Nervescape

Data & intelligence

Data science

Forecasting, experimentation, model evaluation and deployment.

Capability in business terms

Why this capability matters

A more complex model earns its place only when it changes a decision. Most projects need a baseline and a business question, not an algorithm.

What we typically scope

Forecasting, experimentation, segmentation, model evaluation, deployment and monitoring.

How we approach the work

Start with a measurable question and a naive baseline. Ship the simplest thing that beats it.

Work we take on

Example engagement — illustrative. Typical scopes of work, not named client case studies.

Demand forecasting that planners will actually use

Why it happens
Planners override the current forecast so often that the forecast has become decorative.
How we work it
Measure current accuracy against a naive seasonal baseline, interview planners on what they know that the model doesn't, add those features, expose forecast intervals rather than point estimates, track override rate as a success metric.
What you end up with
An accuracy comparison against baseline, a forecast with stated uncertainty, and override rate instrumented as the measure of trust.

Churn scores nobody acts on

Why it happens
A churn model produces scores weekly; the retention team has no defined action for any score band.
How we work it
Work back from available interventions, define score bands against those, estimate intervention cost and expected saved value per band, design a holdout so the effect can be measured.
What you end up with
Score bands tied to specific actions, a holdout group, and a measurable answer to whether the programme pays for itself.

An experiment that has to survive scrutiny

Why it happens
A pricing or product change is proposed and the argument for it is anecdotal.
How we work it
Define the primary metric and minimum detectable effect up front, calculate required sample and duration, pre-register guardrail metrics, agree the decision rule before seeing data.
What you end up with
A pre-registered test design, an agreed decision rule, and a result nobody can relitigate after the fact.

A model that has quietly stopped working

Why it happens
A model deployed eighteen months ago has never been re-evaluated and the world has moved.
How we work it
Reconstruct a recent evaluation set, measure current performance against the original and against a fresh baseline, test for feature and population drift, quantify the cost of its errors.
What you end up with
A current performance figure, a drift assessment, and a retrain, replace or retire recommendation with the reasoning written down.

Segmentation that survives contact with marketing

Why it happens
Customer segments were defined by workshop and are now used for targeting with unclear results.
How we work it
Cluster on behaviour rather than demographics, test each segment's predictive value against the outcome that matters, check each is reachable through an actual channel, size them.
What you end up with
Segments with measured predictive lift, each mapped to a reachable audience, and the ones that failed both tests discarded.

Related business solutions

Management visibilityReports take too long to prepare, and exceptions stay hidden…Workflow automationTeams chase approvals and repeat the same manual steps acros…ERP and system integrationERP, CRM and spreadsheets disagree, so finance and operation…

Ways to engage

By the hourA specialist, for the time you need.By the sprintA defined outcome. A defined finish.By the squadA team you can get to know.Managed operationsA system with someone accountable.

How we approach the work

Understand the estate

Map dependencies, constraints, decision owners and the current operating model.

Make the decision explicit

Record options, risks, costs and acceptance criteria before implementation.

Make the handover usable

Document changes, operating instructions and the remaining risks.

Delivery is subject to specialist fit and availability.

Start with clarity.

What would make your business easier to run?

Tell us the operational challenge. We will discuss whether a short conversation, a scoped assessment or a delivery engagement is the right next step.

Discuss your business challenge