Transformation · Healthcare

Where the next collection centre goes, tested first on the centres already open.

A national diagnostics chain was entering a new state at speed. The model was first tested on the centres it already ran, exposing overlap and weak placements, and then used to find sites across underpenetrated districts.

Diagnostics chain

Labs and collection centres

Tier 4 to 6 markets

What the analysis found

Two observations from the client network, and the scope they came from.

~75%
Observed
Share of revenue produced by the top 30% of collection centres.
30%
Observed
Share of centres underperforming. Vintage does not matter.
48,598
Scope
Micro-markets scored, each 0.7 km², across 100+ signals.
5
Scope
Districts covered in a state the chain had not yet entered.

The case

01 · The retail pivot

A rollout measured in hundreds

The chain was moving into private retail collection centres in markets competitors had not reached.

02 · The scouting bottleneck

Expansion capped by executive calendars

Site selection ran on field scouting, with leadership travelling to each district to sign off in person.

03 · The reporting blindspot

Bad location, or bad execution

Consolidated reporting could not tell the two apart. The standard answer was usually more marketing.

The model was run backwards before it was run forwards

EPIC scored collection centres whose revenue performance was already known. The model separated strong sites from weak ones on data the business already held. Three findings shaped everything after.

01 · Revenue concentration

Three quarters of revenue from under a third of the network. A distribution this skewed points at placement.

02 · The vintage paradox

Almost no difference. The assumption is that a new centre needs time to find its feet. The data says a badly placed centre stays badly placed.

03 · Internal cannibalisation

Centres sharing a single 0.7 km² catchment.Two nodes splitting one patient pool, each looking weak in the MIS, while aggregate reporting hid the overlap.

What was scored

Each 0.7 km² hexagon carries over 100 signals.

Clinical points of interest

Clinics, doctors and hospitals across each catchment.

The closest direct predictor of sample volume.

Competitor footprints

Diagnostic labs, chains and hospital in-house facilities.

Establishes unsaturated volume, separate from total demand.

Micro-demographics

Ward-level population density and household income layers.

Matches catchments to high-value test panels.

Logistics and travel

Road networks and transit times to processing labs.

Ensures samples arrive within testing stability windows.

Three kinds of white space

Unserved markets differ in what they are worth and in how they should be entered.

Grade

Market profile

Strategic action

Priority Growth

Dense, affluent, clinic-rich catchments with zero chain presence.

Immediate high-confidence site acquisition

Clinical Corridors

High-volume hospital clusters with active sample flow already present.

Fast-track centres to capture immediate volume

Growth Fringes

Developing outer corridors where demand is still forming.

Low-cost early-mover entry before saturation

What each team received

One output per role, built around the decisions each one makes.

Leadership and board

Sign-off without the site visit

Data-backed virtual site reports replace travel to every district, and preserve the reasoning so location logic survives field team attrition.

Expansion teams

Ranked shortlists per catchment

Graded micro-markets within each lab catchment, with cannibalisation flagged before a lease is signed, and location scoring on prospective partners before onboarding.

Field operations

Spatial tools for the ground

Doctor outreach directed at dense patient pockets, sample-runner routes built around real catchments, and rep coverage rebalanced against sample demand.

How the workflow changed

Before

Field teams walked new states to find candidate sites

Leadership travelled to each district to approve in person

Centres opened inside each other catchments unnoticed

Underperformance was attributed to marketing

Site knowledge left when the field team did

After

Districts scored before anyone enters the state

Sign-off runs on virtual site reports

Cannibalisation flagged before a lease is signed

Location and execution separated in the diagnosis

Reasoning recorded and preserved through attrition

Where the return comes from

Three levers, modelled against the client district economics and agreed before work began.
These are projections, and they are stated as projections.

Status quo

Three field staff per district cluster scouting manually, plus leadership travel
to sign off.

50% leaner

With EPIC

Scouting sized against a shortlist instead of a district.

Status quo

Roughly 15% of new centres fail after launch, with signage, training and partner churn already sunk.

15% to 2.5%

With EPIC

Failure risk removed from the shortlist before capital is committed.

Status quo

Centres take six months or more to reach steady-state revenue because of weak initial placement.

3 months
faster

With EPIC

Stronger placement shortens the ramp to peak revenue.

Together these models to roughly four times return on the engagement, with payback inside the first corrected centre in each district.

It took us 12+ months to build this network, and with this intelligence
we could've built it in 2 months.

Expansion Leadership

The model was back-tested on their own network before it recommended anything. Showing a field team that the model identified centres they already knew were struggling earns more trust than an accuracy claim about markets they have never seen.

Start with the network you already have.

Share your existing locations and we will score them against our signals, before we recommend a single new site.

EPIC Intelligence · Client identity withheld