A seller sent me a screenshot last month with real pride in it: 240 direct messages in one week. Her reels were doing well. Her follower graph was going up and to the right. And her sales were flat.
So I asked two questions. How many of those 240 people got an answer? She wasn't sure. How many of them ended up paying for something? She wasn't sure about that either. She could tell me her follower count to the exact digit.
That gap is what I want to talk about. Most Instagram and WhatsApp sellers measure the top of the funnel obsessively and the middle of it not at all. The middle is where the money is decided, and it's measurable with a notebook and twenty minutes a week.
Here are the six numbers I'd track if I sold anything through chat, what each one actually tells you, and how to get them without hiring anybody.
Volume is a vanity number. Ignore it first.
More DMs is not a goal. I've watched sellers double their message volume by running a giveaway and end the month with fewer orders, because the extra 300 conversations buried the twelve people who were ready to buy.
Message count only means something as the denominator of a rate. On its own it's a story about your reach, not your shop. Every metric below is either a rate or a time. That's on purpose.
1. Median first-response time
How long a customer waits between "hi, how much is this?" and your first reply, human or automated.
Track the median, not the average. One conversation you answered three days late will drag an average into fiction. The median tells you what a typical customer experiences, which is the only version that matters.
Buying intent in DMs decays fast. Someone asking about a price at 11pm is usually comparing you against two other pages open in adjacent tabs, and whoever answers first often gets the order. I've written before about how tightly reply speed and conversion move together, and the short version is that the drop-off starts within minutes, not hours.
What it tells you to fix: if your median is over an hour, no amount of better copywriting will save the month. Fix speed before you touch anything else. That usually means either coverage during your peak DM hours or an assistant that answers the first message instantly and hands you the ones that need a human.
2. Reply rate: the conversations you never answered at all
Of every 100 people who wrote to you, how many got any reply? Not a fast reply. Any reply.
This is the number that shocks people. Sellers who feel like they answer everything usually sit somewhere between 70% and 85% once they count, because DMs get read on a phone in a taxi, mentally filed as "answer later", and then pushed off the visible screen by newer messages. The customer never sees an apology. They just see silence and buy elsewhere.
Count it once by hand and you'll stop arguing with the number. Open your inbox, scroll back seven days, and count the conversations where the last message is theirs. That's your leak, and it is almost always bigger than you'd guess.
What it tells you to fix: unanswered conversations are lost revenue you already paid to acquire. Before you spend anything on ads, close this. Auto-replies help, but only if they actually answer the question instead of saying "we'll get back to you soon", which is a message customers correctly read as "we won't".
3. DM-to-order conversion
Orders that started in chat, divided by conversations that started in chat. If I could put one number on the wall, it would be this one.
Be strict with the denominator. Not every DM is a shopper. A large share of first-contact messages are collaboration pitches, wrong-page messages, and people who wanted a follow-back. If you count all of them, your conversion rate will look terrible and you'll draw the wrong conclusion. Count only conversations where someone asked about a product, a price, a size, or shipping.
A healthy rate depends heavily on price point and category, so compare yourself to yourself, week over week. If you want an outside reference, I've laid out what a realistic DM conversion rate looks like for sellers at different price points.
What it tells you to fix: this number moves when your answers get better, not when your traffic gets bigger. If it's dropping while volume rises, your audience quality changed. If it's dropping while volume is flat, look at your replies: are you sending prices without photos, or photos without the "here's how to order" step?
4. Average order value from chat
Add up the value of chat-sourced orders and divide by the number of orders. Then compare it to your overall store average.
Chat baskets are usually bigger than site baskets, and sellers keep being surprised by this. When someone is talking to a person, they ask "do you have this in another colour too?" and buy both. That behaviour is exactly what a well-timed second suggestion in the conversation is built to catch.
What it tells you to fix: a flat basket size usually means nobody is suggesting anything. One extra line after the customer says yes ("this pairs with X, want me to add it?") is the cheapest revenue in the whole business. If your basket is big but conversion is low, you're probably scaring off the smaller buyers with a too-heavy pitch.
5. Repeat buyers from chat
What share of chat orders came from someone who had bought before? Track it monthly, not weekly, or the sample is too thin to mean anything.
A returning customer costs you nothing to acquire and decides several times faster than a stranger, because the trust step is already done. Most sellers I talk to have never counted this, which means they're paying for new eyeballs while ignoring a list of people who already handed them money.
What it tells you to fix: a low repeat share is rarely a product problem. It's a follow-up problem. A message a couple of weeks after delivery, asking how the thing worked out, does more for this number than any discount code.
6. Questions your side couldn't answer
The one nobody tracks. Every week, note the questions that got a vague answer, a "let me check", or nothing at all.
These cluster tightly. Shipping time to a specific city, whether an out-of-stock size is coming back, whether the fabric is the same as the one in the reel. Three or four questions usually account for most of your silent losses, and each one is fixable in ten minutes once you can see it written down.
I've made the case elsewhere that a handful of unanswered questions quietly kills more sales than bad pricing does. Writing them down is how the pattern stops being invisible.
How to measure all six without a data team
You do not need a dashboard. You need a habit and a spreadsheet with seven columns.
- Pick one day a week and stick to it. Friday morning, coffee, twenty minutes.
- Sample, don't census. Take the last 40 shopping conversations. Forty is enough to see a rate move; you don't need all 240.
- For each one, mark: did they get a reply, roughly how long it took, was there an order, and what it was worth.
- Log the stumper questions in a second tab as you go.
- Compute five numbers, write them in one row, and close the file.
Four weeks of that and you have a trend line, which is worth more than a perfect single measurement. The point isn't precision. The point is noticing that your median response time went from nine minutes to forty after you hired the new assistant, or that conversion fell the week you started answering price questions with a bare number.
If you'd rather not count by hand, an assistant that handles your chats can keep these numbers for you as a side effect of doing the work. Vardast answers Instagram, WhatsApp, Telegram and website messages from your own product information, hands the tricky ones to you, and shows you response times, conversion and unanswered questions in one place. If counting by hand sounds like something you'll do twice and abandon, have a look at how Vardast handles it.
Where I'd start on Monday
Pick the reply rate. Just that one. Count the conversations from the last seven days where the customer got the last word, and answer them today, however late it is. Some of those people still want to buy something, and a late reply converts better than no reply.
Then measure it again next week. The first metric you track is always the one that improves fastest, purely because you're looking at it.