Lead Scoring for Your Instagram & TikTok Leads
ChattyFlow team · Updated
ChattyFlow scores every conversation its AI Lead Engine has on Instagram, TikTok or your website from 0 to 100, based on what the person actually asked, shared and agreed to. 70 and up is Hot, 45 to 69 is Warm, and every point shows its reason. Included in Growth plans and the 14-day trial; not on Free.
What lead scoring is, and what it needs
Lead scoring is part of the AI Lead Engine, the paid layer of ChattyFlow. When the Social Agent answers a comment or a DM on Instagram or TikTok, or a visitor writes in the AI chat on your website, that conversation becomes a lead. After every message, ChattyFlow recomputes the lead's score from 0 to 100 using everything stored about the conversation, then labels it.
The score is calculated in code from stored facts, not left to the AI's judgement of the moment: the same conversation always gets the same score, and each point is kept with the evidence behind it. When the AI agent rates the conversation itself, that rating is one signal among the others, never the whole score.
Two limits to know. Lead scoring needs a Growth plan (Growth 20, 75 or 150) or Business; the 14-day trial every new account starts with includes it, the Free plan does not. On TikTok, direct messages are not available in the United States, the EEA, Switzerland or the United Kingdom, and a comment can open a DM only for accounts in Vietnam, Thailand and Indonesia. Elsewhere the agent answers TikTok comments publicly, and those public conversations are scored the same way.
Hot, Warm, In conversation, Filtered
Each lead carries one of four labels, decided only by its score and your rules:
| Label | Score | What it usually means |
|---|---|---|
| Hot | 70 to 100 | Several strong signals at once: a need, a near timeline, contact details, consent. |
| Warm | 45 to 69 | Real interest with something missing, often a timeline or a way to reach them. |
| In conversation | 0 to 44 | Talking, but not enough signal yet. There is no "cold" label: a lead stays here until it earns more points. |
| Filtered | 0 | Turned away by your own rules; the score is set to 0. |
Every signal and its points
These are the signals the score reads today, and the points each one adds or removes. Points add up, and the total stays between 0 and 100.
| Signal | Example | Points |
|---|---|---|
| Named a problem | "Our customers' messages get lost at night" | +12 |
| Asked whether you can do it | "Can you…?", "Do you deliver…?", "Is it possible…?" | +12 |
| Answer inside a band you marked as a fit | "We're 12 people" when 10 to 50 is a fit | +12 per band |
| Left contact details | An email or a phone number | +10 |
| Agreed to be contacted | Said yes to the consent question | +8 |
| Asked about price | how much, price, cost, quote, combien, prix, tarif, devis, بكم, السعر, قداش | +8 |
| Kept the conversation going | 2 messages / 3 or 4 / 5 or more | +2 / +5 / +8 |
| Reacted to a specific part of your video | Call to action / objection / claim / other / hook | +8 / +7 / +6 / +3 / +2 |
| Gave a timeline | Any timeline, then: near ("this week", "next month") / within 4 to 6 months / far ("next year", "later") | +6, then +6 / +3 / −4 |
| Warm attitude | "Perfect, I'm interested" | +6 |
| Asked whether it fits them | "Is this for a shop like mine?" | +5 |
| Gave a role, and it is a decision-maker role | "I'm the owner", "founder", "manager" | +3, then +5 |
| Reply speed (median) | Within 90 seconds / within 10 minutes / a day or more | +5 / +3 / −3 |
| Talked about the company | Its name or its size | +4 |
| Message length | 80 characters or more on average / one-word replies | +4 / −3 |
| Used your trigger word | Commented the automation's keyword | +4 |
| Other questions | Practical (delivery, hours, address) or about you | +2 |
| Attitude | Hesitant / worried about price / disengaged / not the decision-maker / frustrated | −4 / −2 / −6 / −6 / −10 |
| Answer inside a band you marked as not a fit | A value you ruled out | −20, and the score is capped at 25 |
| Your team's feedback | After 3 or more verdicts on your account | up to ±10 |
| The AI agent's own rating, when it gives one | Fit and intent from 0 to 5, engagement from 0 to 5 | up to +25, +25 and +10 |
A worked example: from a comment to Hot
A fitness studio runs the Social Agent on a Reel, with COACH as the comment keyword. Here is one conversation scored with the real weights, assuming the agent records the problem, role and timeline it hears, which is its job:
| What the person does | Signal | Points |
|---|---|---|
| Comments "COACH" under the Reel | Used the trigger word | +4 |
| "Can you run private sessions for my team?" | Asked whether you can do it | +12 |
| "Most of us have back pain from sitting all day" | Named a problem | +12 |
| "I'm the office manager" | Gave a role, and it is a decision-maker role | +8 |
| "We'd like to start next week" | Gave a timeline, and it is near | +12 |
| Answers each message within a minute | Replies fast | +5 |
| Leaves an email and says yes to being contacted | Left contact details, agreed to be contacted | +18 |
| Six messages in total | Kept the conversation going | +8 |
Teach it your bar: definitions, feedback and re-scoring
Scores get sharper when the scoring knows your rules. In the Lead Engine, add your own definitions: the question the agent should ask and bands for the answer, as a number range or a list of choices, each marked a fit or not a fit. A measured answer inside a fit band adds 12 points; an answer inside a band marked not a fit removes 20 and caps the score at 25.
Claims are not measurements. "We're small" or "quite big" can't be compared with your bands, so they earn nothing until the person gives a figure. Figures typed as words ("a dozen", "twenty") or in Arabic-Indic digits such as ٣ or ٢٠ still count.
Then teach it. On any lead, answer "Does this score look right to you?" with Agree or Disagree, and give your own score from 0 to 100 with an optional reason. Once your account has three or more verdicts, the average gap between your scores and ChattyFlow's moves future scores by up to 10 points in your direction.
Finally, Re-score all, on the Leads list, recomputes up to your 500 most recent conversations from what is stored, without calling the AI. Owners and managers can run it.
What lead scoring tells you
Who to answer first
Hot and Warm leads rise to the top of your Leads list, so your team starts with the people closest to buying.
Every point has a reason
Open a lead to see each signal with its points and the words it came from, such as "Asked about price +8" or "Left contact details +10".
It learns your bar
Agree or disagree with a score and suggest your own. After three verdicts, your feedback moves future scores by up to 10 points.
How lead scoring works
- Start a Growth plan or your 14-day trial, and connect Instagram, TikTok or your website chat.
- Fill in the Knowledge Base and your own definitions, such as "up to 20 = quick job, over 100 = custom quote", and mark which answers are a fit.
- The AI Lead Engine answers comments and DMs. After every message, the score is recomputed from the whole conversation.
- Open Leads, filter Hot, Warm or In conversation, and tap Why to see the points behind a score.
- Agree or disagree with scores, then use Re-score all so earlier conversations benefit from what the scoring has learned.
Why it matters for shops and service teams
- A price question is interest. A price question plus a timeline and a phone number is a buyer. The score adds it up for you.
- Unhappy existing customers are never scored as prospects: a complaint caps the score at 20 and flags the conversation for a person.
- An answer that lands in a band you marked as not a fit caps the score at 25, so a request outside your rules can't jump the queue.
- Your team's verdicts feed the next scores, so the bar becomes yours rather than a generic one.
- The same scoring on Instagram, TikTok and your website chat, reading English, French and Arabic.
Frequently asked questions
What is lead scoring?
A 0 to 100 score for each conversation, recomputed after every message from what the person asked, shared and agreed to. It tells your team who to answer first: Hot from 70, Warm from 45.
Which signals count the most?
Naming a problem, asking whether you can do the job, and giving an answer inside a band you marked as a fit are worth 12 points each. Leaving contact details adds 10, agreeing to be contacted 8, and a price question 8.
Is there a cold label?
No. Below 45 a lead is shown as In conversation until it earns more points. Leads your own rules turn away are Filtered, with a score of 0.
Does the score use orders or returns?
No. It is built from the conversation itself: messages, answers, consent, reply speed and, when the video has been analysed, the moment the person reacted to. Orders, returns and payments are not part of it.
Can I see why a lead got its score?
Yes. Every lead lists its signals with their points and the words they came from, and you can agree or disagree with the score.
Does it work on TikTok everywhere?
The agent answers TikTok comments publicly in every country, and those conversations are scored. TikTok direct messages are not available in the United States, the EEA, Switzerland or the United Kingdom, and a comment can open a DM only for accounts in Vietnam, Thailand and Indonesia. Everywhere else, people who message you first can be answered and scored in DMs.
Which plans include it?
Lead scoring is part of the AI Lead Engine, included in Growth 20, Growth 75, Growth 150 and Business, and in the 14-day trial every new account starts with. The Free plan does not include it. Each Growth tier includes a monthly number of qualified leads: 20, 75 or 150.
Know which conversations are worth your time
Start your 14-day trial, connect Instagram, TikTok or your website, and let every conversation earn its score.
Start your 14-day trial