Fay

The Handbook

Contents of the handbook

Chapter XIII · Every page, explained

Match suggestions

Possible duplicates and households surfaced for review, with a confidence you can read at a glance.

What match suggestions are for

Import a season of box-office CSVs and a few people will inevitably arrive twice, “Peg” one week and “Margaret” the next, or a couple who each bought their own ticket. Match suggestions is the queue where Fay surfaces those pairs and asks you to decide. Fay works out who might be who, and you say whether it is right: Fay proposes, you dispose, and nothing is ever merged automatically.

The queue lives at /match_suggestions, as the third tab of the People row, People, Households, Matches, because merging duplicates is directory work. The tab carries a count pill when anything is waiting (it might read 7), so you can see from the directory how much there is to review.

Two kinds of match

Every card is one of two kinds, named in its title:

  • Possible duplicate: Fay thinks these two records are the same person, entered twice. The fix is to merge them into one.
  • Possible household: Fay thinks these are different people who share an address or a phone, like a family or a shared flat. The fix is to link them, not merge them.

Telling the two apart is the whole job, which is why every card lays out its evidence before it offers you a button.

How Fay finds them

Nothing here is typed in by hand. Every night at around 3 a.m. Fay looks across your kept people, groups them by a normalized address fingerprint and a normalized phone number, so “12 Elm St.” and “12 Elm Street” land together, and scores every pair it finds inside a group. Four rules produce a suggestion:

  • Same address: with no conflicting email and a name similarity of 0.5 or higher, the pair is flagged a Possible duplicate and scored 0.90.
  • Same address: where the two also share a surname (a Same last name signal), a Possible household, scored 0.80.
  • Same address: on its own, a Possible household, scored 0.50.
  • Same phone: a Possible household, scored 0.60.

Confidence & signals

You never have to read a raw number. The score becomes a plain-language badge in the card’s corner:

  • Very likely: a score of 0.80 or higher.
  • Likely: a score of 0.50 or higher.
  • Possible: anything below that.

Hover the badge to see the exact score. Underneath, the card shows the raw signals that earned it, each as its own amber pill, Same address, Same last name, or Same phone, beside a neutral Name similarity 0.xx that reports how close the two names are on a 0-to-1 scale. The pills are the “why” behind the confidence badge.

The comparison grid

Every card puts the two people side by side in a four-row grid, Name, Email, Phone, and Address, so the field that matches, and the field that doesn’t, is obvious at a glance.

Under each name Fay prints a one-line summary of that person’s history, so you can tell a full record from a thin one before you touch anything. It reads like 3 gifts · $1,240 lifetime · 11 events · Member: the number of gifts, lifetime giving, events attended, and any roles held. When one side of a pair carries all the history and the other is nearly empty, the empty one is almost always the duplicate to merge away.

Deciding: merge, link, or dismiss

Fay never merges on its own. Each card offers three decisions.

Keep one, merge the other

The mirrored pair of buttons, on a card for “Ada” and “Ben” they read Keep Ada, merge Ben and Keep Ben, merge Ada, is how you resolve a duplicate. Merging re-parents all of the merged person’s history, roles, gifts, contacts, tickets, and even the match suggestions they were part of, onto the one you keep, then tombstones the loser so it vanishes from every list. Nothing is ever hard-deleted, and Fay writes a system contact onto the survivor noting exactly what was merged.

Link as household

When the two really are different people who simply share a roof, roommates, a couple, a parent and an adult child, choose Link as household. Both records survive untouched and are grouped into a household, so letters, receipts, and giving tiers roll up together.

Not related

Not related dismisses the pair. Fay remembers the decision and will never suggest that pairing again, which is how the queue stays short instead of resurfacing the same two people every night.

Every one of these buttons opens a confirmation that spells out the consequence in plain words before anything happens. A merge in particular is easy to trigger and impossible to fully undo, so the dialog names both people and what will move. When the last pair has been handled the queue shows its empty state, “No matches to review” over “Every possible duplicate and household has been resolved.” For the wider story of keeping records tidy, see Search, filters & archiving.

Maria’s day

Saturday’s concert is imported and reconciled, and before the coffee cools Maria opens the Matches tab to clear whatever the overnight run turned up. Top of the queue: Peg Whitfield and Margaret Whitfield, tagged Very likely with a Same address pill. The two context lines settle it, “Margaret” carries eleven concerts and every gift going back three seasons, while “Peg” is a single ticket bought on Saturday. Same person, two spellings.

Maria clicks Keep Margaret, merge Peg. The confirmation names both records and warns that Peg’s history will move onto Margaret; she confirms, and the two collapse into one clean record, eleven concerts, every gift intact, and a line in Margaret’s contact log recording the merge. Margaret is the “prefers to be called Peg” from her own notes; the merge finally makes the record match the person.

The next card is a Likely Possible household: two people at one address with different surnames, roommates who each bought a ticket, not one person entered twice. Maria chooses Link as household rather than merging, so both keep their own record while their letters and receipts roll up together. Two cards, two right calls, and the People list is that much cleaner before her second sip.