logoGreenWhite

Please enter your email address to access the 2019 State of the Industry: CNC Machining

Intro
It’s an exciting time to be a manufacturer.

With IoT coming to maturity and achieving market penetration, we have on our hands a whole new arsenal of data to drive decisions with.

In an increasingly complex manufacturing environment, manufacturers are looking for any edge they can get to remain competitive in a heavily globalized economy.

Over the last 3 years, MachineMetrics has collected data directly from the controls and sensors of thousands of machine tools. These machines span across hundreds of companies in the United States, representing all kinds of machine tools in all types of industries. Today, we are releasing this data in a “State of the Industry” report for the first time ever.

Utilization heatmap, all companies
01_MM19_Big_heat_map_v2
The Survey In Numbers
First
State of the Industry Report by MachineMetrics
Thousands
of machines sampled
1,000+
Machine-years of data, collected over three years
The Challenge
The difficulty of collecting such a dataset cannot be understated.

Data in the manufacturing industry continues to be siloed by individual companies, with each manufacturer keeping their data for themselves. This is understandable, as without a neutral third-party, sharing data can feel risky and unwarranted.

This isn’t to say that third-parties haven’t attempted to collect industry data. Industry associations have conducted national surveys for decades through traditional means. These surveys are all voluntary surveys that require data from the contributor in return for survey results. Companies send their data up to the Association and receive overall industry trends in exchange for their data. While this may seem like the most straightforward way of getting insights into the industry, the approach actually has an array of drawbacks.

np_survey_1051332_393b3e
A note on survey bias
We find that the top three reasons for lack of accuracy in self-reporting surveys are as follows:
circle
Nonresponsive or altered repsonses
Due to embarrassment of low numbers
folder
Imprecision
Due to lack of accurate reporting controls
malic
Malicious Reporting
To sway results and throw off competitors

It takes time for the respondent to compile and send their answers every month, which can burden the company with significant overhead especially if there is no infrastructure set up for reporting.

Economic downturns or slowdowns in business often cause non-reporting, as more imperative business matters trump using resources to report to a survey. This, unfortunately, is when your business needs industry information the most -- in a month where there’s a blip in sales, knowing where you stand and what segments to focus marketing efforts on can be of paramount importance to recovery.

With the State of Discrete Manufacturing report, manufacturers don’t need to worry about reporting or not reporting. The insights will be there regardless.

Machine Data
Two Critical types of data are collected:
Details
Metadata collected Includes:

• Machine type
• Geography (down to city)
• Timestamp down to the millisecond level
• Company information
• Machine shop type

Types of machines include:

• Mills
• Lathes
• Swiss CNCs
• Grinders
• Stamps

Data Collection
Our unique architecture cleans and standardizes data, without human error
edge-circle
Edge Connectivity
  • Plug-and-play machine connectivity from PLCs, digital I/O, and analog sensors
  • Transformation of machine data to standard data structures
  • High-frequency data analytics at the edge
infrastructure2
IIoT INFRASTRUCTURE
  • Multi-tenant cloud infrastructure optimized for machine data
  • Machine performance and condition reporting via APIs and BI-Integrations
  • Rules-based workflow triggers for any shop floor data item
applications
Applications
  • Vertically-focused use cases digitize legacy manufacturing processes
  • Text and email notifications enhance reaction time to problems
  • Step-by-step workflows deliver optimized processes to factory workers
cluster
Data is algorithmically anonymized, cleaned and compiled before it makes its way into our report.
sml1
What differentiates our report is that there is no human in the loop to corrupt the data collection process.

Through the adoption of the MTConnect standard, we collect normalized data items from all of our machines and aggregate them programmatically. Our data is not obtained through questionnaires, but through automated data pipelines. There is no manual process of phoning participants, entering data into forms, or dealing with survey dropouts.

sml2
A large part of what we’ve been doing over the last half-decade is compiling a library of adapters that can pull data directly off the PLCs (programmable logic controls) of many different types of machines.

These adapters are constantly pumping fresh data into our pipelines every second, and we’ve gotten to the point now where we’ve compiled thousands of machine-years worth of machining data. While our machines span across the globe, this report focuses on the United States and comprise a representative sample of American machine shops.

Topline Numbers
25.2%
Average Utilization Q3 2019
-6.5%
Percent Change Quarter over Quarter
-19.4%
Percent Change Year over Year
combined-shape
Key Takeaway

Machines were generally utilized between ¼ and ⅓ of all time (24/7, 365 days a year), with a significant decrease year-over-year. This is about one-third of what has been previously reported.

Machine-Level Insights
Machine Type Utilization by Time (Month to Month)
Over the course of this year, we see trends for each machine type, with Grinders being especially high performers. Utilizations for most machine types vary approximately in tandem with each other, showing stronger performance earlier in the year. This potentially reflects larger economic trends.
2019 Utilization Trend Over Time, By Machine Type
02_MM19_Utilization_Trend_by_Machine_Type-1
  • Machine Type Utilization by Day of Week/Hour

    Day-specific trends revealed themselves, across all machine types:

    • Production falls in the wee hours of the morning due to lack of personnel, but by different degrees for different machine types. For example, for Swiss CNCs, barfeeders require periodic manual reloading. In contrast, Grinders tend to be run all night, suggesting that they are easier to automate.
    • Saturday utilization is certainly lower, and Sunday is the real “day of rest” in manufacturing.

    “People don’t run lights out very effectively”

    03_MM19_Utilization_by_dayf_weekhour
  • Density Plot

    A density plot can show us the distribution of utilizations across all of the machines in our dataset, again broken down by machine type. For example, we can see that stamps are consistently modest performers, with utilization typically around 25%, whereas grinders are routinely run above 50%. For most machine types, the distribution is relatively broad, with a fat tail of highperformers that skews the mean utilization upwards.

    04_Density_Plot_v2
  • Downtime Reasons

    In addition to collecting raw utilization data, we also prompt operators to provide reasons for long periods of downtime. Looking at the rates at which various downtime reasons occur reveals a kind of “fingerprint” for each machine type in terms of its average operational and maintenance needs. Stamps stand out as particularly prone to losing productivity due to lack of an operator, and also tend to be stopped more often for changeovers. This is consistent with their overall low utilization. Horizontal lathes exhibit relatively high rates of cleaning and planned mechanical service compared to other machine types. By contrast, Grinders and Swiss CNCs encounter downtime more rarely and generally for broader sets of reasons. A shop owner could learn a lot about their own machines and operations by comparing against these patterns.

    00_MM19_Downtime_Reasons
Factory Level Insights
Let’s dig into the factory level numbers, which aggregate all types of machines together.
  • Density Plot

    We can compute utilization for each of the companies that we work with by averaging over all of the machines in their individual shops. What happens when look at the distribution of these company-level utilizations among our customers? Unsurprisingly, the curve is quite similar to the density plots over machines. Most companies appear to perform at the 25% utilization level, with a fat tail of high-performers stretching into the 60’s. A shop owner can get an immediate sense of how competitive they are by just learning where they live on this plot.

    06_MM19_Density_Plot_v3
  • Utilization by hour

    When we look at average utilization over the course of a day, certain “key hours” are revealed. Factories seem to have “high utilization” hours from 8 am to 4 pm, with a slow and steady ramp-down period from there until the next day at 8 am. If we look at the line plot to the right we can see that 4 AM is the slowest period, with 10 AM being the time at which most machine shops are firing on all cylinders.

    07_MM19_Utilization_by_hour_v2
  • Utilization by day of week

    Aggregating utilization on day of week, we confirm some biases that may have been common sense to factory owners, but putting numbers behind it has been hard until now. It’s known that weekends tend to have lower utilization, but an interesting find is that Mondays and Fridays are also significantly lower than middle of the week utilization. Perhaps having a case of the Mondays is a thing after all, at least in manufacturing.

    08_MM19_Utilization_by_dayf_week_v2
  • Heat Map

    A way to look at hourly and weekly data simultaneously is in the form of a heatmap. In this form, we can quickly identify the dropoffs in overnight productivity in and around weekends, as well as the fact that the mid-day peaks are strongest in the middle of the week. We even spot typical break times as lightercolored horizontal bands.

    00_MM19_Heat_Map
  • Utilization by GEOGRAPHY

    Another insight we can get is for regional trends. The Northeast appears to have the strongest utilization, with the South’s utilization trending downwards over time so far this year. How can this inform your business? Perhaps if you’re a manufacturer in the South, you can rest assured that a downtrend in business may be systematic, rather than due strictly to your business practices. Or perhaps if you’re looking to move your factory to another region of the US, this can help you make an informed decision on if the cost/benefit analysis is worth it. Is the 4% additional utilization you’d get from moving to the Northeast worth a 15% increase in labor costs?

    09_MM19_Utilization_by_Geography-1
  • Downtime reasons

    What about downtime? What are the most common reasons for downtime? Should you be worried that lack of utilization at your facility is due to labor not being available? According to our data, the most common reason is actually because there are not enough operators. So if you are constantly in need of labor, you can rest assured that this has been quantified as an industry norm. It confirms the widespread belief that manufacturers can’t find enough skilled labor to run their shops. You hear about all those great vocational jobs that aren’t being filled — and we can see it reflected in the data right here.

    You can also use downtime benchmarks to inform your business planning practices -- perhaps if you have a disproportionate amount of downtime due to changeovers compared to the industry, you should consider analyzing ways in which this could be better streamlined to better align with your peers.

    10_MM19_Downtime_Reasons_v3
ECONOMIST LEVEL INSIGHTS
OVERALL TRENDS

Aggregating over all machines for all of our companies, industry-wide trends are revealed. One of the most basic questions that we can ask is: “On what days of the calendar year are machine shops actually up and running?”

UTILIZATION HEAT MAP, ALL COMPANIES
11_MM19_Big_Heat_Map_V2
Holidays

Utilization is lower for the week of Christmas, suggesting it’s when many shops are shut down, and not for many federal holidays like Columbus day or MLK day.

These insights can help inform your HR policy. What days are other employees taking off? Perhaps it makes the most sense to align your operation’s holidays with those of the industry, which can help align resources to best serve your customers.

  • DAY OF WEEK UTILIZATION OVER TIME

    When we break down our aggregated trend over the past calendar year by day of week, we can of course immediately infer that weekdays are more productive than weekends. But we also see that utilization becomes less stable as we move away from mid-week. Weekends are particularly volatile, fractionally speaking, but in a manner that is correlated with what is going on during the weekdays. Might ups and downs in the manufacturing market have an outsized effect on overtime work?

    12_MM19_Dayf_week_utilizationver_time
  • HOUR OF DAY UTILIZATION BY MONTH

    When we break this down by hour of day utilization trends over time, one immediate thing jumps out at us: the lines are essentially parallel to each other. This means that the relative rank of each hour remains remarkably consistent over time — hour X may fluctuate in utilization, but relatively speaking, it always remains as the Y most utilized hour when compared to all other hours. The only exception we see to this is 6 AM. For some reason, that hour seems to have the most rank-volatility. Perhaps this is when first shift is getting in, or when many changeovers are happening, causing increased variance in utilization.

    13_MM19_Hourf_day_utilization_by_month
  • Correlation to economic indicators

    And how do we know our insights are legitimate? We’ve been tracking our correlation to several economic indicators, and we’ve found ourselves to be highly correlated to several key economic series. Since mid-2017, we’ve been very closely tracking two series - Industrial Production for Miscellaneous Metal Goods (the Fed’s proxy for Medical Device Manufacturing) and Value of Manufacturers’ Shipments for Motor Vehicle Components. This makes sense to us, as automotive and medical are two of the biggest industries discrete manufacturers serve. When MachineMetrics’ customers manufacture more of the component parts of cars and medical devices, their utilization goes up. The output of machined goods like engines, frames, and medical screws serves as a very tight correlate to the production of the products they ultimately go into: motor vehicles, trucks, metal implants, etc.

    14_MM19_Utilization_vs_Valuef_Auto
“The data will always be honest”
BENCHMARKING IN ACTION

A previous shop owner has said that if they had these reports, they would go straight to the bank for funding for new machines, citing third-party evidence that they were best-of-class and already above the industry norm in terms of utilization. He would also bring it to customers and potential customers to demonstrate that his shop is one of the top shops, and that they had data to prove it.

Conclusion
DATA TELLS A STORY

We release a MachineMetrics index of manufacturing, which is calculated by indexing the utilization of January 2018 to 100. We see that the index has a slight downwards trend, but is overall quite stable. Through this index, we can get a good idea of how manufacturing is progressing over time.

This report may confirm your gut feelings, or it may offer unexpected insights that you never would have anticipated. Analyzing information from machines is like analyzing the heartbeat of a person— many insights can be realized, problems can be diagnosed, and truth can be revealed through the study of data.

15_MM19_Index_for_Conclusion-1