Finding the signal

The analysis

We compared eighteen months of behavioural, billing, and support data across churned and retained accounts, controlling for plan, tenure, and industry to isolate genuine predictors from coincidence.

The signal discovered

Onboarding quality was a far stronger predictor of churn than price. Accounts that reached their first meaningful outcome within two weeks retained at more than double the rate.

The business impact

Retention initiatives were redesigned around onboarding and time-to-value rather than discounting. Annual churn fell materially within two quarters.

The analysis

Using geo holdout experiments and a marketing-mix model, we measured the incremental contribution of each channel rather than the conversions it claimed via last-click attribution.

The signal discovered

The best-looking channel was largely capturing demand that already existed. Two undervalued channels carried most of the genuine incremental lift.

The business impact

Budget was reallocated toward incremental channels. The same spend produced significantly more net-new revenue.

The analysis

We linked feature-level usage to retention and expansion using matched cohorts, separating features that correlate with good customers from features that cause them to stay.

The signal discovered

A single underused workflow feature was the strongest causal driver of long-term retention,  far ahead of the most-requested items.

The business impact

The roadmap was reordered around adoption of that workflow. Activation of the feature became a north-star metric.

The analysis

We modelled price elasticity across segments using historical price variation and a controlled rollout, estimating demand response with explicit confidence ranges.

The signal discovered

Demand was inelastic for established segments but highly sensitive among new, price-led customers — two groups that had been treated identically.

The business impact

A segmented pricing approach raised prices where demand was inelastic and protected acquisition elsewhere.

The analysis

We applied survival analysis to HR, engagement, and workload data to find which factors predicted who would leave, controlling for role and tenure.

The signal discovered

Manager quality and early-tenure workload predicted attrition far more than compensation. The first ninety days were decisive.

The business impact

The company invested in manager enablement and early-tenure workload balancing rather than across-the-board raises.

The analysis

We rebuilt the forecast as a time-series model with seasonality, promotions, and external regressors, and quantified uncertainty rather than producing a single number.

The signal discovered

Most error came from a handful of predictable promotional and seasonal patterns the existing process ignored.

The business impact

Inventory was planned against probabilistic forecasts. Stockouts and excess inventory both fell sharply.

The analysis

We combined funnel analytics with controlled experiments to isolate the factors that causally moved completion rather than those that merely correlated with it.

The signal discovered

A mid-funnel verification step, not the landing page, was the true constraint. Trust signals at that step mattered more than headline copy.

The business impact

The team redesigned the verification step. Conversion improved without additional traffic spend.

The analysis

We modelled lifetime value and identified the early, observable traits that distinguished the highest-value accounts within their first quarter.

The signal discovered

A specific combination of company size and early multi-team usage predicted high lifetime value long before revenue confirmed it.

The business impact

Targeting and sales prioritisation shifted toward those traits, raising the quality of the acquired base.