Aug 27 2021
My last two long posts were about evaluating sales forecasts. They begged the question of how you generate these forecasts. This is a partial answer, about what you can tell from a history of sales through both classical methods and recent developments, particularly probability forecasting.
Apr 2 2021
Pfizer-BioNTech just announced the results of a COVID-19 vaccine clinical trial of 12- to 15-year olds. Because the vaccinated group had 0 infections, the news media jumped to the conclusion that the vaccine has “100% efficacy” on 12- to 15-year olds. That is what the chyron said on NBC news. A look at the published trial results and quick analysis with current methods confirms that the vaccine works for that population but not that it eliminates 100% of the infections.
Sep 26 2020
The post on Does Toyota Use SPC? elicited many comments on LinkedIn. Some suggested that it was scoping SPC too narrowly when contrasting it with Toyota’s approach. In fact, SPC as referenced in the post is the body of knowledge described in the American literature on quality and taught in professional courses.
As to why Toyota is not using SPC, the answer is simple: SPC is about process capability and the quality problems Toyota addresses in 2020 are not due to lack of process capability. In industries that lack process capability, modern data science outguns the old SPC toolkit but that is a different discussion. The most vital question raised in the comments was why we have been not learning Toyota’s approach to quality. In the past 30 years, American industry has learned “Lean Six Sigma” instead.
The comments also enriched the public sources of information cited in the post with corroboration by current and former employees of Toyota.
Sep 21 2020
As part of a discussion started by Lance Richardson on LinkedIn, I stated as a fact that SPC was not part of the Toyota Production System (TPS), which prompted several contradictors to tell me I didn’t know what I was talking about. The “evidence” they provided, however, does not refute my statement. It confirms it instead.
Sep 10 2020
Factories are controlled environments, designed to put out consistent products in volumes according to a plan. Controls, however, are never perfect, and managers respond to series of events of both internal and external origin.
An event is an instantaneous state change, with a timestamp but no duration. An operation on a manufacturing shop floor is not an event but its start and its completion are. Machine failures and quality problem reports from customers are all events. Customer orders and truck arrivals are also events.
For rare events, you measure times between occurrences and mark each occurrence on a timeline. For frequent events, you count occurrences per unit of time and plot these numbers over time. On individual machines, you record the times between failures. For incoming online orders for a product, you count how many you receive each hour or each day. Every 200,000 years, compass needles reverse directions. That is a rare event, recorded in basalt oozing from the mid-Atlantic ridge as it cools down, as on a magnetic tape. (See Paleomagnetism.)
Once a year, an American TV viewer tuning in to the Academy Awards is a frequent event. For this, we care how many millions do it, but not how much time elapses between two tune-ins.
Responding to series of events is central to many businesses. Stores respond to customers coming in, airlines to passengers with reservations showing up to fly — or not, maintenance crews to machine failures, social networks to subscriber activities,… The challenges posed by the randomness of the arrival and service processes has given rise to queuing theory and commonly used results like Little’s Law about lead times, throughput and inventory in steady-state, or Kingman’s Rule about the way lead times explode when systems saturate.
We are concerned here with a different and simpler topic: monitoring series of events to detect changes in their rates of occurrence and tell fluctuations apart from shifts with assignable causes. If the arrival rate of quality problem reports from the field suddenly doubles, a sophisticated analysis is not needed. If it increases by 10%, the conclusion is not so obvious.
Jan 9 2022
Always the Hurricanes Blowing (Part 2)
This post and the previous one use Atlantic hurricanes as a vehicle to show what various visualizations can do. It’s not about second-guessing the data scientists at NOAA who have produced similar displays and much deeper analyses. The point is to show tools anyone can apply to data that may have nothing to do with hurricanes:
While the previous post aimed to show richer visualizations than possible with 100-year old techniques but it was still limited to a few static displays. This means charts that look the same in print and on a screen. This post includes dynamic displays, with animation and interactivity, that you can only use on a screen, and analyses of more of the columns in the HURDAT2 database.
The technology I used to produce these charts takes work but didn’t cost me a dime in license fees. The resulting charts are trivially easy for readers to understand and routinely used in publications like the online New York Times.
Continue reading…
Contents
Share this:
Like this:
By Michel Baudin • Data science • 0 • Tags: Data visualization, Dynamic Charts, Ridgelines