Operating Intelligence for Manufacturing Companies: OEE, Throughput & Beyond

Why Manufacturing Operations Stay Fragmented

A plant running at 62% OEE is losing roughly a third of its available productive capacity. The machine data says so clearly. But that 62% number rarely appears next to the COGS-per-unit figure it is directly inflating, or next to the on-time delivery rate it is directly suppressing. The data exists — in the MES, in the ERP, in the quality system — but it lives in three different places with three different owners.

This is the fragmentation problem that operating intelligence solves. It is not a technology problem. Modern manufacturers are not short on data-collection infrastructure. The gap is the operating layer that sits above the individual systems and connects production signals to business outcomes.

Without that layer, the COO is doing the connection manually. Every week, someone exports from the MES, exports from the ERP, pastes into a spreadsheet, and tries to reconcile numbers that were never designed to talk to each other. By the time the report is clean, it is already describing conditions that existed four days ago.

The manufacturers who close the gap gain a specific and measurable advantage: they catch problems while they are still cheap to fix. A scrap spike that appears in the quality system on Tuesday and gets reconciled into the weekly operations report on Friday is four days of compounding loss. The same signal surfaced in an operating intelligence view on Tuesday morning is an actionable problem on Tuesday morning.

The Core Manufacturing Metrics Framework

Every manufacturing operating intelligence system is built on the same set of foundational metrics. The frameworks vary — lean, theory of constraints, six sigma — but the underlying numbers are consistent. Understanding each one precisely, and understanding how they connect, is the prerequisite to building a useful operating view.

OEE: Overall Equipment Effectiveness

OEE is the single most widely used metric in manufacturing operations. It measures what percentage of planned production time is truly productive — producing good parts at the intended speed with no unplanned downtime. The formula is:

Each component reveals a different category of loss. Availability loss comes from unplanned downtime — equipment failures, changeovers that run long, or upstream starvation. Performance loss comes from running slower than the ideal cycle time — micro-stops, speed reductions, operator pacing. Quality loss comes from producing units that fail to spec and require rework or scrapping.

Industry benchmarks are well-established. World-class OEE is 85% or higher for discrete manufacturers. Most plants average 55–65%. Across nine primary discrete manufacturing sectors, the 2025 average sits at approximately 66.8%, with medical device manufacturers achieving the highest performance at 78.2% and trailer/RV manufacturers the lowest at 57.2%. Continuous-process industries — chemicals, food and beverage, refining — typically target 90% or above because scheduled downtime is structurally minimal.

A useful mental model: a plant at 65% OEE has effectively lost 35% of its planned capacity to some combination of unplanned stops, speed loss, and defects. Identifying which component is driving the gap tells you where to direct engineering and maintenance resources.

TEEP: Total Effective Equipment Performance

TEEP extends OEE by including scheduled downtime in the denominator. Where OEE asks "how productively did we run during our planned hours?", TEEP asks "how productively did we run against all available calendar time?"

A plant with 78% OEE running two eight-hour shifts five days a week has a Utilization of roughly 48% (80 hours planned out of 168 available per week). Its TEEP is approximately 37% — meaning it is using just over a third of its theoretical maximum capacity.

TEEP is not a daily operational metric. It is a strategic capacity metric. It answers the question: if demand increased 30%, could this plant absorb it by adding shifts, or would it require capital investment? A TEEP below 40% with growing demand is a strong signal that additional shifts are the better path before any capex discussion begins.

Throughput

Throughput is the count of good units produced per unit of time — usually per hour, per shift, or per day. The critical precision in throughput is that scrapped units and rework units do not count. A line producing 1,000 units per hour with a 5% scrap rate has actual throughput of 950 good units per hour, not 1,000.

Throughput is the most direct expression of a line's revenue-generating capacity. It connects directly to COGS per unit: if your fixed overhead for a line is $10,000 per shift and throughput drops from 2,000 units to 1,600 units, your overhead cost per unit rises from $5.00 to $6.25 — a 25% margin hit that never appears in any single system's dashboard without the cross-system connection.

First Pass Yield and Scrap Rate

First Pass Yield (FPY) is the percentage of units that exit a process meeting spec on the first attempt, with no rework required. Scrap rate is the complement of quality yield — units that cannot be reworked and are discarded entirely.

FPY is often more revealing than OEE's quality component alone because it tracks rework — units that eventually ship but consumed additional machine time and labor to get there. A line running at 95% throughput efficiency with 80% FPY is effectively running at 76% useful output. The rework hours come out of capacity that could have produced new units.

Mature, well-run plants target scrap rates below 2–3% in most discrete manufacturing contexts. In high-precision industries — aerospace, medical devices, semiconductors — scrap rates are measured in parts-per-million. Benchmarks vary significantly by process, so internal trend analysis matters more than cross-industry comparisons.

Cycle Time

Cycle time is the elapsed time to produce one unit from start to finish of a defined process step or the full production sequence. It has three distinct variants that manufacturing teams conflate at their peril:

• Ideal (theoretical) cycle time: The fastest a process can run under perfect conditions — a machine specification, not an operational target.
• Actual cycle time: What the process is currently averaging, including micro-stops, material delays, and operator variability.
• Takt time: The rate at which units must be produced to meet customer demand. Calculated as available production time divided by customer demand in units.

The relationship between these three numbers tells a precise story. When actual cycle time drifts above takt time, the line cannot meet demand without overtime or added capacity. When actual cycle time drops significantly below takt time, you are overproducing — building inventory that ties up working capital and increases carrying costs. Lean manufacturing uses takt time as the metronome that keeps production synchronized with demand rather than running at maximum mechanical speed.

Inventory Turns

Inventory turns measure how many times inventory cycles through in a year. The formula is COGS divided by average inventory value. Industry average for manufacturers is approximately 5–6 turns per year. World-class lean manufacturers in high-velocity industries achieve 12–15 or more.

Declining turns are a leading indicator of compounding problems: excess raw material, work-in-process that cannot clear quality holds, or finished goods building because demand forecasts were wrong. Each of these shows up as a balance sheet problem before it shows up as a P&L problem — which is why many operators miss the signal until it has already affected cash flow. An operating intelligence system that connects inventory turn data to production throughput and quality hold queues surfaces the signal days or weeks earlier.

COGS Per Unit

COGS per unit is the total manufacturing cost — direct materials, direct labor, and allocated overhead — divided by good units produced. It is the financial summary of everything the plant-floor metrics are describing at a technical level.

The reason COGS per unit belongs in an operating intelligence system rather than just a finance report is that it changes in real time based on what the plant is doing. A downtime event raises it. A scrap spike raises it. A throughput improvement lowers it. Waiting for the monthly close to see COGS per unit means operating blind for three to four weeks after the conditions that drove the change.

COGS as a percentage of revenue typically runs 50–70% in discrete manufacturing, depending on the industry and product mix. A 1-point improvement in gross margin at a $50M revenue manufacturer represents $500K in annual profit. That is the scale of the prize that operating intelligence makes visible and actionable.

Connecting Plant-Floor Data to Financial Outcomes

The individual metrics above are well understood. The operating leverage comes from connecting them. Here is how the connections work in practice.

The OEE-to-Margin Link

Every point of OEE improvement reduces cost per unit because fixed overhead is spread over more good output. Consider a line with $15,000 per shift in fixed costs producing 3,000 units at 75% OEE. Moving to 80% OEE produces 3,200 units — the same $15,000 fixed cost now averages $4.69 per unit instead of $5.00. Over a full year across multiple lines, this is a material COGS improvement that is entirely invisible in a standard operations report unless someone explicitly builds the OEE-to-cost bridge.

Scrap Rate and Inventory Turns

Elevated scrap rates affect inventory turns through two channels. The direct channel: scrapped material is inventory that was consumed without producing revenue, effectively reducing the COGS denominator for turns without reducing average inventory. The indirect channel: high scrap rates often trigger safety stock increases upstream — procurement buffers against yield uncertainty by ordering more raw material, which inflates average inventory and suppresses turns further.

An operating intelligence system that surfaces this chain — scrap spike leads to safety stock increase leads to turns deterioration — gives procurement and production teams the information to break the cycle rather than each team responding rationally to local signals while the system-level problem persists.

Cycle Time and On-Time Delivery

When actual cycle time drifts above takt time — even intermittently — the schedule absorbs the slack by pushing out delivery dates or authorizing overtime. Neither outcome is visible in any single system. The MES shows cycle time. The ERP shows delivery dates. The finance system shows the overtime cost. Only by connecting them does a COO see the full cost of a cycle time problem: not just slower throughput, but the customer service cost and labor premium that came with it.

The Operating Intelligence Layer for Manufacturing

An operating intelligence system for manufacturing is not the MES, the ERP, or the quality management system. It is the layer that connects them into a coherent operating view. Its job is to answer three questions continuously:

1. What is performing below standard right now? — Equipment, lines, shifts, or SKUs that have deviated from target on OEE, FPY, throughput, or cycle time.
2. What is the business impact of that deviation? — The margin and delivery consequence of current plant-floor conditions, not last week's conditions.
3. Where is the highest-leverage intervention? — Which single problem, if fixed, would move COGS per unit, OEE, or on-time delivery the most.

None of these questions can be answered from within a single source system. The MES answers question one for its own domain but cannot answer questions two or three. The ERP answers question two at the end of the month but cannot surface the real-time signal from the plant floor that explains why. Only the cross-system operating view answers all three, continuously.

What a Manufacturing Operating Intelligence Dashboard Tracks

A well-designed manufacturing operating intelligence view includes five interconnected layers:

• Equipment layer: OEE by machine and line, broken down into availability, performance, and quality components. Downtime logs with reason codes. TEEP for capacity planning context.
• Quality layer: First Pass Yield by product and process step. Scrap rate trends. Rework hours consumed. Quality hold queue depth and aging.
• Output layer: Throughput versus plan by shift, line, and SKU. Actual versus ideal cycle time. Takt time adherence.
• Inventory layer: Raw material turns and days on hand. WIP queue depths by stage. Finished goods turns and aged inventory exposure.
• Financial layer: COGS per unit by product line. Overhead absorption rate. Gross margin by SKU — connecting plant-floor performance to the P&L in real time.

The Data Integration Challenge

The most common obstacle is not analytics capability — it is data connectivity. MES systems (Siemens Opcenter, Rockwell FactoryTalk, SAP ME), ERP systems (SAP S/4, Oracle, Infor), and quality systems all use different data models and update frequencies. Machine-level sensor data updates in milliseconds. ERP transactional data updates on order completions. Financial data updates on shift close or daily batch.

A viable operating intelligence architecture for manufacturing handles these update frequencies deliberately — using near-real-time streams for equipment and quality signals and batch synchronization for financial data — and reconciles them into a common time dimension so that the operating view is internally consistent even when source systems update at different cadences.

This is where most homegrown solutions break down. It is straightforward to pull OEE from the MES and COGS from the ERP. It is substantially harder to build the reconciliation logic that keeps them interpretable together, handles missing data gracefully, and surfaces anomalies automatically rather than requiring someone to look for them.

From Reporting to Action

The distinction between a manufacturing intelligence system and a manufacturing report is simple: a report describes what happened. An intelligence system triggers action. The operational difference is significant.

When OEE drops below threshold on a high-volume line, a report captures it in the weekly summary. An intelligence system surfaces it to the shift supervisor and the maintenance queue while the shift is still running. When scrap rate on a particular part number begins trending above baseline three standard deviations from norm, a report shows it on Friday. An intelligence system flags it on Tuesday when the trend is two days old rather than a week old.

The financial scale of this difference is not small. A high-volume line running below OEE threshold for a full week at the cost of $2,000 per point of OEE per shift is a $50,000–$80,000 problem by the time a weekly report surfaces it. The same signal caught on day two is a $10,000–$15,000 problem — the same root cause, but caught before it compounds.

Building the Business Case for Operating Intelligence in Manufacturing

Manufacturing executives evaluating an operating intelligence investment typically need to answer two questions: what is the addressable improvement, and what does it take to capture it?

On the first question, a useful benchmark is the gap between current OEE and world-class OEE. A plant running at 63% OEE with a world-class target of 85% has a theoretical improvement opportunity of 22 percentage points of productive capacity. Even capturing 25% of that gap — moving from 63% to 69% — represents a material COGS improvement and throughput increase without adding headcount or capital equipment.

The path to capturing that improvement runs through visibility. Operators cannot fix what they cannot see, and they cannot see what lives in three separate systems that never update together. The operating intelligence investment is, in practice, the investment in making the gap visible and actionable in real time rather than retrospectively.

On the second question, the critical requirement is data connectivity. The operating intelligence system is only as useful as the data it can access. A realistic deployment sequence starts with the highest-value data sources — typically the MES for OEE and throughput, the ERP for inventory and COGS, and the quality system for yield and scrap — and builds out from there. A pilot on one high-volume line with clean data integration typically demonstrates ROI within 60–90 days, which is the foundation for the broader deployment case.

FAQ: Operating Intelligence for Manufacturing Companies