Lesson 6 of 6
Cutover and billing-cycle continuity
9 min
Listen to this lesson
A manufacturer can stop shipping for a weekend. A utility cannot stop reading meters, and it must bill on schedule because bills drive cash and regulators watch billing delays. The cutover of an IS-U conversion is planned around the billing calendar: which billing portions fall in the cutover window, how meter reads that arrive during downtime are queued, and how the backlog is cleared afterwards.
The billing calendar
Northwind reads and bills in 21 portions, one per working day of the month, about 30,500 contract accounts each. Meter reading results arrive daily from the meter data system; mass billing runs at night, invoicing and bill print follow. A cutover from Friday evening to Monday morning touches one portion directly (Friday's) and delays the read upload for the weekend. The plan must say, portion by portion, what happens.
| Item | Decision (illustrative) | Risk addressed |
|---|---|---|
| Portion falling in the window | Bill portion 18 one day early, before the freeze | No portion skipped |
| Meter reads during downtime | Queued in the meter data system, uploaded after go-live | No lost reads |
| Payments during downtime | Bank files held and loaded in sequence after go-live | No duplicate or lost payments |
| Customer service | Read-only mode and scripted messages during downtime | No changes lost |
| Dunning | Paused for the first week; first run reviewed manually | No wrong dunning from cutover gaps |
| Bill print | First post-go-live print batch reviewed before release | No wrong bills mailed |
Billing continuity decisions for the cutover.
Go/no-go criteria
- Do FI-CA open items and G/L reconcile after conversion?No → stop; no billing until fixed · Yes → next
- Do the golden bills match exactly in production?No → stop; investigate before any mass billing · Yes → next
- Does a simulation of the next portion match the dress rehearsal comparison profile?No → investigate differences · Yes → next
- Are queued reads and payments loaded without errors?No → fix interfaces before billing · Yes → go: bill the catch-up portions
A simulation of the next portion in production, compared with the profile from the dress rehearsal, is the strongest go-live test available: it uses the real converted data and the real configuration, and it changes nothing. If the share of bills within tolerance, the outsort rate and the totals look like the dress rehearsal's, billing can start.
Hypercare: watching the first cycles
| Measure | Normal (illustrative) | Action if exceeded |
|---|---|---|
| Bills created per portion | About 30,500 | Find missing installations before print |
| Billing and invoicing outsorts | Under 0.5% | Analyze by reason; adjust validations only with the billing owner |
| Estimated bills | Under 3% | Check read uploads and plausibility |
| Customer complaints about bills | At pre-go-live levels | Sample complaints for patterns |
| Days to bill each portion | Same day | Escalate batch chain performance |
Billing measures watched daily for the first full cycle after go-live.
What would you do?
On the first night after go-live, mass billing of portion 19 produces 4.8% outsorts against 0.4% in the dress rehearsal. Bills have not been printed yet. What should the billing lead do?
What goes wrong in testing
- The dress rehearsal skips the billing catch-up, so its duration is unknown on go-live night.
- Queued reads are not tested at volume, so the interface fails when the weekend backlog arrives.
- Print and mail vendors are not part of the rehearsal, so the first production print batch is the first test.
Communicating with customers is part of continuity. Northwind (illustrative) told customers a month ahead that bills would look slightly different and that online services would be unavailable for one weekend, and it briefed the contact center with a script for the first cycle. A bill that looks different but is correct still causes calls; a bill that is late causes more. The contact center's call volume is therefore one of the hypercare measures.
Fallback
A utility's fallback decision has a billing dimension too. If go/no-go 2 fails and the program falls back to ECC, the reads and payments queued during the weekend must be loaded into ECC instead, and the portion billed early remains valid. Rehearse the fallback load in the dress rehearsal as seriously as the forward path: a fallback that has never been run is not a fallback.
- Plan cutover around the billing calendar, portion by portion.
- Queue reads and payments during downtime and load them in sequence.
- Golden bills and a next-portion simulation in production are the go-live proof.
- Hypercare lasts one full billing cycle, with daily billing measures.
Check your understanding · not graded
Lesson check
1. Why is an IS-U cutover planned around billing portions?
2. What is the strongest go-live test before the first production billing run?
3. Outsorts jump from 0.4% to 4.8% on the first night. What is right?
4. How long should billing hypercare last?