Revenue

Sizing, shipping and returns decide the basket, and none of them are on the product page.

Ecommerce objections are unusually predictable. Three topics account for most hesitation, all three are documented somewhere on the site, and all three are one click away from the moment they matter.

Before and after on a revenue siteThe same visitor question, handled by the current site and then by an assistant grounded in that site's own content.BEFOREShopper hesitates on sizingat the product pageOpens the sizing guide in anew tab, or does notLoses the thread, compareselsewhereAFTERShopper asks about sizingwithout leaving the productpageAssistant answers from thesizing guide and cites itDelivery and returnsanswered in the sameconversation
  • Private launchOnboarding selected teams now
  • Built by CreoglyphSeven client engagements behind the product

One click away is one click too many

A shopper on a product page wonders whether the size runs small. The answer is in a sizing guide, linked in the footer. Getting there means leaving the product page, which means losing the basket state and the momentum that came with it.

The same applies to delivery timing and to returns. Each has a dedicated page. Each page is well written. Each is in the wrong place, because the decision happens on the product page and the answer lives elsewhere.

The measurable effect is an exit from a page with high purchase intent, which reads in analytics as a product page problem. It is not. It is a proximity problem.

What changes

Today

  1. Shopper hesitates on sizing at the product page
  2. Opens the sizing guide in a new tab, or does not
  3. Loses the thread, compares elsewhere
  4. Basket abandoned, no record of why
  5. The sizing guide is judged useful because it gets traffic

With an assistant

  1. Shopper asks about sizing without leaving the product page
  2. Assistant answers from the sizing guide and cites it
  3. Delivery and returns answered in the same conversation
  4. Objections resolved at the point of decision
  5. The topics that keep coming up are logged for the product page

Answering an objection without leaving the page

Index the policy pages

Sizing, delivery, returns, care instructions, stock policy. Small and high value. Product descriptions are usually a worse source than the policy pages, because they are written to sell rather than to answer.

Answer in place

The widget opens over the product page. The basket persists, the context persists, and the answer arrives with the product still on screen.

Cite, because trust is the objection

On a returns question the shopper is assessing risk. An answer that names the policy page is a different object from an answer that asserts.

Refuse on stock and order status

Anything requiring a live lookup is outside what a content trained assistant can do. Guessing at stock is worse than saying you cannot check.

The three ways a conversation leaves the flowMost conversations end in a grounded answer. The exits are a logged refusal, an escalation to a person, or a capture before continuing.IN THE FLOWQuestionRetrieveGrounded answermost conversationsEXITSRefuse, log gapEscalateCapture, then continue

Where the widget should not appear

This is the ecommerce specific decision, and getting it wrong costs more than the assistant earns.

Checkout: off

Checkout has one intended action, the launcher competes with it for the same thumb on mobile, and a shopper who opens chat during payment is a shopper who has stopped paying.

Basket: allow, but disable at payment

Objections do surface there, but so does distraction. Allowing it on the basket and turning it off at the payment step is the defensible position.

Product and category: always on

That is where objections form and where the answer is furthest away.

This is a conversion decision, not a styling one, and it is worth measuring rather than assuming. The counterfactual for chat on checkout is not a lost answer, it is a completed purchase.

Where the data goes

The assistant answers from published content only. It cannot check live stock, look up an order, or see a customer account, and those should be escalation triggers rather than attempted answers. Security and data boundaries.

Questions

No. It answers from published pages and has no connection to inventory. A stock question should escalate rather than be guessed at, because a wrong stock answer produces a cancelled order and a refund.

No. Order status needs an account lookup, which is outside what a content trained assistant does. Route those to your existing order tracking.

Usually not. Checkout has one intended action and the launcher competes with it. Allowing it on the basket and disabling it at the payment step is the defensible position.

It can describe products your content describes. It is not a recommendation engine and does not rank by margin or stock, so treat it as an objection handler rather than a merchandising tool.

It is a script embed, so it works anywhere you can add one. There is no deeper platform integration built, and we are not going to describe an embed as an integration.

Returns. It is the highest anxiety topic, the answer is unambiguous, and it is the one shoppers most often leave to check.

Signal

Is this your situation?

Access is by waitlist, demo request or public-content pilot. Describe the question your visitors keep asking and we will tell you honestly whether it fits.