Supplier Performance Scorecard

SCMS Global Health Supply Chain (2006 to 2015)

At a glance

I scored the ten highest-volume suppliers in the SCMS shipment records on delivery reliability, delay severity, and price against the category median.

Four vendors were cleared without any action. Four are currently in Monitor. Two have fallen below the review threshold.
PHARMACY DIRECT, although it has a perfect 100% on-time record, was categorized under Monitor instead of Approved because its prices are 110.2% above the median for its product categories.
CIPLA has the better on-time rate of the two review-flagged vendors, but longer delays drop it below Aurobindo on the composite.

Note : In 85% of these shipments, the recorded delivery date matches the scheduled date exactly, suggesting the actual date was filled in from the schedule rather than captured at delivery. On-time delivery carries 60% of the composite score, so the reliability ranking may partly reflect record-keeping rather than performance. I've set out what this means in the limitations.

The scorecard

Each row is one vendor ranked by composite score.
The composite weights on-time delivery at 60%, delay severity at 25%, and price against the category median at 15%.
A threshold above 0.90 clears without action. Below 0.75 triggers a formal review. Everything in between goes on quarterly monitoring.

The four approved vendors are highly reliable and priced at or below the category median. The four vendors (0.75 < x < 0.90) flagged for quarterly review fall there for one of two reasons.

Findings

1. Delivery reliability alone would have ranked these vendors incorrectly

PHARMACY DIRECT completed 326 shipments on time every time, with no delays, but still appears in Monitor. Its unit prices are 110.2% higher than the median for its product categories.
Meanwhile, HETERO LABS ranks highest with a 99.3% on-time rate and prices 8.7% below the category median.
CIPLA is on schedule 86.9% of the time compared to Aurobindo's 85.9%, yet it averages 5.4 days late against Aurobindo's 5.0. So, while CIPLA misses slightly less frequently, its delays are longer.
Seven of ten vendors sit above 98% on time, so reliability alone offers limited distinction across most of the table.

2. Two vendors fall below the review threshold

Aurobindo scores 0.676, and CIPLA scores 0.668, both comfortably below the 0.75 threshold.
Both vendors have on-time rates below 87%, with average delays around five days. Price is not a concern for either, as both are priced below the category median.
Aurobindo shipped 668 consignments, ranking third in the set, indicating it is not a marginal vendor with a small footprint.

3. The Monitor band holds two different problems

Four vendors have scores between 0.75 and 0.90, and the same status label applies to opposite causes.
PHARMACY DIRECT, ABBVIE, and S. BUYS all deliver reliably, with on-time rates above 98%. However, their prices range from 94.7% to 149.6% above the category median.
Orgenics is positioned roughly at the median price, at 4.0% above, with 87.0% on time and an average delay of 1.6 days.
These require different responses: one is about price negotiation, and the other concerns delivery. A composite score merges these distinctions, which is why the input columns remain visible on the scorecard.

Where each vendor sits on delivery and price

S. BUYS, PHARMACY DIRECT, and ABBVIE cluster high on both axes because they deliver reliably and charge well above the category median.
Orgenics sits alone on the left, priced near the median but delivering at 87%. Same status label, opposite causes.

Freight: Air vs Ocean

Ocean

The median cost is $1.68 per kg.
The median lead time is 167 days.
The dataset includes 371 shipments, so this side is based on a much smaller sample.

Air

The median cost is $10.02 per kg.
The median lead time is 99 days.
With 6,113 shipments in the dataset, air transport is the default mode, not the exception.

The tradeoff

Air transportation costs an additional $8.34 per kilogram and reduces delivery time by 68 days, with a speed value of $0.12 per kilogram per day.
The median shipment value is $71 per kg. With a 20% carrying rate, holding costs are about $0.04 per kg per day, making air charges about three times higher than delay costs.
Air only breaks even above $224 per kg, and just 8.4% of shipments reach that.
The rate cannot account for stockouts. These are antiretrovirals and test kits supplied to treatment programs, and that is where the remaining case for air has to come from.

Method

1. Scope

I filtered the SCMS Delivery History to the ten vendors with the highest shipment counts, which left 4,044 shipments to score. Vendors below that cut had too few shipments for an on-time rate to mean much.
Rows marked "SCMS from RDC" were excluded. Those are transfers out of a regional distribution center, so the delivery date reflects warehouse movement rather than vendor performance.
The freight comparison uses a different cut of the same file: all air and ocean shipments rather than the top ten vendors, because mode choice belongs to the lane, not the supplier.

2. On-time delivery and delay severity

On-time delivery is a binary flag: a delivered date on or before the scheduled date counts as on time; anything later counts as a miss. It measures how often a vendor misses, not by how much.
Delay severity is the average days late across a vendor's shipments.
I kept them as separate inputs rather than combining them because they disagree. CIPLA misses less often than Aurobindo but averages longer delays, and a single reliability number would hide that.

3. Price position

Unit price alone isn't comparable across this catalog, since an antiretroviral and a test kit don't share a price scale. Instead, I compared each line item against the median unit price for its own Molecule/Test Type category.
That gives a percentage above or below the category median, so a vendor selling only expensive products is not penalized for it. The scored shipments span 45 categories.
I dropped 39 rows with a zero unit price, leaving 4,005 priced shipments because these are donated or bundled lines where price carries no information.

4. Composite Score

Each measure is normalized to a 0–1 scale, then weighted: 60% on-time delivery, 25% delay severity, 15% price position.
Reliability carries the most weight because a missed delivery in a health supply chain has consequences that a price premium does not.
Severity sits second as the check on frequency. Price is smallest because it is the easiest of the three to renegotiate.
The weighting is a judgment, not a derived optimum. Different weights would reorder the middle of the table, which is why the three input columns stay visible on the scorecard rather than collapsing into the score.

5. Thresholds and Status

Scores are sorted into three bands: 0.90 and above is Approved, 0.75 to 0.89 is Monitor, and below 0.75 is Review Required.
The 0.90 cutoff sits inside the largest natural break in the upper table, between 0.977 and 0.887. The 0.75 cutoff is a round number placed in a wide empty span between 0.824 and 0.676, so any value in that range would produce the same two groups.
Each band maps to a fixed action: no action, quarterly review, or formal supplier review. The point of a scorecard is to trigger something, so the status column carries an instruction rather than just a label.

6. Freight comparison

This uses median cost per kg and median lead time, not averages. Air's mean cost per kg is $33.52 against a median of $10.02, so a handful of emergency shipments at extreme rates would have made air look roughly three times worse than it routinely is.
I excluded truck and air charter because they aren't substitutable for air and ocean.
The 20% annual carrying rate in the tradeoff is my assumption, not a figure from the data.
At 15%, the breakeven moves to $299 per kg, and at 25% it moves to $179, so the conclusion holds across that range.

Limitations

1. Delivery dates may be backfilled from schedules

In 85% of scored shipments, the delivered date matches the scheduled date, suggesting the date was copied from the schedule rather than captured at receipt. This falls under on-time delivery, which carries 60% of the composite.
The pattern is uneven. The four Approved vendors run 88% to 99% identical dates, with PHARMACY DIRECT at 100% across 326 shipments.
The two Review Required vendors sit lower, at 82.6% and 83.4%, so their poor scores are not a record-keeping effect.
The key point is that the top of this table provides weaker evidence than the bottom. A vendor flagged for review indicates a genuine concern, whereas a vendor marked 'Approved' might simply have well-organized paperwork.

2. Different weights would produce a different ranking

The 60/25/15 split is my judgment, not a derived optimum, because nothing in the data says reliability should be worth four times price.
Vendors with scores between 0.82 and 0.89 are close enough that slight reweighting could change their ranking.
In contrast, the Approved and Review Required groups are more stable.
This is why the three input columns stay on the scorecard. A reader who disagrees with the weighting can rank on the raw measures instead.

3. A high score here does not mean a good supplier

The composite includes delivery, delay, and price, but omits quality, fill rate, order accuracy, and responsiveness.
A vendor might score 0.99 on this scorecard despite delivering defective products on time and at a fair price.
Since the dataset concludes in 2015, all scores reflect past performance and do not represent each vendor's current position.

Workbook Download

Download all Project Files (.zip)

Raw SCMS dataset, working file, and chart exports

Download the workbook (.xlsx)

The live Excel file with all formulas

V.S. Rishi · September 2026 · Source: SCMS Delivery History (USAID), via Kaggle · Built in Excel.

Google Sites

Report abuse