Enum & Status-Value Drift Sweep

A field is SET CLOSED, VARIANTS PRESENT, or SET BROKEN. The sweep takes the worst field: SETS HOLD, DRIFT FOUND, or SETS BROKEN.

Most drift tools stop at spelling: here are the values that do not match, go and fix them. That list is usually enormous and mostly harmless, which is why nobody works through it.

The interesting question is not whether a variant exists. It is how many rows are sitting on it. Below a threshold it is a typo somebody made once. Above it, that variant is a value your systems write regularly and your code does not branch on — which means somewhere a report is undercounting and a workflow is silently skipping rows. At or above 5% of a field’s rows, the field reads SET BROKEN even with zero out-of-set values. The threshold is a cell on the Dashboard; move it if your tolerance differs.

ticket.status and order.status hold the same 76,088 rows across the same five distinct values. Move how many of them sit on the variant.

ENUM & STATUS-VALUE DRIFT SWEEP
==========================================================================

ticket.status  (branch-driving)  VARIANTS PRESENT
    76,088 rows across 5 distinct values; 0% of rows on variants
      open                     12,400  IN SET
      in_progress               8,100  IN SET
      resolved                 15,200  IN SET
      closed                   40,300  IN SET
      In Progress                  88  VARIANT  -> in_progress
    fix first: In Progress

order.status  (branch-driving)  SET BROKEN
    76,088 rows across 5 distinct values; 9% of rows on variants
      open                     12,400  IN SET
      in_progress               8,100  IN SET
      resolved                 15,200  IN SET
      closed                   33,900  IN SET
      In Progress               6,488  VARIANT  -> in_progress
    [GATE] variants carry enough rows to be a second value in production, not a typo
        values alone read VARIANTS PRESENT; the volume overrides it
    fix first: In Progress

invoice.state  (branch-driving)  SET BROKEN
    25,560 rows across 5 distinct values; 0% of rows on variants
      draft                       900  IN SET
      sent                      4,200  IN SET
      paid                     18,700  IN SET
      void                        310  IN SET
      partially_paid            1,450  OUT OF SET
    fix first: partially_paid

lead.source_channel  (display only)  SET BROKEN
    10,900 rows across 4 distinct values; 8% of rows on variants
      organic                   5,000  IN SET
      paid                      3,000  IN SET
      Referral                    900  VARIANT  -> referral
      email                     2,000  IN SET
    [GATE] variants carry enough rows to be a second value in production, not a typo
        values alone read VARIANTS PRESENT; the volume overrides it
    fix first: Referral

subscription.plan  (branch-driving)  SET CLOSED
    32,300 rows across 3 distinct values; 0% of rows on variants
      free                     22,000  IN SET
      pro                       9,100  IN SET
      enterprise                1,200  IN SET

--------------------------------------------------------------------------
SWEEP VERDICT: SETS BROKEN
  1 closed / 1 with variants / 3 broken  (of 5 fields)
  4% of 220,936 rows sit off the declared set  -- context only, it decides nothing
  gate fired on 2 field(s) (>= 5% of rows on variants)
  fix first: order.status -- In Progress

Grades the values you paste in. Not legal or compliance advice.

The .xlsx reproduces every verdict above exactly — the parity check behind this kit asserts 371 of them, cell by cell, with zero error values. Three things to notice. The twin pair differs only in how 76,088 rows split. invoice.state is SET BROKEN from a genuinely new value with no variants at all, so the volume gate gets no credit there. And 4% of all rows sit off-set across the whole sweep, which reads SETS BROKEN — the share is context, not the verdict.

• It does not connect to your database. You run the distinct-value query and paste the result; it grades what you paste.
• It cannot tell you which of two values is the right one. It tells you a variant carries 6,488 rows; deciding whether to migrate them or widen the set is your call.
• It does not check whether your declared set is a good set. A closed set with the wrong values in it reads as closed.
• A field that reads SET CLOSED today can drift tomorrow. This grades values that already exist, so there is no evaluation date anywhere and two runs a quarter apart compare directly.

Deterministic and offline. It grades fields, never people, and is a working aid rather than legal or compliance advice.