## TL;DR
Widget placement converts to self-serve when the widget appears at the moment of confusion: inside the billing page for billing questions, next to the complex form, on the error screen. A widget parked in the global corner gets ignored. Pair placement with search-first content: the widget must answer from the help center instantly, not just offer a chat. Placement without good answers is a faster path to the same ticket.

## The query

```text
support widget placement: what converts to self-serve
```

## Use this when

- Deciding where support widgets live in the product
- Widget engagement is low
- Building self-serve strategy
- Reducing ticket volume with placement

## Not for

- Widget visual or UX design
- Help-center information architecture
- Chatbot conversation design
- Marketing popups

## Steps

### 1. Map confusion points from ticket data

Top ticket intents plus the page or feature each comes from. Billing questions cluster on the billing page; setup questions on the onboarding screens. Place widgets at the top three confusion points first.

Expected output: confusion points mapped to pages.

### 2. Put the widget at the confusion, not in the corner

Contextual placement: the billing page widget opens with billing help, the error screen widget offers the error's article. Global corner widgets are fine as a backstop, but contextual placement is what deflects.

Expected output: widgets placed at the top confusion points.

### 3. Make the widget search-first

The widget's first screen is search over the help center, with the top articles for that page pre-suggested. Chat is the fallback, not the headline. A widget that opens straight to "chat with us" is just a ticket form with extra steps.

Expected output: search-first widget with contextual suggestions.

### 4. Measure deflection per placement

Track: widget opens, searches, article views, and tickets avoided (survey or ticket-rate comparison). Per placement, not globally. Kill placements that do not deflect; double down on ones that do.

Expected output: per-placement deflection numbers.

### 5. Iterate on the content, not just the placement

Low deflection usually means the articles are bad, not the placement. Review the top failed searches per placement monthly and write the missing articles. Placement gets them to the door; content gets them through it.

Expected output: a monthly failed-search review.

## Template: the placement plan

```text
PLACEMENT PLAN
Confusion point -> Page -> Widget mode -> Suggested content
Billing questions -> /billing -> contextual, billing articles first -> plan changes, invoices, refunds
Setup friction -> /onboarding/step-2 -> contextual -> setup guide, video
Error confusion -> error screens -> contextual, error-specific -> the error's article
Backstop -> global corner -> search-first -> top 10 articles

Measure per placement: opens, searches, article views, deflection.
Review failed searches monthly; write the missing articles.
```

## Variant phrasings

### where to put help widget

Steps 1 and 2. Map confusion, place contextually.

### self-serve widget strategy

Full sequence. Placement plus content iteration.

### widget not reducing tickets

Steps 4 and 5. Measure per placement, then fix the content.

## Why it works

Self-serve fails on timing: help offered after frustration is ignored, and help offered generically is skimmed. Contextual placement catches the user at the exact moment of confusion with the exact relevant articles. The search-first design respects that most users prefer to solve it themselves if the answer is one search away.

## Edge cases

- Mobile: screen space is tight. Use a single entry point with strong contextual suggestions.
- Logged-out users: they hit different confusion points (pricing, signup). Place accordingly.
- The widget annoys power users: add a dismiss that sticks, and do not re-prompt aggressively.
- Accessibility: the widget must be keyboard-navigable and screen-reader friendly. Test it.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_vWQiaS8-ruNgV8fbIkdt1g