"Feels wrong" almost always means one of four specific errors is
inflating it. Here's the correct calculation and the traps, so you can
see which one you hit.

CORRECT MRR — normalize everything to one month
  Monthly plans      → face value
  Annual plans       → contract ÷ 12  (NOT counted in full the signup
                       month — this is the #1 inflation bug)
  Discounts          → actual paid, not list price
  One-time fees      → EXCLUDED (setup, onboarding, overages aren't
                       recurring)
  Usage that varies  → excluded unless it's a committed floor

THE FOUR THINGS THAT MAKE IT "FEEL WRONG"
  1. Annual contracts booked at full value in the signup month. A $12K
     annual deal is $1K MRR, not a $12K spike. This one inflation makes
     MRR jump and then look like it "churned" the next month.
  2. One-time setup fees counted as recurring. They're revenue, not MRR.
  3. List price instead of paid price on discounted accounts.
  4. Failed payments still counted. A card that's declining isn't MRR
     until it clears.

THE MOVEMENTS — where the real story is
Your total MRR is less useful than its composition:
  New MRR         + from new customers
  Expansion MRR   + from upgrades / seats added
  Contraction MRR − from downgrades
  Churned MRR     − from cancellations
  ─────────────────────────────────────
  Net new MRR     = the number that actually tells you if you're growing

From these:
  Gross retention = (start − contraction − churn) ÷ start   [≤100%]
  Net retention   = (start − contraction − churn + expansion) ÷ start
                    [can exceed 100% — the metric investors care about]
  Magic number    = net new ARR ÷ prior-quarter S&M spend

WHAT TO CHECK FIRST
Pull whether annual deals are booked at full value in month one. That's
the error that most often makes MRR "feel wrong" — it's simultaneously
too high at signup and creates phantom churn later.

Share the subscription data and I'll compute the real MRR plus the four
movements.