# The parallel run

**Four Sundays. Both systems. Same figures, or find out why.**

The workbook is not retired because the new system looks right. It is retired
after the two have produced the same figures, independently, for four
consecutive Sundays — and after the session has minuted the decision. Banks
call this parallel running. A parish's money deserves it for the same reason.

---

## Before the first Sunday

1. **The districts must match exactly.** Their names in the system must be the
   names in the workbook, spelled the same. The comparison matches on name; a
   "Tudor Tum" against a "Tudor tum" reads as two districts and one missing.
2. **Set the financial year to the months the parish actually uses.**
   Makupa's runs October to September. Every target follows this date.
3. **Enter the budgets** from the workbook's Budget column.
4. **Do not import the old year yet.** Import it after the parallel run
   succeeds. Importing first means the two systems are no longer independent,
   and an agreement proves only that a copy equals its original.

---

## Each Sunday, for four weeks

**Keep the workbook exactly as you always have.** Nothing changes about how it
is filled in. If the parallel run causes anybody to fill the workbook
differently, the test has been spoiled.

Then, separately, enter the same Sunday into the system — from the counting
sheets, not from the workbook. Two people entering from the same source, not
one person copying.

Then compare:

```
php tools\parallel.php "C:\path\to\District Report.xlsx" --week 08.04.2025
```

It prints one line per week, or every difference by district and column, with
the exact shilling. It writes to neither side; running it cannot change
anything.

---

## When a difference appears

**A difference is good news the week it happens and bad news in June.** Chase
every one before the next Sunday. In practice they are:

| What you see | Usually means |
|---|---|
| One district, one purpose, a round difference | An entry typed into one place and not the other |
| One district, small difference, week and year both | A keying slip — 2,520 entered as 2,250 |
| One district, everything zero in the system | That district's sheet was not entered |
| A district in one and not the other | A name spelled differently |
| Year-to-date differs but the week agrees | An earlier week diverged; go back and find it |

Record which side was wrong. If it is the workbook more often than the system,
that is worth telling the session — politely, and with the weeks named.

---

## Retiring the workbook

After **four consecutive agreeing Sundays**, the tool says so. The decision
is still the session's, not the software's:

1. Table the four comparison printouts at the deacons' court.
2. Minute the decision to retire the workbook.
3. Import the historical year (`Import old workbook` in the system), which
   brings the district-level history across. Note that the old book never
   recorded who gave, so individual giving history begins now.
4. Keep the final workbook file where the **parish** controls it — not only on
   the office PC and not only with the developer.

Fewer than four? Keep both running. A system that has agreed three times has
agreed by luck about as often as by construction.