SystemsSheet 08 of 08

Analytical dashboards your teams can trust

A dashboard is only as good as the numbers under it. We define each metric once, test it against source data, and publish it in the BI tool you already use. Then we add an assistant that answers follow-up questions from the same definitions.

Tools: Power BI, Tableau, Looker, Databricks AI/BI, Superset
Built before: Reporting for 80+ US healthcare clients
Report families: Eight, from member months to provider performance

Metrics defined once, served everywhere

Switch the environment to see the services it runs on. Play the walkthrough, or pick a step.

Deterministic codeAI stepHuman reviewYour existing systemAI assist, approved by a person

Metrics defined once, served everywhere

Curated data, one semantic layer with owners and tests, and every dashboard and assistant reading from it.

Power BI semantic models with row-level security, and Copilot in Power BI for questions.

See it run

Metrics from the semantic layer, and a plain-language question answered with the query shown.

Payer performanceExample data, not real membersIllustrative example

Member months128,420March 2026

PMPM$412.60+3.1% vs February

Open care gaps3,9125 measures

IBNR estimate$6.1Mfrom lag triangles

PMPM, last 12 months

Open care gaps by measure

Breast cancer screening1,120

Colorectal cancer screening980

Blood pressure control740

Diabetes, HbA1c control610

Cervical cancer screening462

Why did PMPM rise in March?

Inpatient paid amount rose 14% in March, from 38 admissions at two facilities. Membership was flat, so PMPM rose 3.1%. Summary reviewed by an analyst before it was sent.

select service_month, setting, sum(paid_amount) as paid
from semantic.claims_paid
where service_month in ('2026-02', '2026-03')
group by 1, 2;

Component view

The same system as an exploded 3D drawing. Each plate is one step, and each block is one component in one of four materials.

Drag to rotate. Select a layer to inspect it.

Parts list

Numbered bottom to top

Dashboards read from curated tables. Where the data is not curated yet, we build that first, as shown on sheets 05 and 06.
- Warehouse or lakehouse Your existing system
- Spreadsheets Your existing system
- SaaS tools Your existing system
Typical toolingWarehouse, lakehouse, spreadsheets, SaaS exports
A clean reporting model at the right grain, with security rules applied once rather than per dashboard.
- Curated tables Deterministic code
- Row-level security Deterministic code
Typical toolingdbt, SQL views, Delta tables, row-level security
Each metric, such as member months, PMPM or IBNR, is defined once and has an owner. Every dashboard and the assistant use the same definition.
- Semantic layer Deterministic code
- Metric owner Human review
Typical toolingdbt semantic layer, Power BI semantic models, LookML, Databricks metric views
Totals are reconciled against source systems on every refresh. If a number moves outside its expected range, the owner is alerted before anyone reads it.
- Source reconciliation Deterministic code
- Range alerts Deterministic code
Typical toolingdbt tests, SQL reconciliation checks, alerting
Views for each audience, from executive summaries to the operational lists people work from every day, with drill-down to the record.
- Executive view Deterministic code
- Operational views Deterministic code
- Drill-down Deterministic code
Typical toolingPower BI, Tableau, Looker, Databricks AI/BI, Apache Superset
People ask questions in plain language and get answers built on the semantic layer, with the query shown. Scheduled summaries explain what changed, and an analyst reviews them before they go out.
- Ask your data AI step
- Change summaries AI step
- Analyst review Human review
Typical toolingDatabricks Genie, Power BI Copilot, text-to-SQL agents over the semantic layer
Refreshes follow the data pipeline, reports arrive where people work, and usage shows which dashboards earn their place.
- Scheduled refresh Deterministic code
- Subscriptions Deterministic code
- Usage tracking Deterministic code
Typical toolingSchedulers, email, Teams and Slack delivery, usage logs

Deterministic codeAI stepHuman reviewYour existing system

Sheet08 of 08

Layers7

SystemAnalytical dashboards

Drawn byA. Adhikari

IssuedSeptember 2026

ScaleNot to scale

Report families we have built

Delivered for 80+ US healthcare clients, with data from payers including Aetna, Anthem and Blue Cross Blue Shield plans. The method carries to any industry: define, test, then visualise.

Member months and PMPM

The denominator behind nearly every per-member metric, derived from enrollment spans with retroactive changes handled.

IBNR reserves

Incurred-but-not-reported estimates built from claims-lag triangles for actuarial teams.

Care gaps

Quality-measure results turned into member lists for outreach programmes.

Leakage

Spend that left the contracted network or missed negotiated rates.

Provider performance

Cost and utilisation compared with peer benchmarks.

High-cost members

Spend and risk stratification that points case management to where it matters most.

What an engagement delivers

Metric catalogue
Every metric written down with its definition, grain, owner and source tables.
Reporting model and tests
The tables behind the dashboards, with reconciliation tests on every refresh.
Dashboards
Views for each audience, built in your BI tool and reviewed with the people who use them.
Assistant and handover
Plain-language questions over the same definitions, plus training and documentation for your team.

Where it runs

The drawing stays the same. The services change with the environment you already run, and the feasibility analysis picks the fit.

Environment

What we use

When it fits

Power BI and Microsoft Fabric

What we useSemantic models, row-level security, Copilot

When it fitsMicrosoft-centred teams.

Tableau

What we useTableau Cloud or Server, published data sources

When it fitsTeams with existing Tableau skills.

Looker

What we useLookML semantic layer

When it fitsGoogle Cloud and BigQuery teams.

Databricks AI/BI

What we useDashboards, Genie, metric views

When it fitsData already in Databricks.

Open source

What we useApache Superset, Metabase

When it fitsSelf-hosted, no per-user licences.

Feasibility first

Before anything is built, we check whether this system is worth building for you, and where it should run.

Which decisions the dashboards support, and who makes them.
Which metrics exist today, and where the numbers disagree.
Data sources, freshness needs and security rules.
BI tool, licensing, and where people already work.

What you receiveA metric catalogue for the first dashboards, a reporting model design, mock-ups reviewed with users, and a go or no-go recommendation.

Built before

Work delivered by Ashish Adhikari, who leads engineering at YoursSherpa.

Payer analytics and reporting suite for 80+ US healthcare clients: eight report families including member months, PMPM, IBNR, care gaps, leakage, provider performance and high-cost members.Healthcare analytics
HEDIS quality measures engine (BCS, CCS, COL) that feeds care-gap reporting.Healthcare quality measures

Start with a feasibility call

Tell us about one process or data problem. In the first call we will say which parts we would automate with code, which need an agent, and which we would leave alone.

Send a short note through the contact form and we will set up the call.

What helps us prepare

The process or system you have in mind, and who works on it today.
Where the data lives: cloud, platform and main tools.
Security or hosting rules we need to work within.