Guide

Using Live Chat for Upsell and Cross-Sell

4 minute read · Updated August 14, 2026

The best upsell doesn't feel like one

Upselling in chat has a bad reputation because it is usually done badly — pushing a bigger plan at someone who just wants help. Done well, it is indistinguishable from good service: the customer wanted an outcome, and the add-on genuinely gets them there faster. The goal is not to sell more; it is to solve better, and sometimes solving better costs a little more.

Spot the genuine opportunity

A real upsell opportunity announces itself in what the customer says: they are hitting a limit, describing a need the bigger option fills, or asking about something the add-on provides. That is the moment a recommendation is welcome. A customer who is just troubleshooting a bug is not that moment — reading the difference is the entire skill.

Recommend, don't push

The line between helpful and pushy is honesty and a graceful exit. Explain why the option fits their stated need, give the real trade-off, and make “no thanks” completely comfortable. A recommendation the customer can easily decline builds trust; one they feel cornered by burns it, and a burned customer costs far more than the upsell was worth.

Know when NOT to sell

Sometimes the right move is to sell nothing — when the customer is frustrated, when the cheaper option genuinely suits them, when they came in with a problem and not a purchase. An agent who tells someone “honestly, the plan you have is the right one” earns a loyalty no upsell can buy. Restraint is part of selling well.

How MyLiveChat fits

MyLiveChat lets you recognize returning and higher-intent visitors, route them to the right person, and keep canned product explainers handy so a recommendation is accurate and fast. The judgment of when to recommend — and when to hold back — stays with your agents; the tool just makes the helpful version easy to deliver.

Earn the right to suggest anything

The difference between a helpful suggestion and an unwelcome pitch is almost entirely sequencing. A recommendation offered before the customer’s actual question is resolved reads as an agenda; the same recommendation after it reads as expertise.

So the order is not negotiable: solve the thing they came for, confirm it is solved, and only then, if it genuinely fits, mention the thing that would help next. If the original problem is still open, there is no upsell — there is just an interruption.

This is also the honest test of whether a suggestion is any good. If it would not survive being made after you have already helped and have nothing left to gain from the conversation, it was never a recommendation.

Suggest from evidence, not from a target

The suggestions that work come out of something the customer said. They were asking about a limit they keep hitting, a feature they assumed existed, or a workaround they have built by hand. Each of those is a specific, observed reason to mention something else.

  • They hit a ceiling. “You have mentioned running out of room twice now — the next tier up removes that limit entirely. Worth a look, or would you rather stay where you are for now?”
  • They are doing it manually. “You do not have to export those by hand — that is what the scheduled reports do.”
  • They asked for something adjacent. A question about one capability often reveals they do not know a related one exists.

Notice that none of these start from what you want to sell. Suggestions generated from a quota rather than from evidence are the ones customers can feel, and they cost more trust than the occasional sale is worth.

Make it easy to say no

Counter-intuitively, an offer that is easy to decline converts better than one that is not — because the customer can consider it without first bracing for a fight.

Two habits do most of this. Attach the exit to the offer itself: “happy to leave it as is” or “no rush on this at all.” And accept the first no completely. One clean suggestion respects the customer; the second attempt after a decline converts a helpful conversation into a sales call, and it is remembered that way.

Then move on visibly. “No problem — anything else I can help with on the original issue?” signals the topic is genuinely closed rather than paused.

Where the context comes from

Good suggestions depend on knowing something about who you are talking to, and most of that is already in front of you: the current conversation, the pre-chat form, and past transcripts showing what this customer has asked about before. A customer who has now raised the same limit three times across three chats is telling you something clearer than any prompt could.

Two boundaries worth keeping. Do not let the widget become the place where account changes get made on the strength of a chat conversation alone — a public chat widget is not an authentication channel, and plan or billing changes belong behind a real login. And keep the suggestion in the conversation rather than following it up through channels the customer did not agree to.

Measure whether it is helping or eroding

The metric most teams reach for — revenue attributed to chat — is the one most likely to mislead, because it counts the sales and not the cost. Track it alongside two things that would show damage: satisfaction on chats where a suggestion was made, and the rate at which suggested upgrades are later reversed or refunded.

A healthy programme shows suggestions on a minority of chats, satisfaction unchanged between chats with and without them, and few reversals. Suggestions on most chats, or satisfaction noticeably lower where they occur, means the team is pitching rather than advising — and that is worth catching early, while it is still a habit rather than a reputation.

Put it into practice

MyLiveChat is free forever for one agent, with unlimited chats and the embed code ready in about a minute.

Free forever for 1 agent

Give every visitor an instant way to reach you.

Launch live chat, connect your knowledge base, and add AI answers when you are ready. No credit card, no trial clock.