Ballad
Product · Market analysis

Know what your market sees, and act on it.

The market analyst you haven't hired: every channel's numbers on one board, your share of the AI answers buyers read, what those buyers are saying about you, and which article or post turned into each customer. When the evidence moves, the plan is asked to move with it.

The job

What a market analyst would own, done while you ship.

  • Performance reporting

    Site traffic, referrers, social numbers, search rankings and AI citations on one live board, read on their own.

  • Share of voice

    Asks Claude and ChatGPT the questions your buyers ask, every week, and tracks who gets named: you or your competitors.

  • Voice of the customer

    Reads replies, comments, threads and interviews, and distils them into the questions, objections and words your buyers use.

  • Search performance

    Your Google queries and positions, read daily from Search Console, with articles that start to slip flagged.

  • Attribution

    Follows each visitor from the post that brought them to the signup, and each person into the account they belong to.

  • Recommendations

    Turns what it learned into proposed plan changes and experiments, applies the safe ones, and reverts what doesn't hold up.

01 · Signals

Every signal, on one board.

The weekly report a market analyst would assemble from six tools is one live page. A map of who's reading right now, then a widget for each kind of signal: your site's sessions and top pages, the people who signed up and the post that first brought each one, the referrer mix including AI answer engines, social, search, and ballad.to.

  • Social numbers, collected. On X, Ballad reads impressions, replies and reposts itself every six hours. Everywhere else it asks two days after a post goes live, and what you type feeds the next plan.
  • Search, daily. The queries you show up for on Google and your position on each, pulled from your own Search Console. The searches you could win, keyword research and links have a page of their own.
  • Ask it anything. ⌘K on Signals or Strategy. “Which post brought this week's signups?” comes back drawn only from your own data, every claim linked to the post, person or plan change it came from.
SEO & Website: searches you could win
app.balladlabs.com
Signals
Everything your content is doing, as it happens.
7d30d90d
Live visitor mapLive · 23 on site
site visitlink click
Web
4,812
+18% wk
Signups
41
+9 this wk
SEO
#4
↑3 · 'job queue'
AI answers
31%
Claude · ChatGPT
ballad.to
918
+40% clicks
02 · AI visibility

Your share of the AI answer.

Buyers increasingly ask an AI before they ask Google. Ballad keeps a list of the questions your buyers ask, asks Claude and ChatGPT each of them every week, and records who gets named and whose site gets cited: you, or the competitors you're watching. The figures combine both, and each engine's own rate is there too. Claude and the weekly read come with Startup and Business; Founders asks ChatGPT once a month.

  • Share of voice. How often you're named against each competitor, how often your own site is the source, and which way it's moving. Sampled and directional, never a made-up rank.
  • Questions suggested. Start from the questions Ballad drafts from your positioning, then add the ones you hear on sales calls.
  • Gaps become experiments. Where a competitor is named and you aren't, one click starts an experiment, and the next drafts aim at that question.
app.balladlabs.com/strategy/market
Market & competitors
12 buyer questions · asked of Claude and ChatGPT every week
Share of voice ↑6
Share of voice
Halyard44%
Driftwood31%
Tidewell25%
named in 58%cited in 17%
Tracked questions
“best background job queue for Postgres”names Driftwood, Halyard
“job queue for Node without Redis”names Halyard, Tidewell
“how do I replay a failed background job”names Driftwood
Where competitors wincited source: a “best job queues” listicle
Queues for Node without Redis: Halyard is named, Driftwood never is.
Target this gapstarts an experiment · next drafts aim at it
03 · Listening

What your market says, in its own words.

Customer research usually means a quarterly survey. Ballad listens all the time: replies to your posts, comments you paste in, Hacker News threads in your space, and customer interviews, distilled every day into the themes your buyers actually repeat.

  • Threads count too. Comments in Hacker News threads about your space are recorded alongside replies and interviews, with the thread they came from.
  • Questions, objections, phrases. What prospects keep asking, what makes them hesitate, and the words they use for the problem, with who said it kept to a handle or a role.
  • Into every brief. The questions come first in every draft Ballad writes, so the content answers what the market is asking this month.
app.balladlabs.com/signals
What your market is saying
From 142 replies, comments, threads and interviews · last 30 days
Listening
Just heard
HN comment · job queue threadquestion

Does anything replay failed jobs without a second tool?

@jkwan · reply on Xobjection

Moving our queue off Redis mid-project sounds painful.

Customer interview · Staff Engineerphrase

Debugging a stuck cron job used to eat my whole afternoon.

What people say
Question · 9×Can I replay a failed job with different arguments?
Objection · 6×Moving off Redis mid-project sounds like a migration mess.
Their words · 11דthe cron box”, never “scheduler”
questions first in every brief
04 · Attribution

From the post to the account.

One line on your site and a token on every link Ballad publishes. When someone signs up, Ballad knows the post that first brought them and the last one they read, and People keeps the record: who they are, the account they belong to, and what they did next.

  • First touch and last. Signups by post, channel and angle, including the ones that arrived from an AI answer, and the share that came from Ballad's content at all.
  • People and accounts. Each person in the company, team or workspace they belong to, with its traits, so a question can be asked per account rather than per email.
  • First-party, and polite. No cookies, no cross-site profile, and nothing identifying under Global Privacy Control. Your records, on your site.
Read the attribution guide
app.balladlabs.com
Attribution
Which post brought the signup.
beacon receiving
Signups · 30d
19
+6 vs last
From Ballad
13
68% of signups
Top source
LinkedIn
7 traced
Signups by post13 of 19 traced
Our cron box ran for 14 months. It never broke. That was the problem.
LinkedIn · Thu 8:30
7
The Postgres bloat fix nobody writes down
Your blog · cited by Claude
4
Six cron horror stories, one root cause
X thread · Sat
2
LinkedIn · 7AI answers · 3X · 2ballad.to · 1→ feeds the next plan
05 · Recommendations

Analysis that changes the plan.

An analyst's report ends with recommendations someone has to act on. Ballad's ends with the action: it writes up what it learned this week, proposes the plan changes that follow as diffs in your Inbox, runs the experiments, and watches what it changed. Strategy shows the whole picture in one place: what Ballad knows, what it's doing about it, and what happened.

  • What it's learning. A plain-language read of what's working and what isn't, with a confidence score that moves as the evidence comes in.
  • Experiments, judged on signups. Threads against single posts, a new angle against the old one, run to a result and scored on what converts.
  • Every change reversible. Safe changes apply themselves and are watched for two weeks. If they don't hold up, they roll back on their own.
app.balladlabs.com/strategy
What Ballad knows
Ballad's readAdjusting
DRIFTWOOD
Background jobs in the Postgres you already run.
Postgres-backed SaaS teamsTeams outgrowing cron
Voices
Maya founder
LinkedIn 3/wk · X 4/wk
Driftwood brand
X 5/wk · Blog 2/wk
Competitors
Halyard Queue
Tidewell
2 battlecards
What Ballad believes9 held · 3 hypotheses
Buyers come to Driftwood after a job ran twice, not to save money. 82%
What it's doing about it
The plan this weekv12 live
Ballad adjusts on its own
2 experiments running1 launch under way
What Ballad learned
Lean in: mistake-led posts earn 2.1× the replies of how-tos.
Ease off: feature lists on LinkedIn.
What happened
Plan adjusted · LinkedIn leans toward image cards2d
auto-reverts if it underperforms in 14d
The rest of the loop

One system, not a stack of tools.

What one job learns, the others use: the calendar writes from the positioning, the launches measure through the same signals, and the analysis changes the plan.

The report writes itself. So does the plan.

Add one line to your site and name your competitors. By next Monday Ballad has read your signals, asked the market's questions, and has its first recommendations waiting.