Qualification is a routing decision, not a judgment about a person. Tags and attributes should describe the customer’s stated need and the business’s ability to serve it. They should never infer wealth, religion, ethnicity, gender, health, or another sensitive characteristic.
After this lesson, you can create a small qualification schema, deterministic scoring rules, and an audit sample that prevents automation from silently excluding valid customers.
Separate Tags, Attributes, and Events
- Attribute: current structured fact, such as
city=lahoreorinterest=product_a. - Tag: operational grouping, such as
needs_human_quote. - Event: timestamped action, such as
qualification_completed.
Do not create three versions of the same fact. WATI’s current help material supports contact attributes, tags, filters, and chatbot attribute updates, though feature availability can depend on plan.
Define the Minimum Qualification Rule
For each question, write the business reason and allowed values:
| Field | Why needed | Allowed values | Unknown behavior |
|---|---|---|---|
| service_code | selects trained team | catalog IDs | show menu/human |
| city | checks coverage | controlled list | manual coverage review |
| timing | prioritizes legitimate urgency | now / 7 days / later | ask once, then unknown |
| budget_band | checks offer fit if necessary | published bands | prefer quote/human |
| consent_marketing | controls future offers | true / false | false |
Use eligibility rules before scores. If a service is unavailable in an area, mark coverage_review or provide alternatives; do not call the person a “bad lead.”
If a score is useful, make every point explainable:
2 points: supported service selected
2 points: supported city confirmed
1 point: timing within seven days
0 points: unknown answer
route ≥4 to sales; otherwise nurture only with consent or human review
The score is a queue aid, not evidence of ability to pay or likelihood to buy.
Worked Example
An Islamabad bookkeeping firm serves registered small businesses and freelancers. Its flow records service, city, business type selected by the customer, urgency, and preferred consultation time. A lead requesting a service outside scope receives a truthful explanation and a human-review option.
One contact selects payroll setup, Islamabad, registered company, within seven days. The system sets attributes and adds sales_priority, then creates an event containing the rule version. Another says “not sure” to business type. The bot does not invent a category; it adds needs_discovery and routes to a consultation queue. A monthly audit compares 30 automated classifications with human review and records false routing.
Failure Cases to Diagnose
- Tags accumulate forever: define who removes them and when.
- Free text becomes an uncontrolled category: map to allowed values or preserve
unknown. - Unknown scores as zero and suppresses the lead: route uncertain cases for review.
- Sensitive traits used as proxies: remove the field and review prior decisions.
- Agent can overwrite consent casually: protect the authoritative consent ledger.
- Rule changes without version: attach a version to qualification events.
- High score triggers marketing without permission: operational priority and marketing consent are separate.
🇵🇰 Pakistan Angle
City coverage can be precise without being discriminatory. Ask the customer’s requested service area rather than inferring location from their mobile prefix, name, or language. For nationwide digital services, avoid using city as a quality score merely because clients from certain areas historically purchased more.
Budget questions can feel intrusive. Publish price bands first and allow need recommendation or human help. Quote in PKR, clarify tax and payment terms, and do not imply that a lower-budget lead deserves slower support if the business accepted the enquiry.
Hands-On Exercise
- Define five or fewer qualification fields and their business purposes.
- Create controlled values plus unknown behavior.
- Write deterministic eligibility and optional scoring rules.
- Configure attributes/tags on internal test contacts.
- Manually audit at least ten classifications and log disagreements.
Done means: any operator can explain a routing result from customer-supplied facts and a versioned rule without using sensitive inference.
Completion Rubric
- Attributes, tags, and events have separate roles.
- Qualification uses only necessary customer-supplied facts.
- Unknown values route safely.
- Consent is independent from priority.
- Rules are versioned and explainable.
- A human audit measures routing errors.
Sources
- WATI Help — Updating contact attributes in chatbots
- WATI Help — Contact attributes
- WATI Help — Search and filter chats
- WhatsApp Business — Messaging policy
Key takeaway: qualify from minimal, explicit facts with versioned rules; preserve unknowns and human review instead of letting labels become hidden judgments.