profile

Miraki23 LLP

60% Override Rate. Nobody at Board Level Knew. You Didn't Have a Technology Problem.


The Architecture Brief — Life Sciences Edition | Miraki23 LLP
Life Sciences Edition
The Architecture Brief
July 28, 2026
 
Editorial Statement

Somewhere in a pharmaceutical company's demand planning office, a spreadsheet is open. It was built in 2019 by a planner who has since been promoted twice. Nobody fully remembers the logic behind every formula anymore, but everyone trusts it more than the ₹4 crore forecasting platform sitting one tab over — because the platform's output, six times out of ten, gets overridden before it reaches the S&OP meeting.

Nobody at Board level knows this is happening. The licence renewal gets approved every year without a second look, because on paper, the company has "invested in AI-driven demand planning." The line item exists. The capability does not. This is not a story about bad software. SAP IBP works. Oracle ASCP works. Kinaxis RapidResponse works. This is a story about what happens when an organisation buys a demand-sensing platform and skips the much harder work of designing the capability around it — the governance, the trust calibration between human judgement and model output, the change management that convinces a planner with eleven years of pattern recognition that the machine is worth listening to.

This week we open Layer 7 of the Pharmaceutical Digital Capability Framework — Enterprise Data & Analytics — and connect it directly to Layer 3's supply chain planning domain, because the two cannot be separated. A forecasting model without governed data underneath it is not intelligence. It is an expensive guess with better graphics.

— Editorial Team  ·  Miraki23 LLP
Story of the Week
Life Sciences · Supply Chain Intelligence · Demand Sensing

Your SAP IBP Licence Cost ₹4 Crore. Your Planners Still Trust Excel.

Your demand planners are overriding 60% of system outputs with Excel models built in 2019. You did not have a technology problem. You had a capability adoption problem — and nobody at Board level knew.

A 60% override rate is not a rounding error. It is a verdict. When six out of every ten forecasts generated by a multi-crore platform are quietly replaced before they reach a decision-maker, the organisation has effectively built a very expensive validation exercise for a spreadsheet. The platform is not driving decisions. It is being consulted, politely ignored, and paid for regardless — an arrangement that survives budget review after budget review because nobody has ever measured override rate as a KPI, let alone reported it upward.

Supply chain intelligence in pharma requires more than an S&OP platform. It requires a demand-sensing capability that ingests real-world prescription data, channel sell-through, patient adherence signals, and epidemiological trends — and processes them through a model architecture the organisation actually trusts. SAP IBP, Oracle ASCP, and Kinaxis RapidResponse all support this kind of ingestion. The technology is not the constraint. The constraint is that most implementations stop at "the platform is live" and never reach "the planners believe the platform's output enough to stop building parallel Excel models as insurance."

Why a ₹4 Crore Platform Loses to a 2019 Spreadsheet
No Trust Calibration Framework
Nobody defined which forecasts the model is reliable for and which require human override by design. Without that boundary, every planner decides individually — and defaults to distrust.
Ungoverned Master Data Feeding the Model
If product master data, customer hierarchies, and channel data are inconsistent, the model's early outputs are wrong often enough that planners stop trusting it — permanently, even after data quality improves.
Override Rate Never Measured
Most organisations track forecast accuracy. Almost none track how often the forecast was overridden before it was ever tested against actuals. Without this metric, the adoption failure is invisible at Board level.
Implementation Without Capability Design
The go-live was treated as the finish line. Model governance, planner training on interpretation (not just navigation), and a change management programme addressing the eleven years of pattern recognition being displaced — none of that shipped with the platform.
The Evidence — When Demand Sensing Is Done Right

Deloitte's 2024 Life Sciences Supply Chain Survey reports that pharma companies with AI-augmented demand sensing achieve 25–35% forecast accuracy improvement and 18% reduction in write-off costs — but only 28% have deployed beyond pilot stage.

— Deloitte Life Sciences & Healthcare, 2024 Supply Chain Report

25–35%
Forecast Accuracy Gain — When Deployed Properly
18%
Reduction in Write-Off Costs
28%
Have Moved Beyond Pilot Stage

Read those three numbers together and the picture sharpens. The upside is real and substantial — a quarter to a third improvement in forecast accuracy is the difference between a stockout that costs a hospital account and a smooth fill rate that protects it. But 72% of pharma companies with these platforms are still stuck in pilot purgatory: technology deployed, capability never built, value never realised. The gap between the 28% who broke through and the 72% who did not is rarely the software vendor. It is whether the organisation treated demand sensing as a capability to be designed — data governance, model governance, human-AI decision integration — or as a system to be switched on.

Technology adoption without capability design is investment without return. The fix is not a bigger platform. It is a governance framework that defines when the model leads, when the planner leads, and how the organisation measures — and reports upward — which one is actually happening in practice.

S&OP Maturity Benchmark — Complimentary

Let us benchmark your S&OP maturity — including the override rate nobody has measured yet.

Benchmark Your S&OP Maturity →
Pharma Digital Capability Framework
Pharmaceutical Digital Capability Framework · Layer 3 + Layer 7
Supply Chain Planning + Enterprise Data & Analytics
9 Layers  ·  36 Domains  ·  ~200 Capabilities  ·  Sandeep Bansal, 2026

Data is the foundational currency of pharmaceutical digital transformation. Without robust data governance, high-quality master data, and enterprise-grade analytics platforms, all other capability layers are constrained in their ability to deliver intelligent, data-driven operations. Layer 7 is the foundation on which AI, machine learning, and advanced analytics depend — and it is the layer this week's demand-sensing failure was actually missing, disguised as a supply chain problem.

Layer 3 · Manufacturing & Supply Operations — The Relevant Domain
Supply Chain Planning ★ This Week's Focus

Demand forecasting, inventory optimization, integrated S&OP, supplier collaboration, capacity planning. This is where SAP IBP, Oracle ASCP, and Kinaxis RapidResponse sit — but the domain only functions when Layer 7's data governance underneath it is sound.

Layer 7 · Enterprise Data & Analytics — Six Domains
Data Governance

Data quality management, metadata management, data lineage, stewardship programs, regulatory data compliance — the layer that determines whether a demand model's inputs can be trusted before the model even runs.

Master Data Management

HCP/HCO master data, product master data, customer master data, reference data management. Inconsistent product or customer hierarchies are the single most common cause of early forecast errors that kill planner trust permanently.

AI & Machine Learning Platforms

Model development environments, MLOps infrastructure, model governance, AI ethics and bias monitoring — the governance layer that defines when the model leads and when the human overrides, by design rather than by default.

Predictive Analytics

Clinical outcome prediction, supply demand forecasting, commercial lead scoring, safety signal detection — this is where real-world prescription data, channel sell-through, and epidemiological trends get converted into a demand signal.

Enterprise Reporting & BI

Executive dashboards, operational reporting, self-service analytics, embedded analytics, data visualization — including, critically, the override rate dashboard most organisations have never built.

Real-World Data & Evidence

RWD integration, real-world evidence generation, patient registry analytics, outcomes research platforms — the patient adherence and epidemiological signal layer that separates true demand sensing from historical trend extrapolation.

Map Your Layer 3 + Layer 7 Maturity →
Miraki23 Perspective

“A ₹4 crore platform that gets overridden 60% of the time is not underperforming technology. It is a Board that never asked the one question that mattered: not 'did we buy the system,' but 'do our people actually trust what it tells them.'”

Aparna Bansal  ·  Founder, Miraki23 LLP
Next Week

Digital unreadiness has a patent-cliff equivalent in regulatory time.

Subscribe to the Newsletter

The Architecture Brief — Every Week

Pharma digital architecture, GxP compliance, AI governance, higher education transformation, GCC strategy, and enterprise advisory — weekly, for leaders who build, not just manage.

Subscribe Free →
Share on LinkedIn

Know a Supply Chain or Demand Planning Leader Who Needs This?

Share this issue with someone whose S&OP meeting still runs on a spreadsheet built years ago, or a CIO whose Board has never asked what the platform's override rate actually is.

Share on LinkedIn →
Topics This Week
#SupplyChain #DemandPlanning #SAP #PharmaCIO #InventoryOptimisation #DemandSensing #S&OP #KinaxisRapidResponse #OracleASCP #PharmaDigitalTransformation #LifeSciences #AIGovernance #Miraki23

Miraki23 LLP

We help organizations move from fragmented initiatives to structured transformation. At Miraki23, we integrate strategy, digital ecosystems, and performance architecture to deliver measurable business impact.

Share this page