X Real-Time Lead & Intent Finder

Turn X (Twitter) into a warm, scored lead pipeline.

Most "Twitter scrapers" hand you a firehose of tweets and leave the hard part — who is actually a buyer? — to you. This Actor does the opposite. It reads tweets, replies, and profile bios the way a good SDR would, looks for real buying and help intent about your product or category, and returns a clean, deduplicated list of prospects, each scored 0–100 with a transparent breakdown of why.

Point it at what you sell, tell it where to look, and get back people you can actually reach out to today.

---

What you get

• Intent-scored leads, not raw tweets. Every row is a person expressing intent, with the exact phrase and category that triggered it.
• A transparent 0–100 relevance score built from four signals — intent strength, topic match, recency, and engagement — so you can dial quality up or down with one number.
• Bio-aware matching. Detects intent in the tweet and the author's bio (e.g. "founder evaluating a new CRM"), and tells you where it matched.
• Eight discovery modes you can mix and match: keyword search, hashtags, mentions, replies, quote-tweets, list timelines, communities, and account timelines.
• Author quality gates to filter out bots and throwaways (followers, account age, verified/Blue, following-ratio heuristic).
• Incremental runs. Schedule it and each run only returns new leads since last time — no duplicates.
• Real-time delivery to your CRM, Slack, Zapier, Make, or Google Sheets via webhook, in addition to the dataset.

---

Who it's for

B2B / SaaS demand generation

Find people asking "what's the best tool for X?" or "any alternative to <competitor>?" and reach them while the need is fresh. Filter to buying_intent and alternative_seeking for the hottest prospects.

Competitor conquesting

Scan mentions of a competitor, replies to their announcements, and quote-tweets of their launches. Catch frustrated or comparison-shopping users at the exact moment they're open to switching.

Agencies & lead-gen shops

Run one configuration per client, schedule it, and pipe scored leads straight into each client's CRM or a shared Slack channel. The per-source incremental cursor means clean daily digests with zero duplicate work.

Social selling

Give founders and reps a daily, ranked shortlist of prospects who are actively asking for what they sell — with the tweet link and a ready-to-read snippet, so they can reply personally.

Support & success triage

Watch mentions and hashtags for pain points ("frustrated with", "not working", "nightmare") about your product and route them to support before they churn or go public.

---

How scoring works

Each lead gets a 0–100 relevance score with a scoreBreakdown you can inspect on every row:

Signal	Max	What it measures
Intent	40	How strong the buying/help signal is (category weight + number of phrases).
Topic	25	How many of your topic keywords the tweet/bio matches.
Recency	20	How fresh the tweet is — newer intent is warmer.
Engagement	15	Log-scaled likes + retweets + replies + quotes (+ views when available).

Optional author-influence bonus (useAuthorInfluence): nudges higher-follower prospects up the ranking. It's kept out of the core 100 unless you turn it on, then reported as scoreBreakdown.authorInfluence.

Choosing minRelevance:

• 0 — keep every match (most volume)
• 40–55 — balanced quality (recommended starting point)
• 60–75 — only strong, on-topic, high-intent leads
• 80+ — very strict; fewer but hottest leads

---

Intent categories

Built-in, weighted phrase bags (you can select any subset and add your own):

Category	Weight	Example signals
Buying intent	5	"looking to buy", "pricing", "free trial", "willing to pay"
Alternative seeking	4	"alternative to", "switching from", "better than", "vs"
Pain point	3	"frustrated with", "too expensive", "nightmare", "sick of"
Seeking recommendation	3	"recommend", "best tool for", "what should i use"
Problem / help	2	"how do i", "need help", "struggling with", "not working"

Add customIntentPhrases for your niche, and choose matchMode: substring, whole word, or full regex.

The topic gate (requireTopicMatch, on by default) ensures a lead matches both a topic keyword and an intent phrase — so you get "looking for a CRM recommendation", not every tweet that says "recommend". Use ignoreKeywords to drop noise (giveaways, hiring, promos).

---

Discovery modes

Enable any combination:

Mode	What it scans	Great for
Search	Keyword/phrase queries	Broad demand-gen
Hashtags	One or more hashtags	Topic & event monitoring
Mentions	Tweets mentioning @handles	Conquesting your brand or competitors
Replies	Replies to target tweets/accounts	Catching responders to a competitor
Quotes	Quote-tweets of target tweets/accounts	Reaction-based intent
List	X List timelines	Curated prospect pools
Community	X Community timelines	Niche buyer groups
Timeline	Specific accounts' recent tweets	Named-account monitoring

Everything streams through the same intent → topic → score pipeline, is deduplicated within a run, and is memory-light thanks to cursor streaming.

---

Output

One row per lead (featureType: "lead"), including:

• leadType (tweet | reply | quote) and discoveryMode
• tweetId, tweetUrl, text, authorUsername, authorName, createdAt, lang
• engagement (likes, retweets, replies, quotes, bookmarks, views)
• matchedIntentCategory, matchedIntentCategories, matchedIntentPhrases
• matchedTopicKeywords, matchedSnippet (centered on the first match), matchedIn (tweet | bio | tweet+bio)
• relevanceScore and scoreBreakdown { intent, topic, recency, engagement }
• authorProfile when enrichment is on (followers, following, verified/Blue, account age, tweet count, bio, location, following ratio)
• sourceQuery, scrapedAt, and optional raw (with includeRaw)

Unreachable, protected, or deleted targets are reported as unavailable rows with a reason — the run never crashes on a bad target.

A per-run summary is written to the key-value store under OUTPUT: sources scanned, tweets examined, leads emitted, upstream request count, and any non-fatal errors.

Dataset views: Overview, Leads, Buying intent, Alternative-seeking, and By source.

---

Incremental & delivery

• Per-source cursors are stored between runs (keyed per query/handle/list/etc.), so scheduled runs only emit tweets newer than last time. Set a manual sinceId floor and a maxLeads cap as needed.
• Webhook (webhookUrl + webhookFormat): every new lead is POSTed to your endpoint in addition to being saved to the dataset. Choose json (full lead) or slack (headline, score, intent, author, topics, tweet link). Delivery failures are ignored so your dataset always fills up.

---

Quick start

1. Set topicKeywords to your product/category/competitors (e.g. ["CRM", "hubspot"]).
2. Turn on one or more sources and give them inputs (e.g. sourceSearch + searchQueries).
3. (Optional) Set minRelevance to 40 and turn on author gates to raise quality.
4. (Optional) Add a webhookUrl to push leads into Slack or your CRM in real time.
5. Run it — or schedule it for a fresh, deduplicated lead digest every day.

---

Notes

• This is a public Actor. You do not need to provide any X login, cookies, or personal proxy — just configure what you sell and where to look.
• Data volume and freshness depend on public availability at run time.

---

Contact me

Questions, feature requests, or custom lead-gen pipelines?

• Email: dubem115@gmail.com
• GitHub: github.com/DrunkCodes