Article summary
Create a separate language-routing table for Cameroon +237 customers instead of substituting demographics, names or avatars for preference.
Support language for Cameroon +237 customers cannot be derived from gender, age, names or avatars. If demographic screening is being used for routing, the data team should split the two purposes into separate tables first.
What belongs in a language-routing table
| Field | Meaning |
|---|---|
| preferred_language | Explicit customer choice |
| preference_source | Form or conversation |
| confirmed_at | Confirmation time |
| fallback | Neutral entry when unknown |
Keep the profile table independent
If gender_observed and age_band_observed have a valid purpose, they still need unknown states, observation time and use limits. They do not join the language decision rule.
Use +237 only for number normalisation
The code does not prove language, region, citizenship or identity. National-looking strings with weak provenance remain country_unknown.
Ask when preference is missing
Offer a short, switchable welcome so the customer chooses. Do not send several long language blocks or infer from a name.
Measure routing errors, not guessed accuracy
Record language switches, requests for another agent and abandoned conversations. Direct feedback is more useful than a statistical correlation between profiles and language.
Keep language and profiles out of TXT
One phone per line is enough, with an internal row_key linking the preference table. Returned Excel does not rewrite preferred_language automatically.
Avatar visibility does not change priority
Customers with blank avatars or unknown profiles receive the same language choice and service entry, preventing systematic disadvantage to less-visible groups.
Make preference updateable
A new customer choice supersedes the old one while source and time remain recorded. Long inactivity may return the value to unknown rather than preserve a permanent label.
The Cameroon outcome
Ask for language when language is needed; prove a separate purpose when demographics are needed. Two independent tables are more reliable than one field expected to solve everything.
