Portfolio laboratory
Exploratory analysis, prediction and operational modelling translated into commercial actions.
All datasets and commercial scenarios on this page are synthetic and created for portfolio demonstration.Each case includes the business question, source structure, working method, selected code or decision logic, measured result, recommendation and limitations.
Who is valuable, becoming inactive and still economically sensible to contact?
A time-based CLV validation, contribution forecast, acquisition-economics view and treatment queue with explicit assumptions and value-capture diagnostics.
367-line Python workflow · controls · holdout test · CLV scoring · intervention economics
Which customers need intervention before the renewal date?
A practical classification workflow that balances predictive performance with the commercial need to explain who is at risk and why.
How much stock protects service without locking up cash?
A six-SKU forecasting and safety-stock model with a transparent holdout test and a simulated operational impact check.
Who is valuable, becoming inactive and still economically sensible to contact?
A time-based CLV validation, contribution forecast, acquisition-economics view and treatment queue with explicit assumptions and value-capture diagnostics.
Business context
A customer team wants to retain valuable customers but has limited contact capacity. The analysis must distinguish observed contribution, forecast value, inactivity and the economics of treatment.
Questions tested
Working method
Validated 16,111 transactions for uniqueness, dates, margin logic and missing values.
Built customer-level frequency, recency, tenure, margin, support and acquisition-cost features using only pre-holdout data.
Predicted six-month contribution on a time holdout and compared error with a median-value baseline.
Converted the model to discounted 12-month CLV, value bands and a transparent inactivity rule.
Subtracted treatment cost and applied an expected reactivation rate before placing a customer in the intervention queue.
Dataset
transaction_id / customer_idOrder and customer identityorder_date / join_dateObservation and customer-age timelinesegment / acquisition_channelCommercial cohort contextrevenue_gbpObserved order revenuegross_margin_gbpContribution basis for valueacquisition_cost_gbpOne-time channel costsupport_contacts_90dRecent service-friction signal| Transaction | Customer | Date | Segment | Channel | Revenue | Margin |
|---|---|---|---|---|---|---|
| CT-0000001 | CLV-00001 | 17 Jul 2025 | Consumer | Paid Search | £63.96 | £31.92 |
| CT-0000002 | CLV-00001 | 05 Feb 2026 | Consumer | Paid Search | £10.00 | £4.86 |
| CT-0000003 | CLV-00001 | 17 Mar 2026 | Consumer | Paid Search | £40.66 | £21.93 |
Preview shows three records; the complete synthetic dataset is available above.
pipeline = Pipeline([
("features", ColumnTransformer([
("numeric", StandardScaler(), numeric_features),
("categorical", OneHotEncoder(handle_unknown="ignore"), categorical_features),
])),
("model", GradientBoostingRegressor(
n_estimators=180, learning_rate=.035,
max_depth=2, min_samples_leaf=16,
loss="huber", random_state=1210,
)),
])
eligible = (
(customer.discounted_12m_clv_gbp >= 120)
& (customer.recency_days >= 120)
& (customer.expected_incremental_margin_gbp > 0)
)Result
Recommendation
Pilot the ranked 152-customer queue with a holdout group, record contact and incremental margin outcomes, and review assumptions by segment before increasing treatment capacity.
Limits & next evidence
The synthetic model predicts value, not causal uplift. A live intervention requires consent controls, calibrated treatment-uplift evidence, fairness review, drift monitoring and cost-to-serve.
Which customers need intervention before the renewal date?
A practical classification workflow that balances predictive performance with the commercial need to explain who is at risk and why.
Business context
A subscription business has limited retention capacity. The goal is not merely to predict churn, but to rank a practical intervention queue and explain the operational signals behind it.
Questions tested
Working method
Validated identifiers, missingness and the 15.2% target prevalence.
Used a stratified 75/25 split and fitted a preprocessing/model pipeline to prevent leakage.
Evaluated both discrimination (ROC-AUC) and threshold accuracy.
Scored the full population and converted probability into operational risk bands.
Dataset
contract_typeMonthly, annual or two-yeartenure_monthsCustomer agesupport_tickets_90dRecent service frictionusage_scoreProduct engagement, 0–100payment_failures_12mRecent payment instabilitychurnedObserved binary outcome| Customer | Region | Contract | Tenure | Tickets | Usage | Churned |
|---|---|---|---|---|---|---|
| CU-00001 | North | Monthly | 28 | 2 | 53 | No |
| CU-00002 | South | Monthly | 72 | 1 | 23 | No |
| CU-00003 | London | Two-year | 21 | 0 | 79 | No |
Preview shows three records; the complete synthetic dataset is available above.
model = Pipeline([
("prep", ColumnTransformer([
("num", StandardScaler(), numeric),
("cat", OneHotEncoder(handle_unknown="ignore"), categorical),
])),
("model", RandomForestClassifier(
n_estimators=220, min_samples_leaf=5, random_state=1210
)),
])
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=.25, stratify=y, random_state=1210
)
model.fit(X_train, y_train)
risk_probability = model.predict_proba(X_test)[:, 1]Result
Recommendation
Contact the highest-risk decile first, pair the score with recent support and usage context, and test at least two retention treatments against a control group.
Limits & next evidence
The synthetic target was generated from known relationships and does not prove causal drivers. Production use would require fairness checks, calibration, drift monitoring and treatment-uplift measurement.
How much stock protects service without locking up cash?
A six-SKU forecasting and safety-stock model with a transparent holdout test and a simulated operational impact check.
Business context
A small distributor wants fewer stockouts without treating every SKU as equally predictable. The case joins demand forecasting with a transparent service-level reorder rule.
Questions tested
Working method
Created trend, day-of-week, month and promotion features for each SKU.
Held out the final 60 days, preserving time order.
Calculated reorder points as lead-time demand plus 1.65 standard deviations of safety stock.
Compared rolling stockout events against a naive replenishment rule.
Dataset
date / skuDaily product grainunits_soldObserved daily demandlead_time_daysSupplier replenishment delayunit_holding_cost_gbpIllustrative carrying costpromotion_flagKnown demand intervention| Date | SKU | Units | Lead time | Holding cost | Promotion |
|---|---|---|---|---|---|
| 01 Jan 2024 | SKU-A | 29 | 6 days | £6.20 | Yes |
| 02 Jan 2024 | SKU-A | 14 | 6 days | £6.20 | No |
| 03 Jan 2024 | SKU-A | 20 | 6 days | £6.20 | No |
Preview shows three records; the complete synthetic dataset is available above.
train, holdout = history.iloc[:-60], history.iloc[-60:]
model = LinearRegression().fit(train[features], train["units_sold"])
prediction = np.clip(model.predict(holdout[features]), 1, None)
recent = history.tail(90)["units_sold"]
expected_lead_demand = recent.mean() * lead_time
safety_stock = 1.65 * recent.std() * sqrt(lead_time)
reorder_point = round(expected_lead_demand + safety_stock)Result
Recommendation
Pilot the new rule on high-volume SKUs, track stockout rate and excess inventory together, and review service factors by item criticality rather than assuming one service level.
Limits & next evidence
The simulation excludes supplier minimums, shelf life, substitution, capacity constraints and forecast bias. It estimates policy direction, not a production purchase order.