- Before
- Runs the migration plan from cutover runbooks and defect logs, and learns about a wrong merge when the member calls.
- Now
- Signs a batch only once each point is accounted for, with the hard pairs already decided as patterns.
Batch seven of a frequent flyer migration, screen by screen
The airline bought another airline. The acquired program's 4,218,600 accounts move into the airline's own program, with its 18.6M accounts, in nine batches before the acquired program closes. Six batches have moved. Tonight is the seventh: 286,450 Silver and Gold accounts. Here is the day in the working solution.
A worked example on sample data. Names, IDs and figures on the screens and in this section come from the demo.
- 01The morning of the move
Seven items need the manager; the agents did the rest overnight
Loyalty integration managerThe batch moves tonight, 11:00 pm to 3:00 am CT, with redemptions paused for 4 hours. So far 2,330,000 of 4,218,600 members have moved, 3,812 pairs need a person and 0 points are unaccounted for. Overnight: 61,920 pairs found at 1:30 am, 1,212 scores lowered by the second look, 37 pairs held by fraud watch, 286,450 notices written by 6:10 am. The manager asks what could go wrong tonight.
Move agent: “The batch ties to the point. What is left is people’s work on the 3,812 hard pairs.”
- 02One member
Two accounts side by side, and the one the member will have
Matching agentOne member holds a Blue account with 18,640 miles on the airline's side and a Silver account with 42,300 points on the acquired side. Between them sits the account after the move: 71,515 miles (18,640 + 42,300 × 1.25), Silver kept, 14,200 status points carried 1 : 1, an upcoming trip re-pointed and the co-brand card earning into it. Each value opens its source line.
“One question: the member uses two emails. I kept the one on our side for sign-in and the other as second. Is that right?”
- 03Why 0.97
The reasons for the score, and an agent that argues against it
Matching agent and second-look agentIn favor: the same Known Traveler Number (+0.31), the same date of birth (+0.22), the same mobile (+0.18), the same home once written the same way (+0.12). Against: two different emails (−0.03). Asked whether the member flew on both at once, the second-look agent checks the flight records.
Second-look agent: “In 24 months the two accounts did not share a flight on any date… I tried to prove them two people and could not.”
- 04What the member will read
A personal before-and-after notice, on the channel allowed
Notice agentThe notice is written for this member in English or Spanish: 42,300 points plus 18,640 miles become 71,515 miles, Silver stays Silver, and the trip and card move too. It goes by email only: texts were off on one account, and the stricter choice per channel is carried.
“Your miles do not expire for 36 months. We will never ask for your password or your card number.”
- 05The hard pairs
3,812 pairs as eight patterns instead of a grid of rows
Program operations analystThe agents were unsure of 3,812 of the 61,920 pairs, so they grouped them by why they are hard: name variants and nicknames (1,606), married or changed surnames (612), shared email or phone (541), family at one address (388), one account matching two (297), names spelled two ways (214), status conflicts (117) and holds by fraud watch (37). Each carries a proposal and its risk. The analyst approved three; 2,606 pairs await a second signature.
- 06Second signature
A father and son at one address stay apart
Loyalty integration managerIn the family pattern, two members share a name and a home address, were born in 1958 and 1986, and flew together side by side. The rule the agents propose for all 388: keep apart where date of birth or Known Traveler Number differ. The manager reads five samples across the range of scores, sees the 3 pairs pulled out to decide one by one, and signs as second.
“The analyst approved this at 8:12 am as different people, same family. Your second signature puts 385 pairs in tonight’s batch.”
- 07Before the move
Contact changes just before cutover are held
Fraud operations manager214 accounts in the batch changed email or phone in the 14 days before the move. Fraud watch held 37 pairs and refused 9 link requests with a mismatched birth date. Five are left. On one, the email changed to a free-mail address created that day, the phone was removed 4 minutes later, and a link request came from a new device at 1:12 am. On another, the member called to confirm the change, so the agents propose release.
Fraud watch holds only; the fraud operations manager releases or confirms each pair.
- 08Convert
Each point accounted for, to the cent
Conversion agent and loyalty controller9,738,412,652 points × 1.25 under the signed rate, rounded up to the whole mile per account, become 12,173,087,124 miles. Liability before, at 1.75¢ a point, is $170,422,221.41; after, at 1.40¢ a mile, the same, a $0.00 difference. The 71,309 rounding miles, $998.33, are booked as goodwill. The controller signed the liability at 8:05 am.
“Each account carries its own line, so any member’s balance can be traced.”
- 09Signed
Nothing moves until three people sign
Integration manager and head of loyaltyThe sign-off lists what will run: 286,450 accounts, 57,108 merges, 224,530 new accounts, the reconciliation, the window and the snapshot taken before the first account moves. 875 pairs still open at 10:00 pm stay as two working accounts for the next batch. The integration manager signs with the reason batch reconciled, and the head of loyalty co-signs.
“Snapshot first · rollback within 72 hours.”
- 10On the call
A wrong merge, undone in one signed step
Member services leadIn the last batch a son was merged into his father's account: same name, same street, a different birth date. He is on the line. From that batch's snapshot the move agent has the undo ready: 41,250 miles and his upcoming trip go to a new account, with Silver restored. His father's four redemptions since used only his own miles, so nothing is clawed back. A personal letter and 5,000 miles of goodwill go with it.
Merges undone to date: 41 of 486,212, or 0.84 in 10,000.