Zebra Spindle Index

What the index is

A coincident measure of U.S. metalworking activity constructed entirely from public federal data — output indices, hours, deflated order dollars, kilograms of carbide crossing the border. No opinion surveys. Every number links to its .gov source. Zebra publishes a separate proprietary Coolant-System Activity Pulse alongside the index as an independent real-world reference signal; the Pulse is not an input to the ZSI calculation. Site copy makes no forecasting claim: the 2026-08 backtest found no out-of-sample lead over output's own history.

How to read it vs. a PMI: a PMI or diffusion index reads 50 = as many respondents up as down. The ZSI reads 100 = the calibration-period norm of measured activity; 10 points = one robust standard deviation. 110 is not "10% growth"; it is one SD above normal.

Recipe (reproducible with a spreadsheet)

1. Transform each series: seasonally-adjusted output/orders → annualized 3-month-on-prior-3-month log growth; hours → 3-month average minus prior 3-month average; NSA trade quantities/values → year-over-year log change of the 3-month average. Nominal orders and export values are deflated by the PPIs listed below before transforming.
2. Standardize with a frozen robust z: z = (x − median) / (1.4826 × MAD), median and MAD computed once over the calibration window below and re-based each January by published note; clipped to ±3.
3. Combine with fixed domain weights (table below). If any component is missing the month is not published — weights are never renormalized.
4. Scale: ZSI = 100 + 10 × Σ(weight × z).

Series

Tooling imports use pinned 10-digit HTS lines: inserts 8209.00.00.30, 8209.00.00.60 (kilograms). Solid cutting tools (satellite, count): 8207.70.30.30, .70.60.30, .70.30.60, .70.60.60, .50.20.55, .50.20.80, .50.40.30, .50.40.60, .80.30.00, .80.60.00, .60.00.30, .60.00.35, .60.00.61, .40.30.00, .40.60.00, .90.30.30, .90.30.85, .90.45.00. Rock-drilling, dies, stamping, files and woodworking lines are excluded by list. Effective duty rate = calculated duties ÷ customs value, from the same Census records.

How to use it — a historical regularity, not a forecast

The index is coincident, and we tested backlog, capital-goods orders, machine-tool orders and Fed expectation surveys as leading additions: none beat the simplest rule out of sample. That rule is itself the useful one. Since 2005, when machine-shop output was growing (positive 3-month-on-3-month growth), output over the following six months ran above the current quarter about four times in five; when it was contracting, below about three times in four (overall ~78%, machine-shop IP, COVID months excluded; details in the project's OUTLOOK_RESULTS). Read the direction of the output components, not the level, and plan against that — knowing that turning points are exactly where the rule fails.

Revisions

Source agencies revise. Two histories are kept: the number shown on each release day (as-published, never rewritten) and the latest-revised history (the headline chart). Both are in the public JSON feed (free data key). A late agency release produces a dated delay notice, never a partial or guessed value.

What it cannot do

1. It is slower than a survey — measurement lags mood by about six weeks.
2. It is coincident, not leading.
3. Export series are deflated values, not units.
4. Imports are a slice of tooling consumption; domestic tooling shows up only through output.
5. It measures the industry, not your shop.
6. It will be revised.
7. Line-level trade data begins in 2010 — one and a half cycles plus COVID; no recession-detection claims.
8. The Zebra Pulse is one company's activity, shown coarsely and with a delay.

Zebra Coolant-System Activity Pulse (separate, proprietary)

Derived from activity in Zebra consumables and replacement parts — never revenue or dollars, capital equipment excluded — aggregated, indexed so that 100 = normal, rounded to whole numbers, and published with a deliberate delay: a Pulse observation is never more recent than the latest index month, and the public chart shows a rolling window of recent months only. Observations may be suppressed when sample size or customer concentration does not meet Zebra's confidentiality standards. A published Pulse value is frozen and not revised. The Pulse is independent of the ZSI, contributes nothing to its calculation, and is Zebra Skimmers' proprietary data — not part of the CC BY 4.0 dataset. It reflects Zebra's share, pricing and execution as well as the market.

Frequently asked

What does the Zebra Spindle Index measure?
Current U.S. metalworking activity, assembled entirely from measured federal data: machine-shop industrial production, weekly hours, deflated new orders, carbide-insert imports by weight, and machine-tool and insert exports. It contains no opinion surveys and no Zebra company data.

What is the Zebra Coolant-System Activity Pulse?
A separate, proprietary reference series from Zebra Skimmers' own consumables and replacement-part activity, published beside the index as a real-world cross-check. It is delayed, aggregated, rounded, subject to confidentiality suppression, and is not an input to the index.

How do I read the number?
100 equals the 2012–2024 norm of measured activity; every 10 points is one robust standard deviation. 110 is one SD above normal, not 10% growth. It is not comparable to a PMI's 50 line.

Is it a leading indicator?
No. The published backtest found no out-of-sample forecasting value beyond output's own history, and a follow-up test of backlog, capital-goods orders and expectation surveys did not beat the naive rule either. What holds historically: when machine-shop output is growing, the next two quarters have run above the current quarter about four times in five since 2005 — a regularity to plan against, not a forecast.

How is it different from the Gardner Business Index?
The GBI is a diffusion survey — the share of respondents saying up versus down. The ZSI is a magnitude index built only from measured quantities, dollars, hours and output. They answer different questions and can disagree.

How do tariffs affect it?
Tooling imports are measured in kilograms, not dollars, and the effective duty rate for each tariff line is published alongside so price effects are visible rather than hidden.

Can I reproduce it?
Yes. Every series id, tariff line, deflator, weight, calibration median and MAD is on this page; the history is available as JSON and CSV with a free data key.