Dependable data, clearly explained.

Data engineering & analytics

Flores DataCore

Data engineering & analytics consultancy

Dependable data, clearly explained.

We design, build, and document the data systems behind your reports and models, from the source database to the dashboard. Every step is tested, owned, and written down in plain English.

Take the 8-question assessment How an engagement runs

Independent Tool-agnostic Founded in 2025

  1. Sources: app database and CRM
  2. Ingestion: scheduled loader
  3. Staging: orders and customers
  4. Models: fct_orders and dim_customers
  5. Mart: finance
  6. Uses: revenue dashboard and demand forecast

Trace a number Hover, tap, or tab to any step to light up everything it depends on and everything that depends on it.

Fig. 1Where one revenue number comes from, in six layers.

A1The problems

Most data problems are trust problems.

The symptoms show up in meetings, not in the database. These are the ones we hear about most, and the fix is rarely another tool.

Analyst reviewing pie and bar charts on a desktop monitor
Fig. 2A number on a screen is only as good as the path behind it.

Two reports, two answers.

Finance and sales both report revenue, and the totals differ. Each team applied its own rules for refunds, currencies, and dates. Nobody is wrong, and nobody can say which number to use.

Pipelines that fail quietly.

A nightly load breaks on a schema change. Dashboards keep showing stale data,2 and the first alert is a confused question from a manager.

Spreadsheets doing a database's job.

A critical workbook pulls from three exports, and one person knows how to refresh it. When that person is out, reporting stops.

Dashboards nobody opens.

There are more dashboards than readers. Several show the same metric with different filters, and none say where the number came from.

AI plans waiting on data.

A forecasting or AI project stalls because history is incomplete, labels are missing,3 or no one can say who may use which data.

Costs that creep.

Warehouse and tool bills grow every quarter, but no one can connect the spend to the reports and decisions it supports.

  1. 1

    Lineage is the recorded path a dataset takes from its sources, through each transformation, to the reports and models that use it. Back

  2. 2

    Freshness compares when a table was last updated with how current its users need it to be. A daily sales table that is 30 hours old is stale; a monthly budget table is not. Back

  3. 3

    In machine learning, labels are the known outcomes a model learns from, such as whether a past customer canceled or which invoices were paid late. Back

A2Services

Six practices, one standard: tested, documented, owned.

  1. 01 Data strategy and architecture A target architecture and a sequenced plan you can fund.
  2. 02 Data engineering and pipelines Reliable, monitored loads from every system that matters.
  3. 03 Analytics engineering and modeling Business rules moved out of spreadsheets and into tested models.
  4. 04 BI and dashboards Fewer dashboards, built on definitions everyone agreed to.
  5. 05 Data quality and governance Checks, owners, and definitions people can actually find.
  6. 06 AI and machine learning data readiness History, labels, and access controls a model can rely on.
All six services, with signals and deliverables

A3Worked example

From messy to modeled, one step at a time.

Scroll, or pick a step, to run the four steps we apply to almost every source. The records are fictional. The problems are not.

Raw. Eight records as they arrived from two loads: one exact copy, one customer loaded twice, three date formats, mixed casing, a missing plan, and an amount that is not a number.

idtext customertext signup_datetext plantext amounttext
1042Harbor Supply2025-03-04Pro120.50
1042Harbor Supply2025-03-04Pro120.50
1043Maple & Co2025-03-06Pro89.00
1044CEDAR POINT8 Mar 2025basic45
1045Linden Works2025/03/09Basicn/a
1043maple & co03/07/2025PRO"89"
1046riverstone2025-03-11pro230
1047Oak Street Bakery2025-03-12NULL-15.00
  • Duplicate
  • Inconsistent format
  • Wrong type
  • Missing value
  • Fails a rule
  • Passed checks
Fig. 3raw.signups: eight fictional records from two loads.

A4Approach

Five phases. Each one ends with something you keep.

Person studying flow diagrams sketched on a whiteboard
Fig. 4Design starts on a whiteboard and ends in version control.
  1. 01Assess2–4 weeks

    Interviews, a source inventory, and a written findings report with priorities.

  2. 02Design2–4 weeks

    Target architecture, data models, and a plan sequenced by value.

  3. 03Build4–12 weeks

    Pipelines, models, tests, and dashboards, delivered in working increments.

  4. 04Enable2–4 weeks

    Documentation, runbooks, and working sessions with your team.

  5. 05OperateOngoing, optional

    Monitoring, fixes, and small changes on a monthly plan.

Read the full approach, with the lineage map
Printed charts held over a desk with a laptop, notebook, and calculator
Fig. 5Charts are easier to argue with once they are on the table.

A5Self-assessment

Where does your data practice stand?

Eight questions about how your data moves, how it is checked, and who uses it. You get a stage from 1 to 4, a chart drawn from your own answers, and a suggested first engagement. It takes about four minutes, and your answers stay in your browser.

Start the assessment

B1Insights

From the insights desk.

Practical notes on building data systems people trust. Written for the people who pay for them and the people who run them.

City lights spreading across a valley at dusk

Correspondence

Tell us what the numbers need to do.

Describe your sources, your team, and the decisions you want the data to support. We reply by email with questions, or with a short plan for an Assess phase.

contact@floresdatacore.com
Use the contact form

Fig. 6 City lights at dusk, set as a halftone screen. Move across it to see the photograph.