashboard calculations.

    Design for adoption

    A dashboard is successful when stakeholders return to it during real operating rhythms.

    Reduce analyst drag

    Good systems remove repeated pulls, one-off reconciliations, and preventable explanations.

    Keep builders close to users

    The person modeling the data should understand the workflow the dashboard supports.

    ↑ Top