Set up lead capture without friction

Setting up lead capture without friction

Lead capture is optional: by design

Fynd is an AI product advisor widget for webshops. Shoppers ask questions; the bot retrieves answers from catalog, sitemap, feeds, Catalog API, and custom FAQs. Lead capture is an optional layer on top of that advice: a short form where someone leaves contact details when a human or personal follow-up step makes sense.

Lead capture is enabled or disabled per bot via collect_leads in bot configuration. Conversations belong to the shop; submitted leads link to shop, bot, session, and optionally conversation_id. Analytics records a lead_submit event so you can read lead rate alongside opens and sessions.

This article helps you decide when to turn lead capture on, how to keep timing and copy frictionless, which fields to keep minimal, how to measure, how to connect unanswered questions, and what privacy-minded wording fits, plus when to leave it off.

When lead capture makes sense

Turn lead capture on when there is a clear next step that helps the shopper, not when you want to "capture" every conversation.

Good moments

• The shopper asks for something you will send later: a comparison by email, a quote, sizing advice your team handles manually.
• A product is (temporarily) unavailable and someone wants a notification.
• The bot gives useful advice but cannot make a final decision, for example B2B configurations, custom work, or project furnishing.
• Your unanswered queue shows shoppers asking for contact after policy or complexity stops the bot, lead capture as a safe exit after explanation.

Poor moments

• Before the first useful reply: then the widget feels like a popup form.
• For simple product questions already covered by catalog and FAQs.
• When you have no process to follow up on leads, a form without follow-through erodes trust.
• When checkout or support chat already works and lead capture only duplicates effort.

The useful question is not "can we ask for an email?" but "does leaving an address help the shopper choose better or get help sooner?"

Timing: value first, then the form

Friction rarely lives in field count alone; it lives in when you ask for data.

Recommended order in the conversation

1. Greeting and disclaimer: clear that this is an AI advisor.
2. Shopper asks a question; bot answers from catalog or knowledge sources.
3. Optionally: product cards or follow-up questions.
4. Only then lead capture, when the bot (or you via copy) offers handoff: "Want us to email this comparison?" or "Leave your details and the team will contact you about custom sizing."

In widget configuration you steer this partly via welcome_message, disclaimer, handoff_message, and any persona copy. handoff_message is what shoppers see after a successful submit, by default something like "Thanks. We will contact you as soon as possible." Adjust that line to your real service level; do not promise a turnaround operations cannot meet.

Triggers to avoid

• Lead capture as soon as the widget opens.
• Required email before the first question can be sent.
• Showing the form repeatedly in the same conversation after dismissal.

Test in the playground: run three realistic shopper flows. Only when at least one flow naturally leads to handoff is collect_leads ready for production.

Minimal fields: email first

The lead submit API requires email; name, phone, and message are optional. Keep the form aligned with what your backend actually needs.

Rule: every extra field needs a clear operational reason. Phone without a call process is noise. Requiring a message when the conversation already holds context is duplicate work for the shopper.

Store context via conversation_id and page_url (passed automatically by the widget where configured) instead of long forms.

Copy that keeps trust

Lead capture copy must stay consistent with disclaimer and welcome_message. Shoppers should understand:

• What they get after submit (comparison, callback, stock alert).
• Who contacts them (team, not a "personal advisor" if the flow is bot-first).
• Why email is needed (one sentence).

Example: after useful advice

I can show you three options in this chat. Want the comparison by email 
too? Leave your address. We will only send this summary, not a 
newsletter.

Example: gap / handoff

For on-site custom work our team needs to take a look. Leave your 
email. We reply on business days using the address you enter here.

Example: handoff_message after submit

Thanks. We received your details and will follow up on your question 
about [topic from conversation].

Avoid:

• "Don't miss this deal", that is marketing, not advice handoff.
• "We will call within 5 minutes" unless that is truly how you operate.
• Vague promises of human chat when nobody connects.

Align copy with support and sales so tone matches email and phone.

Measuring lead rate without misleading yourself

Analytics shows Leads with the percentage of sessions (lead rate). That is your primary metric, not raw lead counts alone.

How to read lead rate

• Lead rate = leads / sessions over the same period (default thirty days).
• Compare lead rate with engagement (messages/opens) and answer gap (unanswered/messages).
• Leads rise while engagement falls? You may ask for data too early and shoppers stop chatting.
• Leads rise while answer gap falls? Handoff may catch gaps; check whether you also fix FAQs or catalog so the bot answers itself next time.

Secondary signals

• Leads linked to conversation_id: open the conversation: was handoff logical?
• lead_submit events in the event mix: ratio to message_sent and unanswered.
• Status in your lead overview (new, followed up): lead-to-sale conversion you measure outside Fynd unless you tie purchase events.

Set a band for yourself, not an absolute target from external benchmarks. After two weeks live: note lead rate, engagement, and answer gap in one line per week. Adjust copy or timing with one change per iteration.

Connecting to unanswered questions

The unanswered queue and lead capture work together but do not replace each other.

Workflow

1. Weekly analytics: sort unanswered questions by theme.
2. Question type A, missing policy → resolve with custom FAQ; no lead capture needed.
3. Question type B, bot may explain but not close → improve handoff copy; collect_leads stays on.
4. Question type C, spam or off-topic → dismiss; lead capture irrelevant.

When the same unanswered wording appears often and shoppers drop off, consider:

• Fix knowledge first (catalog or FAQ).
• If human follow-up stays structurally needed, explicit handoff in persona/system prompt: "Offer lead capture when …"

Document in your team log: "Lead #123 over unanswered on refurbished warranty; FAQ added, handoff copy softened."

That prevents lead capture from becoming a permanent bandage on knowledge gaps you can close.

Privacy-minded wording

Lead capture collects personal data. Be transparent without inventing legal claims your team cannot support.

Practical guidelines

• Link to your privacy policy on the site (footer or disclaimer), not to non-existent "GDPR certifications."
• Briefly state why you use email: follow-up on this question, not generic marketing, unless you actually send marketing too; say so honestly and separate consent if your shop requires it.
• Do not ask for data the advisor never needs (national ID, payment card, passwords). System prompts and FAQs should steer shoppers away from sharing those.
• Keep only fields your processes use; consider periodic review of open leads holding personal data.

Example disclaimer addition (if it fits your shop)

If you leave contact details, we use them to respond to your question. 
More information is in our privacy policy.

Have legal or a privacy owner approve final copy if you serve regulated sectors or multiple countries. This article is not legal advice.

When to keep lead capture off

Keep collect_leads off in these situations:

• Self-service checkout is the goal: every handoff is friction toward cart.
• Your team cannot follow up within a reasonable time: better no form than a broken promise.
• Assortment and FAQs cover almost everything: focus on answer gap and product clicks, not email collection.
• You already run a CRM popup or exit intent: double capture annoys and pollutes metrics.
• Trial week 1: embed, catalog, and unanswered loop stable first; lead capture in week 2 when engagement is healthy.

You can toggle lead capture seasonally (for example on during peak advice, off in quiet months) as long as you document the change so lead-rate comparisons stay explainable.

Setting up in Fynd: step plan

1. Open bot settings for the default bot or the bot on your storefront.
2. Enable collect_leads only after playground tests with realistic flows.
3. Write handoff_message that matches your follow-up time.
4. Check disclaimer and welcome for consistency (AI advisor, not human).
5. Embed live; exclude playground from reporting.
6. Week 1: monitor engagement, answer gap, product clicks, no major lead copy experiments yet.
7. Week 2+: read lead rate alongside unanswered queue; resolve FAQs where possible; tune handoff copy.

Optionally set handoff_email if your team wants internal notifications for new leads; ensure the mailbox is staffed.

Checklist before go-live with lead capture

• Clear shopper value before the form tested
• Email only required; extra fields justified
• handoff_message realistic and consistent with support
• Privacy policy reachable; no overstated promises
• Process for following up new leads defined
• Analytics: lead rate baseline noted
• Unanswered themes split: FAQ fix vs handoff

Lead capture should finish the advice conversation, not interrupt it. With the right timing, minimal fields, and weekly ties to analytics, it aligns with what Fynd already measures: opens, messages, gaps, and follow-up contact, without making shoppers choose a form before the bot has done something useful.