Some of what is written about your brand on X (formerly Twitter) is genuine feedback about a store visit, much of it is advertising and unrelated noise, and a small but valuable share comes from prospective customers asking where they can buy a product. Separating the three by hand is not sustainable for a chain. This post explains how X (Twitter) customer feedback is collected in BranchVoice, how AI screens out the noise, and how the same flow produces sales leads.
Why customer feedback on X is different
A Google review is written to a store and a Şikayetvar complaint to a brand; a post on X may be addressed to nobody at all. The customer writes to their followers without tagging the brand or naming the store. On top of that, the same search returns the brand's own announcements, influencer collaborations and giveaways. What a regional manager needs is not the whole stream but the real experiences inside it.
The second difference is speed. A complaint on X can spread within minutes, so a weekly reporting cycle is useless here. The third difference is opportunity: no map or complaint platform ever shows you a customer who has not yet decided to buy. X does.
Two separate passes: reviews and leads
BranchVoice monitors X from a single settings screen, but with two separate passes:
- Review pass: collects brand-related posts through search terms such as the brand name, the brand's accounts and mentions of those accounts, and stores them as reviews.
- Lead pass: searches with keywords in need-and-category language, such as "lingerie recommendation" or "winter coat suggestions", and flags posts that carry purchase intent as leads.
Scheduling, limits and filters are configured separately for each pass: how many posts to fetch at most, verified accounts only or not, a minimum follower count, whether retweets and replies are included, language and location. Each run requests only posts newer than the last sync, so the same content is never paid for twice, and every run is visible on the sync jobs page.
The relevance gate: not every post is saved
Unlike the other platforms, a post from X goes through an AI screen before it is written to the database at all. In the review pass the question is simple: is this post actually about the brand? In the lead pass three conditions must hold together: the author is a prospective customer, the product mentioned falls into one of the company's categories, and the post is not an advert, a priced product list, a giveaway or a brand announcement. A post that fails the gate is never stored, so no noise reaches the dashboard and analysis cost is incurred only for relevant content.
For the gate to work well, the AI fields in the company profile must be filled in: brand aliases (including account handles) and product categories. If the categories are empty, the lead gate lets everyone through; once the profile is complete, the screen tightens.
The review pass: posts about the brand
Posts that pass the gate are stored as reviews and go through the same AI questions as reviews from any other platform: sentiment, risk level, topics, responsible unit, employee mentions. A post is first attached at brand level to the company's virtual "General" store; if the text contains a clue such as a mall, a district or a store name, AI compares it against candidate stores from your list and moves it to the store when the decision is confident enough. Posts that cannot be resolved stay in the General pool. The decision logic is described in matching reviews to stores.
A critical post, say a hygiene or safety complaint, goes to the alert engine and, depending on the rule you defined, lands as an email with the regional manager or the customer experience team; see critical review alerts for rule design. The post's like count and the author's follower count are stored as well, which helps you spot a complaint with high spread potential.
The lead pass: from purchase intent to a sale
Posts that pass the lead gate live in the review table and also become a lead card in the Lead Center. When the AI analysis completes, every lead receives a score from 0 to 100 and a tier: hot, warm or cold. The score is a weighted average of the lead questions, such as purchase intent, urgency, recommendation seeking and brand affinity, with a small bonus for the author's reach (followers, engagement, verified account). Until the analysis finishes, the lead shows as "awaiting assessment".
For the sales or marketing team the flow looks like this: hot leads are listed at the top of the Lead Center, each lead's status is updated as new, contacted, qualified, converted or dismissed, and the team leaves notes on the card. The demand trend card shows which categories and keywords have picked up in recent days, and the AI summary comments on waiting hot leads and keyword performance. The lead side is covered in full in capturing leads from social media.
| Need | Where it lives in BranchVoice |
|---|---|
| Isolate genuine experiences with the brand | AI relevance gate |
| Get a post to the right store | Store matching and the General pool |
| See a spreading complaint the same day | Alert rules and email |
| Turn purchase intent into a sale | Lead pass, score and tier, Lead Center |
Next steps
For a chain, X is both the noisiest and the most opportunity-rich source; the relevance gate and the two-pass design mean your team sees only real feedback and real buyers. To see the flow with your own brand and category keywords, request a demo.