Skip to main content

Writing good queries

Queries are the heart of Kinn. They're how all the discovery happens — you ask a question in plain language and Kinn answers from the conversations it has ingested across your connected sources.

The most common mistake isn't asking a "wrong" question — it's asking a thin one. Faced with an open-ended box, it's easy to fall back on something like "summarize Discord this week." That'll give you something interesting, but it leaves most of Kinn's power on the table. The habits below are how you unlock the rest.

tip

You don't have to start with a perfect query. Give Kinn a little context and refine from there — a good answer is usually a short conversation, not a single flawless prompt.

Start by asking the agent​

If you're not sure what to ask, ask the agent itself. It's been built and tuned around the needs of real partners, so it usually has good ideas about how it can help. Something as simple as "I'm planning a marketing push and could use ideas for creative — what do you have on our audience?" is enough to get going. When in doubt, this is always a safe first move.

Give context before you ask​

Kinn is good at spotting opportunities you might miss — but only if it knows what you're trying to do. It's easy to ask for a metric and forget the goal behind it, and the goal is exactly what helps Kinn weight the answer for you.

Compare these two:

How are our creators doing on Instagram and TikTok?

We spent $50K on creator promotions this month and have a $10K budget for a follow-up campaign. Looking at engagement and comment sentiment across the creators we're tracking on Instagram and TikTok, which ones should we prioritize for the next round?

Same underlying data — but the second tells Kinn what "good" means for you, so the answer comes back weighted toward the decision you actually have to make.

Be specific​

When you leave things vague, Kinn fills the gaps with assumptions. It won't invent data — but you'll get sharper results when you pin down the specifics: dates, platforms, KPIs, and the current version, patch, or release (or DLC, for a game).

Specificity is also where Kinn has a real edge. Because it answers from your sources rather than the open web, it doesn't get pulled toward stale, viral posts the way web-grounded tools do. If a problem went viral a year ago and was patched since, a general model will often resurface that old post; Kinn won't, especially when you tell it which version or time window you care about. Naming the current patch lets you cleanly ignore feedback that's already been addressed.

Ask for visuals​

AI tends to produce a lot of text, which is great when you're the one digging — but sometimes you need to convey a result quickly. Kinn has a large chart library and a good sense of which chart fits which question, so just ask. "Chart engagement by creator over the last month" or "show sentiment as a trend" turns a wall of text into something you can drop into a deck.

Compare across platforms​

Analyzing several sources at once is one of Kinn's most novel capabilities, and precisely because it's novel, it gets overlooked. The key is knowing that each platform skews differently:

  • Discord is usually your core community — passionate, invested, close to the product.
  • Reddit tends to be more neutral and detached.
  • Steam skews toward strong reactions — complaints and intense feedback show up here first.
  • Social sources (YouTube, TikTok, Instagram) lean casual and creator-driven.

If you're hunting bugs, gauging feedback, or looking for insight, asking across platforms — and reading each one's bias into the answer — is a great place to start. "Are people reporting this crash across Discord and Reddit, or just in one place?" tells you far more than either source alone.

Standard vs. deep research​

Kinn has two modes, and it's worth using both:

  • Standard is your everyday mode — fast, and great for most questions.
  • Deep research is for when you want depth and maximum accuracy. It runs more services, expands the context window, and writes more thorough answers. Reach for it when the question really matters.

Next​

  • Example queries — a library you can copy and adapt, including before/after transformations.
  • Common pitfalls — the usual reasons an answer comes back thin, and how to fix each.