How to tell a real buying post from noise: a scoring rubric for B2B teams
Why volume is the real problem
Finding public posts that mention your category is the easy part. A keyword feed can return a lot of posts that mention your category, and many may come from students, recruiters, vendors promoting themselves and people venting with no plan to buy.
The cost lands on your reps. Every post that reaches a person has to be read, judged and acted on or dropped. If each rep judges differently, the feed becomes a queue nobody trusts, and the one post that mattered sits under many that did not.
A shared rubric fixes the judging step. It needs six questions, a scale everyone uses the same way, and a rule for what each total means. Below are all three, then five invented posts scored.
New to the approach? Start with what buying-intent social listening is. If you have not yet tried reading posts by hand, how to find buying signals manually covers the search side.
Six questions to ask of every post
Writing the questions down makes the judging consistent across reps.
- Explicit need. Is the person asking for a product, a recommendation or a fix, or only discussing a topic?
- Named options. Are they comparing, shortlisting or naming what they use now?
- Timeframe. Is there a date, an event or a deadline behind the post?
- Authority hints. Do they sound like someone who decides or influences the purchase, from their role, the team they mention or how they talk about budget?
- Fit. Does their situation match the kind of company you sell to, in size, sector, region and problem?
- Recency. How old is the post? Requests go stale, and one from three weeks ago may already be solved.
Each question gets a score from 0 to 5. Add them up for a total out of 30.
The rubric
Copy this table into a spreadsheet. Score from what the post and public profile say.
| Question | 0 | 3 | 5 |
|---|---|---|---|
| Explicit need | No request at all | Asks a general question about the category | Directly asks for a tool, recommendation or replacement |
| Named options | None mentioned | Mentions a count or a type of option, no names | Names the options being compared or the tool in use |
| Timeframe | Nothing | Vague ("soon", "this year") | A date, deadline or event tied to the decision |
| Authority hints | Student, intern or unclear | Team member who would use the product | States a role that owns the budget or the decision |
| Fit | Wrong size, sector or problem | Partly matches your ideal customer | Clearly matches your ideal customer profile |
| Recency | Several weeks old | A few days old | Posted today |
Use 1, 2 and 4 for in-between cases. Two override rules keep the total honest:
- If explicit need is 0 or 1, the post is a topic discussion. Do not route it.
- If fit is 0, drop it. A perfect request from the wrong kind of company is still the wrong lead.
Then read the total against bands you set yourself. A reasonable starting point:
- 22 to 30: send to a rep.
- 12 to 21: hold for a second look by a person.
- Under 12: discard.
If you use the strong, possible and weak labels from our manual guide, strong is 22 to 30, possible is 12 to 21 and weak is under 12. The 0 to 100 scores on our landing page illustrate a planned automated score, not this rubric.
Five illustrative posts, scored
For these examples, imagine you sell shift scheduling software to employers with many sites. Every post below is invented to show the method. None is a real post, and none comes from a real person.
Post A. Illustrative example, not a real post. "We are replacing our spreadsheet rota before January across our 14 sites. Shortlisting three tools. Any recommendations for scheduling software that handles shift swaps? I run operations and need budget sign-off by the end of the month." (posted today)
Post B. Illustrative example, not a real post. "Anyone know a decent scheduling app for a small team?" (posted two days earlier)
Post C. Illustrative example, not a real post. "Our current scheduler keeps double booking shifts. Seriously thinking about switching. Has anyone moved to something else and been happy?" (posted three weeks earlier)
Post D. Illustrative example, not a real post. "Hiring: scheduling software developer, remote, apply below." (posted today)
Post E. Illustrative example, not a real post. "Just finished my operations module assignment on rota planning. What software do real companies use? Presentation is tomorrow." (posted today)
| Post | Need | Options | Timeframe | Authority | Fit | Recency | Total |
|---|---|---|---|---|---|---|---|
| A | 5 | 3 | 5 | 5 | 5 | 5 | 28 |
| B | 5 | 0 | 0 | 0 | 2 | 4 | 11 |
| C | 4 | 2 | 2 | 1 | 4 | 2 | 15 |
| D | 0 | 0 | 0 | 0 | 0 | 5 | 5 |
| E | 1 | 0 | 1 | 0 | 0 | 5 | 7 |
Verdicts.
- A, 28: send to a rep. It has a request, a date, a decision role and a matching company. Named options score 3 because it counts three tools without naming them.
- B, 11: discard. The need is explicit, but nothing tells you who is asking, how big they are or when they need an answer. The fit score of 2 is a judgement call. If small teams suit your ideal customer profile, it may deserve a human glance.
- C, 15: hold for a person. There is real pain and a leaning towards switching, but only a vague timeframe, unclear authority and an old post. A person can check whether the poster is still looking.
- D, 5: discard. It is a recruitment post that happens to use your category words. Need is 0, so the override applies. The perfect recency score does nothing for it.
- E, 7: discard. The need is homework, and the fit is 0.
If you would like to discuss how we plan the scored feed to work, request a demo. It is a conversation, not a product walkthrough.
Noise patterns
Discards often fall into groups like these four.
Recruitment posts. They use your category words because an employer wants someone to build or run that kind of product. The tell is a job title or "apply" wording. Need is 0.
Self-promotion. Someone describes a problem to introduce their own product. The tell is a pitch with a link and no real question. Check the profile before giving it any credit.
Students and homework. Real questions with real deadlines, which inflates the timeframe score. The tell is "assignment", "module", "thesis" or "for class". Fit should pull the total down.
Vendors fishing. Another supplier asks a leading question to collect ideas or leads. The tell is a question built to draw replies from many people, often from an account that mostly posts promotion. Treat it as noise unless the profile shows a genuine buyer.
When you drop a post, note the pattern in one word. After a few weeks those notes show which search terms attract the most noise, and you can tighten them.
Deciding who acts and how fast
A score tells you how much attention a post deserves. A separate rule decides who gives it. A starting point:
- High total with a named decision role: an account executive or the rep who owns that segment.
- High total from a user, not a buyer: an SDR, who can start a conversation and find the decision maker through normal means.
- Middle total: a reviewer on your team, whose job is judging, not selling.
How fast to act depends on your team and capacity, so set it yourself and write it down. Every band needs an owner, so no scored post sits unassigned. Routing social buying signals goes into ownership rules in more detail.
When someone does act, they should reply as themselves and say who they work for. See our reply policy guide. Read each network's terms before you act, because they govern what you may do on that network. This is not legal advice.
These posts come from identifiable people. Keep notes to the post link, the score and the date. Respect anyone who says they do not want contact, and check what rules such as GDPR or CCPA mean for your own use, with proper advice where you need it.
Calibrating the rubric against your own closed deals
The bands above are a starting guess. Your market may need different ones.
Take deals your team closed and deals it lost or never started. For each, find the public post, if any, where the buyer first talked about the problem, and score it without looking at the outcome first, so the result does not colour the scores.
Then see which questions separated the groups. In some markets a named option is the strongest sign. In others authority hints may matter more, for example where the person asking is not the person who signs. Adjust the bands or override rules to match.
Keep the sample small enough to do by hand and treat what you learn as a working assumption. Revisit it when your product, price or customer changes. Agree in advance how far apart two reps' scores may be, and clarify the anchors if they differ by more.
Where automation could help, and where a human should stay
Hand scoring is possible, but reading time grows with the feed. The repetitive parts suit software, where each network's terms allow it: finding public posts that match your terms, suggesting scores and flagging the noise patterns above.
Hushlead is in development, with early access planned, and plans to automate this. It is not a live product yet. Its planned design includes a list of 13 intent tags, such as Buying Intent, Recommendation Request, Comparison, Seeking Alternative and Hiring, that would sit alongside a score and route the result to the right person. That list is a plan and may change, and we make no claim about how accurate any scoring will be.
A human should stay on the judgement calls: whether an unclear profile is a buyer, whether a reply would be welcome, and whether contacting someone is appropriate at all. A person should decide whether to make contact.