About the FXR Interest Index

Each month we pull US interest-over-time for the basket from Google Trends (Google Trends). Trends reports relative interest on a 0 to 100 scale inside the export window (share of Google search interest, not a count of searches). It is not comparable to Keyword Planner volume estimates.

The composite is the equal-weight mean of the basket's raw Trends values for that month. The series is then scaled so the average of calendar year 2012 equals 100. Published points start at 2010-01. Months in 2010 and 2011 are shown even when early Trends sampling is sparse (fewer than 4 active basket terms); they are not dropped from the series.

A month needs at least 4 active basket terms (Trends value > 0, with literal <1 treated as 0.5) to stay on the chart after 2012. Calendar-2010 through 2012 months are kept even when fewer than 4 terms are active (2012 still anchors the base average). Months that fail the min-4 rule after 2012 are dropped, not plotted as zero. The series is monthly; seasonal cycling is visible (spring/summer riding season vs winter). Headline latest, MoM and YoY (percent change on the normalized index) use the last complete calendar month; the current month stays on the chart as provisional.

Unexplained spikes. When harley fxr is an extreme outlier versus its own same-calendar-month seasonal norm and the other basket terms barely move (currently 2012-09 and 2026-08), the chart still marks the raw spike as an annotated marker, but the main index line replaces harley fxr with a typical same-month value: the median of other years' same calendar month in 2010 to 2025 (excluding the spike year). Both values are stored (index normalized, indexRaw unadjusted). A 3-year trailing average of the normalized index is plotted as a second line: at each published month it is the mean of the prior 36 published index points (about three years). It starts after the 36th point so the line stays continuous.

Equal weights across all thirteen terms. Bare fxrs is excluded from the composite (too much non-Harley noise). Bare fxr is not in the basket either: the FXR clothing and apparel brand dominates that query, so motorcycle interest would be swamped. Batch exports that share the overlap term harley fxr are joined on that common scale. Dropped from the basket as dead or near-empty in the Trends exports: eric buell fxr (all zeros), cvo fxr, and fxr issues. Low-signal terms that stay because they show real activity: harley fxr parts, fxr vs dyna, fxr2, fxr3, fxr4, shovelhead fxr, and fxrd.

• Do not read the index as absolute US search volume. Trends is relative interest only.
• Website Trends re-scales each export. We use matching all-years monthly US exports for the published series.
• Series starts 2010-01; 2010-2011 months are shown even when sparse. After 2012, months with fewer than 4 active terms are omitted.
• Seasonality shows up in the monthly series; treat short MoM percent moves with care. The in-progress calendar month is provisional on Trends; headline stats skip it.
• Unexplained harley fxr spike months are normalized on the main line; the raw spike remains as a chart annotation only.
• Refresh is monthly by hand for now. An official Trends API path may automate later (Google Trends API alpha).
• FXR Archive is not affiliated with Google. Index methodology © FXR Archive.