4.1
Strategic procurement
RFP evaluation, tender coordination, PPRA compliance, supplier negotiation — vendor management across 35+ external manufacturers and distributors.
IqraTahir
Supply chain & procurement — Islamabad
02 / The work
Six thousand SKUs of personal care and cosmetics, across twenty-eight shops, under demand that refuses to sit still.
My work is deciding what gets bought, from whom, and at what risk — forecasts that are honest about uncertainty, negotiations that respect the vendor on the other side of the table, and cost trade-offs made out loud, from the warehouse floor to head office.
03 / The network
Thirty-five vendors in. Twenty-eight branches out.
One procurement desk between them — Punjab Cash & Carry HQ, Islamabad.
- Active SKUs routed
- 6,000+
- Brands carried
- 200+
04 / The practice
4.2
Predictive analytics
Demand forecasting, stochastic optimization, multi-criteria decision analysis, predictive modeling — the quantitative core of both the thesis and the day job.
4.3
Retail operations
Inventory control, capacity planning, assortment and shelf-velocity analysis — dead-stock discipline across 6,000 SKUs.
05 / The thesis
MS Logistics & Supply Chain Management · NUST · complete draft
Analyzing current procurement practices and optimizing cosmetics supply chain through machine learning-based demand forecasting.
A case study of a leading retail chain across Islamabad and Rawalpindi — the same category, and the same kind of shelf, she buys for daily.
5.1 / The practice as it is
Half the thesis is diagnosis, not modelling. Before anything can be optimised it has to be described: how orders actually get raised, what a reorder decision is really based on, where lead time is absorbed, and which of those steps runs on judgement rather than on a number.
Method
Process mapping of the current procurement cycle, against the demand history the same operation generates.
5.2 / The signal
Weekly demand for a single cosmetics line. What makes it hard is not the week-to-week noise — it is the promotional and festival spikes, which a reorder point cannot anticipate and a model can be taught to.
- Lagged demand
- Rolling means
- Seasonality & calendar
- Promotion periods
- Price movement
Illustrative figures — a diagram of the method.
Weekly demand
3,180
+420 units (+15.2%) vs Feb
Monthly
- JAN
- FEB
- MAR
- APR
- MAY
- JUN
- JUL
5.3 / The forecast
History fits, then the split, then the horizon — the forecast runs forward inside a widening interval, above the line where the shelf runs out before the next delivery lands. The interval is the honest part: a forecast that does not widen is not a forecast.
- Tree ensembles
- Random Forest · XGBoost · LightGBM
- Sequence models
- LSTM · RNN · Prophet
Illustrative figures — a diagram of the method.
Forecast · next 20 weeks
4,240
± 680 units at the far horizon
Weekly
- W32
- W36
- W40
- W44
- W48
- W52
Which family wins, and by how much, is held for the defence. What can be said now: the models are trained on the same category this site is about — cosmetics, bought for real shelves, in the city she works in.
06 / The category
One lipstick, taken apart.
Six components, four cities, one shelf date. The longest lead time — not the price — decides when buying has to begin.
- 6.1Pigment — mica, iron oxidesJaipur, India45 days
- 6.2Base — waxes, butters, oilsKarachi, blended from import30 days
- 6.3Mechanism — aluminium, ABSShenzhen, China60 days
- 6.4Bullet moulding & fillLahore21 days
- 6.5Folding cartonLahore14 days
- 6.6Label & foil stampKarachi10 days
Representative lead times for a mass-market lipstick sold in Pakistan — the decomposition a category manager carries in her head.
07 / The role
Cosmetics procurement, at retail scale.
- Active SKUs
- 6,000+
- Personal care & cosmetics brands
- 200+
- Retail branches supplied
- 28
- External vendors & manufacturers
- 35+
Dead stock down 10% — by analysing demand patterns and shelf-space velocity, and proactively discontinuing non-performing and high-expiry SKUs.
Before this: operations manager for a 70-bed residential complex in Multan — budgets, strategic supply procurement, a five-person team, utility and material waste down ~65%.
