The Configure step is where you tell the Genie how to match. Every knob on this page influences the trade-off between precision (fewer false positives) and recall (finding more matches).
The main decisions
Match profile
A dropdown of preset profiles plus a Custom option. The Genie recommends a profile based on the column profiles of your two files; override if you have context the Genie doesn't.
The available profiles and their specifics are documented in Match profiles. Most common choices:
- Person for customer/contact/lead matching
- Company for B2B account/supplier matching
- Identifier when a reliable unique ID exists in both files
- Contact dedupe when you only have one file
Confidence threshold
A slider from 50 (loose) to 95 (strict). Default is 70. The slider controls the primary match threshold; the review threshold moves in lockstep, always 15 points lower. See Setting the confidence threshold for detailed tuning advice.
Field mapping
If your source and master columns have different names, map them here. The Genie auto-maps columns with identical or very similar names; explicit mapping is needed when names diverge (e.g. First Name in source vs fname in master). See Field mapping.
What the Genie shows you before you hit Run:
- Side-by-side data preview under each mapping row — three source samples next to three master samples, pulled straight from your files. Makes it obvious at a glance when columns carry different kinds of data (source
Region= "Lombardia, Lazio" vs mastercity= "Milano, Roma" — clearly not a match). - Mis-map warnings with Genie-voice explanations. Fires when the samples suggest the columns are semantically different (length, token count, or character mix — numeric vs alpha). Example: "These look like different kinds of data. 'city' contains values much longer than 'Province'. That usually means one is a full text field and the other is a short code."
- Match-impact badges translating your mapping choices into plain English: "1,800 rows won't exact-match on this field" (red, blocking), "3,400 flagged cells — fuzzy still works" (amber, advisory), "230 flagged — no impact (ignored)" (grey). No more surprises at results time.
- Auto-Ignore on unmapped fields. If a source column has no master counterpart picked, the match role locks to Ignore automatically — no silently-contributing-nothing signals.
Match roles
Each mapped pair has a role that tells the engine how to compare the values:
- Exact match — treated as identity-grade evidence. Perfect agreement = instant match. Use for unique IDs (SSN, NPI, EIN, email, phone, custom IDs).
- Fuzzy (Name) — for people's names. Tolerates typos, nicknames, diacritics, hyphenation variants.
- Fuzzy (Address) — for street/city/state. Tolerates abbreviation variance (St ↔ Street, Ave ↔ Avenue).
- Fuzzy (Company) — for organization names. Tolerates suffix variance (Inc ↔ Incorporated, LLC ↔ L.L.C.).
- Fuzzy (Category) — for classification text (specialty, department, cuisine, genre). Agreement boosts the match score; disagreement is a mild penalty, not a veto. Handles "Hematology" ↔ "Hematology-oncology", "Cardiology" ↔ "Cardiologist", multi-language variants.
- ZIP Radius — distance-based postal match; see below.
- Ignore — skip the column entirely.
Combining columns at map time
When your source carries a single full-name column (Full Name, HCP Name, Physician Name) and your master has first and last split across columns, pick "first + last (combined)" from the master dropdown. Two sub-pickers appear where you name the exact first and last columns to concatenate. The engine builds the combined name on the master side at match time — no name-search heuristics, no magic.
The Genie auto-includes a middlename column when your source names average more than 2.4 tokens (e.g. Spanish compound names like "Alberto Eterio Velasco Valdazo").
Optional toggles
Phonetic matching
Default: on for Person, off for Company. Matches names by sound (Smith ↔ Smyth) using phonetic coding. Toggle off for product codes or other data where phonetics add noise. See Phonetic matching.
ZIP / postal radius
Default: off. When on, postal code comparison becomes distance-based — source and master records within a configurable mile radius can match even if their ZIPs don't exactly agree. Useful for data with sparse or imprecise location info. See ZIP radius matching.
One-to-one matching
Default: off. When on, each master row can only match once — preventing the situation where two source rows both claim the same master as their best match. The engine finds the globally optimal assignment. See One-to-one vs one-to-many.
Max candidates per match
Default: 1 (best match only). When raised, the engine returns up to N top candidates per source row in the output — useful for research or review workflows where you want to see what other matches were plausible.
Saving as a custom profile
Once you've tuned settings for your use case, click Save as profile and give it a name. The profile becomes selectable on future Configure steps under the Custom section. Great for recurring workflows (monthly lead match, quarterly supplier reconciliation).
Advancing
The Next button advances to Run and queues the match. You can still abandon the run from there; matching doesn't consume quota until completion.