PredictionHero
prediction-market aggregatorA greenfield product in one of the internet's most competitive categories — handed a content engine its team runs on AI.
PredictionHero normalises live odds from Polymarket, Kalshi, Opinion, Limitless and 10+ other platforms into one searchable dashboard — a genuinely strong product with almost no organic visibility. Domain Rating 0, three ranked keywords, in a category that got loud fast after the 2024 US election. We rebuilt the growth strategy on honest data, caught a launch-blocking bug before a penny was spent on content, and stood up a programmatic content pipeline the founder's team runs themselves — with us alongside as ongoing advisors.
We were invisible and I couldn't tell you why. Saher threw out the flattering keyword numbers, rebuilt our whole SEO strategy on data that actually held up, and caught a launch-blocking bug before we'd spent a penny on content — I came away with a plan my team can actually run.
Honest data over flattering data
The first keyword research had been run on a source that inflated search volume roughly 3× and badly understated difficulty — a median KD of 12 against a real 47. Acting on it would have meant chasing traffic that didn't exist against competition we couldn't beat.
Rebuilt on data that holds. We rebuilt the entire keyword universe on Ahrefs as the primary source, kept the old data only as a cross-check, and ran a second full re-sweep two weeks later to confirm the numbers held.
Found the actual opening. The head terms are locked — ~2.18M searches/mo sit behind a wall a DR-0 site can't climb. But the event and matchup layer scores KD 0–11: ~330k searches/mo genuinely winnable now, with a defensible 45k–80k visits/mo target.
The engine we built
The thesis: a programmatic content engine on the wide-open event layer, not a head-term arm-wrestle — and a pipeline the team runs without us.
A programmatic content thesis. Win the event/matchup layer with pages that match the product's real strength, instead of an unwinnable fight for locked head terms.
A 24-piece, E-E-A-T content architecture. A clean URL structure — /learn/ explainers, /compare/ head-to-heads, platform hubs, state-by-state legal pages — with thin affiliate ideas steered out as YMYL risks and replaced with link-earning assets, including an original-data study built to attract backlinks and AI citations.
An AI pipeline the team runs. A self-contained starter pack — content workflow, prompt library, a 'write like a human' de-slopping guide, an internal-linking playbook and a catalog of marketing plays — so the founder's team executes on AI directly, with us coaching the judgment calls.
Caught before it cost them
The technical audit came owner-assigned, with impact ratings, effort estimates and a sequenced roadmap — not just a list.
A launch-blocking bug, caught pre-spend. Every editorial post under /insights was inheriting the homepage's title, meta and Open Graph tags — meaning the entire content program would have been invisible to search before it began. Caught before the content budget was spent against it.
Foundations set. Trailing-slash redirect chains resolved, author/editorial-policy (E-E-A-T) signals added, and Search Console access established.
The win isn't a traffic chart yet — this is early and ongoing, and we say so. It's that a greenfield product went from 'invisible and unsure why' to a de-risked, data-backed content engine its own team runs on AI, pointed at the one door in a crowded category that's still open.