ecommerce
By Saurav | Founder of saavos | Building in public toward $10k MRR
[!TLDR] Ecommerce stores lose 2–4% of revenue to unanswered pre-purchase questions. A chatbot trained on your product specs, shipping policy, and return window deflects those questions 24/7 and converts uncertain browsers into buyers. Setup takes under 30 minutes. ROI is positive by week three for stores doing 50+ orders per month. The difference between a chatbot that converts and one that doesn't: training on factual specs instead of marketing copy.
Most cart abandonment analysis focuses on price and friction at checkout. But a significant chunk — probably 20–30% in my observation across ecommerce builds — abandons before they even add to cart. Not because the price is wrong. Because they had a question and couldn't get an answer fast enough.
"Does this fit a king-size mattress?" "Will this ship in time for the 15th?" "Is this the same material as the old version?"
These are answerable questions. Your FAQ probably covers most of them. But the FAQ is buried in the footer, the visitor doesn't want to leave the product page, and it's 11pm so there's nobody in your chat.
That's the gap a pre-purchase chatbot fills. Not magic — just availability.
A pre-purchase question is anything a visitor asks before deciding to buy. The categories are predictable:
Fit and compatibility. "Will this work with X?" "What sizes does this come in?" "Is this compatible with [device/model/standard]?" These are high-anxiety questions because a wrong answer means a return. Visitors who can't get a confident answer often bail rather than risk it.
Shipping and timing. "How long is delivery?" "Do you ship to [country]?" "Can I get this by [date]?" Urgency-driven buyers abandon if they can't confirm timing instantly. A chatbot that can answer "we typically ship within 2 business days and deliver in 5–7 via USPS" converts those buyers.
Return and risk. "What's the return window?" "Can I exchange for a different size?" "What if it doesn't work?" First-time buyers from unfamiliar stores need to know the safety net exists. A chatbot that recites your real return policy removes the risk objection.
Comparison and differentiation. "How is this different from the [cheaper version]?" "Why should I get this over a competitor's?" These are high-intent questions from buyers who are close to committing but want confirmation.
Most ecommerce chatbots fail because they're trained on product description copy, which is written for SEO and search discovery, not for answering these specific questions. The fix is structural.
This is the most important decision you'll make, and it's worth spending 20 minutes getting right before you launch anything.
Good sources for pre-purchase ecommerce chatbot:
What to exclude:
A 3,000-word training set of pure factual content dramatically outperforms a 30,000-word training set that includes marketing noise. Narrower and more accurate wins every time.
You can get a working pre-purchase chatbot live in 30 minutes with a managed platform like saavos. Here's the exact sequence:
Gather your sources. Download your FAQ page as text. Copy your shipping policy. Export your sizing guide or compatibility table. Put everything in one document. Total target: 2,000–5,000 words of clean factual content.
Create the bot and set the persona. Name it something recognizable ("the [Brand] assistant" is fine — no need to name it Sophie or make it seem human). Write a greeting that sets expectations: "Hi — I can answer questions about sizing, shipping, returns, and product details. What can I help with?"
Configure the fallback. This is non-negotiable. When the bot can't answer, it should route to something real: "I don't have that in our product info. Email us at support@yourbrand.com — we typically reply within 4 hours." A bot that says "I'm sorry, I can't help" and stops there creates more frustration than no bot at all.
Test against your real questions. Ask the five questions your support inbox answers every week. If the bot gets any of them wrong, find the gap in the source content, fill it, and re-train. Takes 10 minutes.
Deploy on high-intent pages. Product pages, cart page, and shipping/FAQ page. Homepage is worth covering but lower priority — visitors there are still orienting, not evaluating.
Let's make this concrete for a store doing $20K/month in revenue (roughly 80 orders).
If 10% of visitors who don't convert have an unanswered pre-purchase question, and your site gets 2,000 visitors a month, that's 200 visitors with answerable questions. A chatbot that answers 60% of those questions and converts 30% of answered visitors into purchases generates ~36 additional purchases. At an average order value of $50, that's $1,800 in recovered revenue per month. Against a $25/month chatbot subscription, the math is obvious.
I'm using conservative numbers. Ecommerce stores with good FAQs and clear policies have reported 4–6% conversion lift from pre-purchase chatbots. I don't have enough data from saavos users to quote that for us yet, but the unit economics hold even at lower conversion rates.
There are real limits to what a pre-purchase chatbot should do. Understanding them avoids building something that frustrates customers.
Real-time inventory. "Is this in stock in medium?" requires a live API call to your inventory system. Most managed chatbot platforms don't do this without custom integration. If inventory is volatile and out-of-stock questions are common, either build a lightweight inventory feed into the bot or handle stock questions as a fallback-to-human case.
Order status. "Where's my order?" is a post-purchase question. Keep it out of a pre-purchase bot. It creates expectations the bot can't meet and muddies your training data.
Subjective fit. "Do you think I'd like this?" is a human question. The bot can recite specs and let the customer decide, but it shouldn't pretend to have taste. Route these to the fallback with a human recommendation.
Price negotiation. "Can I get a discount?" Should route directly to a human. Training the bot to say no or to apply discount codes opens a can of edge cases.
The rule: the bot handles fact-based questions, humans handle judgment calls.
If you're on Shopify, a few practical details:
The native Shopify Inbox app is free and handles some of this with limited AI. It works fine for basic automated replies but isn't good at RAG-based retrieval from your actual content. Worth starting there if you want zero setup, but you'll hit its limits fast.
For a more capable bot, use an external platform with Shopify integration (including saavos). Embed via the theme footer or a Shopify app block. The setup is paste one script tag — under 5 minutes.
Product pages are your most valuable placement. If you can only put the chatbot on one surface, put it there, triggered on time-on-page of 30 seconds (you can configure this in most platforms, or use a scroll-depth trigger).
A pre-purchase chatbot doesn't just convert — it teaches you. The conversation logs are a real-time feed of what your customers actually want to know, phrased in their own words. That's better market research than any survey.
After 30 days, read through the logs. Find the five questions that came up most that weren't in your original FAQ. Add them. Update the training. The bot gets better, and so does your product copy.
That loop — conversations to insights to content improvement — is where the compounding value lives.
If you're on Shopify specifically, the Shopify Chatbot Setup guide for 2026 covers the platform-specific installation steps and the three embed options in more detail. And if you're also thinking about capturing leads from visitors who research but don't buy — the chatbot can do both — Lead Capture Chatbot: Turn Site Visitors Into Qualified Emails covers the qualification sequence that converts.
Start free on saavos — you can have a pre-purchase chatbot trained on your product content in under 5 minutes.
Get the next post in your inbox
Honest writing on building, embedding, and shipping AI chatbots. No spam. Unsubscribe anytime.
Four categories cover 80% of pre-purchase volume: fit and compatibility ("Will this work with X?"), shipping and timing ("Can I get this by [date]?"), returns and risk ("What is your return policy?"), and product comparison ("How is this different from the cheaper version?"). A chatbot trained on your FAQ, shipping policy, and structured product specs handles all four confidently. The categories that require a human: real-time inventory lookups, subjective fit recommendations, and price negotiation.
Ecommerce stores with well-trained pre-purchase chatbots report 2–6% conversion lift in published case studies; saavos has not yet measured this in-house. The mechanism is simple: a meaningful share of would-be buyers abandons before adding to cart because they had an unanswered question and did not want to risk a return. A chatbot that answers those questions 24/7 — including off-hours and weekends — converts those visitors. At 2% lift on a store with 2,000 monthly visitors and a $50 average order value, that is 40 additional orders, or $2,000/month in recovered revenue.
Training on product description copy instead of factual content. Product descriptions are written for SEO and discovery — they emphasize benefits and brand voice, not dimensions, compatibility, or return windows. A chatbot trained on marketing copy sounds confident but gives vague or wrong answers to specific pre-purchase questions. Train tightly on your FAQ, verbatim policies, and structured spec sheets. A 3,000-word factual training set outperforms a 30,000-word mixed set that includes marketing noise.
Prioritize product pages, the cart page, and the shipping/returns FAQ page. These are the surfaces where pre-purchase questions are most likely. The homepage has mixed-intent visitors and is worth covering but is the lowest-ROI surface. If your platform lets you configure trigger conditions, set the chatbot to appear on product pages at 30 seconds on-page or 50% scroll depth — this catches buyers who are actively evaluating, not just landing.
Most managed chatbot platforms do not access live inventory data without custom API integration. A well-configured bot routes inventory questions to the fallback: "For current stock availability, please check the product page or email us at support@yourbrand.com — we respond within 4 hours." This is the honest, correct answer. Avoid training the bot to make inventory claims it cannot verify — a wrong in-stock answer that turns into a fulfillment failure destroys more trust than a simple handoff to a human.
Builds tools for solopreneurs and small SaaS teams who don't have an afternoon to spare.
Paste your URL. Train your bot. Drop one script tag. No credit card.