Identify the event being counted
A permit authorises work; a start records the beginning of construction under the survey definition. Completion comes later. The unit of observation is a housing unit, not a construction site: a building with many apartments can move a large number of units into one stage together. Permits cover places where permits are required, while the construction statistics also estimate activity in non-permit areas. Comparing a national permits total directly with a completions total therefore mixes timing and coverage.
Match the number to the supplier
A supplier of foundations is exposed to starts; a maker of fitted kitchens generally serves a later stage. Suppose eight developments of forty apartments each are permitted in June and none breaks ground that month. The illustrative count is 320 authorised units, zero new starts for those developments and no immediate evidence of kitchen deliveries. If five developments start in September, the timing of demand for fittings still depends on construction progress and the contractual delivery milestone. The same number of single-family homes could be spread across many independent sites with quite different delivery schedules.
Read the stock between stages
Units currently under construction can reveal a queue that monthly starts miss. A large queue may herald future deliveries or an inability to finish projects because financing, labour or materials are unavailable. Separate single-family and multifamily projects and look at completion times before choosing between the explanations. Seasonally adjusted annual rates are rates, not actual units delivered in the stated month. Use unadjusted counts when reconstructing a specific period and avoid subtracting annualised flow measures from a stock of unfinished units.
Practical check
For each building product, name its installation stage and lead time before attaching a US housing indicator to an earnings forecast.