You added a chatbot two months ago. Is it working? "It feels busy" isn't an answer, and neither is the raw count of conversations. The number that tells you whether the bot earns its place is its conversion rate, and most people either never define it or measure the wrong version. Let's fix that.
Decide what a conversion even is
Conversion rate is meaningless until you name the goal. A bot on a plumber's site and a bot on a software pricing page are chasing different outcomes, so they need different definitions.
Pick the one action that matters most for your business. Common goals:
- A captured lead with contact details
- A booked call, demo, or reservation
- A completed sale or checkout
- A qualified lead that meets your criteria, not just any email
Choose a single primary goal to start. You can track secondary ones later, but a bot optimized for everything is optimized for nothing. Once you've named the goal, the rate is simple arithmetic: conversations that reached the goal, divided by total conversations, times a hundred.
One caution on the denominator. Decide whether "total conversations" means every visitor who opened the chat, or only those who sent at least one message. The second is usually more honest, because a widget that pops open and gets ignored shouldn't count as a failed conversion. Just pick one definition and stick with it so your numbers stay comparable.
Set a baseline before you judge anything
The most common mistake is looking at a conversion rate in isolation and deciding it's good or bad. Twelve percent means nothing on its own. Is that strong for your goal and traffic, or weak? You can't know without a point of comparison, and the only fair comparison is your own bot over time.
So measure quietly for the first few weeks and treat that as your baseline. Don't chase someone else's benchmark from a blog; their goal, traffic, and industry aren't yours. Once you have a baseline, every change you make has something to beat. Did rewording the opening question lift the rate from your baseline? Then it worked. That's the whole loop.
The metrics that keep conversion rate honest
Conversion rate is the headline, but on its own it can lie to you. A bot could show a decent conversion rate while quietly frustrating most of the people who use it. So track a small set of companions.
| Metric | What it tells you |
|---|---|
| Conversion rate | Share of chats that hit your goal |
| Engagement rate | Share of visitors who actually chat |
| Containment rate | Share of questions resolved without a human |
| Escalation rate | Share handed off to a person |
| Answer quality | Whether answers were correct and complete |
Read them together. A high conversion rate with a rock-bottom engagement rate means the bot converts a tiny, self-selected few and everyone else bounces. A high containment rate with a low conversion rate might mean the bot is great at answering support questions but never nudges toward the goal. The picture only makes sense in combination.
A worked example
Take Summit Gear Co., a small outdoor retailer whose goal is captured leads for their gear-fitting service.
In month one, the bot had 400 conversations and produced 32 leads. That's an 8 percent conversion rate, and it's their baseline. Fine, but they wanted more. Reading transcripts, they noticed the bot answered product questions well but rarely mentioned the free fitting service, so most chats ended with a satisfied customer and no lead.
They made one change: after answering a gear question, the bot now offers, "Want a free fitting session to make sure it's the right size?" Month two: 420 conversations, 54 leads. That's about 13 percent, up from their 8 percent baseline.
Same traffic, same bot, one deliberate change measured against a known number. Without the baseline, the jump would've been invisible. With it, they knew exactly which tweak paid off, and where to poke next.
Don't let a big number fool you
Some chatbot dashboards lead with totals that look impressive and mean little. A count of "messages handled" or "conversations started" climbs no matter what, so it always feels like progress. It isn't a conversion rate, and it won't tell you whether the bot moves the needle on anything you care about.
Anchor on rates, not raw counts. A rate has a denominator, which forces an honest comparison. Two thousand conversations sounds terrific until you learn only twelve turned into a lead. The rate would have told you that in a single number. So when a report hands you a big total, ask what it's a fraction of before you celebrate it.
Connect the goal to your real numbers
A conversion rate trapped inside the chatbot dashboard is only half useful. The point of measuring is to compare the bot against your other channels and to see whether those leads become revenue. That means getting the goal event out of the widget and into the tools where you track everything else.
At minimum, make sure a converted conversation shows up in your analytics or CRM as an event, tagged as coming from the bot. Then you can answer the questions that actually matter: do chatbot leads close at a similar rate to other leads, or better? Is the bot cheaper per lead than your ads? Those comparisons are where "the bot feels busy" turns into "the bot brings in leads at a third the cost."
In SpideyChat you can see conversations and the leads they produce, and connect that goal to your analytics so the bot's numbers sit next to your other channels instead of in a silo.
Turn the number into a habit
Tracking conversion rate isn't a one-time report. It's a rhythm. Here's a workable one:
- Define your single primary goal and how you count a conversation.
- Measure for a few weeks to set a baseline. Don't touch anything yet.
- Read transcripts to find where conversions leak, like Summit Gear's missing fitting offer.
- Change one thing at a time so you know what moved the number.
- Compare against your baseline and keep what wins.
- Recheck monthly, and update the baseline as the bot improves.
The reason to bother with all this is simple. A chatbot is easy to add and easy to forget, and a forgotten bot drifts toward mediocre. A tracked one gets better, because every month you're looking at a real number and asking how to beat it. That's the difference between a bot that "feels busy" and one you can prove is pulling its weight.