Sprawl is a symptom,
not the disease
Every mature BI deployment ends up with hundreds of dashboards and a widespread belief that most of them are wrong. The tidy-up projects never hold, because the thing generating them was never addressed.
Why the tidy-up never sticks
Somebody runs an audit, finds that a third of dashboards have not been opened in a year, and archives them. Within eighteen months the count is back, because the mechanism that produced them is untouched.
That mechanism is simple. Somebody needed a number, searching for an existing dashboard was slow and returned four plausible candidates with no way to tell which was authoritative, and building a new one took twenty minutes. They made the rational choice.
Search is not the fix people expect
Better search helps only if the results are decidable. If a query for revenue returns eleven dashboards and the user cannot tell which is trustworthy, more results make it worse rather than better.
What makes results decidable is a definition they all share. When every dashboard showing revenue is showing the same governed metric, picking the wrong one costs you a layout you dislike rather than a number that is wrong.
The number that actually matters
Dashboard count is a vanity metric in both directions. A business with four hundred dashboards built on forty governed metrics is in good shape. A business with forty dashboards built on forty private definitions is in trouble and does not know it yet.
Count the definitions, not the dashboards.
What to do about the four hundred you have
Not much, initially. Leave them. Define the metrics that matter properly, point the dashboards people actually use at those definitions, and let the rest decay in place.
Deleting them is satisfying and largely pointless. An unused dashboard costs almost nothing. An untrustworthy definition costs a meeting a month.