Alternative Data Strategies: Turning Signals Into Portfolios

An alternative data strategy is a rule for turning non-traditional signals into positions. Instead of trading on price momentum or a valuation screen, you trade on what the data says a business or asset is actually doing — how fast it is being adopted, whether attention is accelerating, how sentiment is turning, and where capital is flowing — and you encode that read into how much you hold and when you buy or sell.

The raw material is alternative data: developer activity, search interest, app usage, product reviews, social reach, market sentiment and real-time order flow. The strategy is the machinery that turns those signals into a repeatable, testable book.

From signal to score

Individual signals are noisy and live on different scales — GitHub stars, review counts and trading volume are not comparable numbers. The first job of any serious alternative data strategy is normalization: convert each signal to a percentile rank within its category, weight it by how much it should be trusted (source reliability × recency × independence), and aggregate the weighted signals into a single interpretable score, with a separate confidence that never hides inside the number. Prismetric calls this a TechScore; the principle generalizes to any composite signal.

Three ways to size positions from a score

Once every candidate has a score, the strategy decides how much to hold of each. Three approaches cover most of the useful space:

• Score-weighted — position size is proportional to how far a name's score sits above an exit floor. Higher conviction earns more capital, with no hand-picked weights. This is the default for a diversified, signal-tilted book.
• Top-N — hold only the N highest-scoring names, equally weighted. A concentrated, momentum-style expression that rotates as the leaderboard changes.
• Threshold / equal-weight — equal-weight everything above an entry bar. The simplest expression of "own what scores well."

Triggers: when the score decides to trade

A strategy is not just an allocation; it is a set of rules that fire over time. Score-driven entry and exit thresholds replace discretionary timing: buy when a name's score crosses the entry level, trim or sell to cash when it falls below the exit level. The gap between the two — hysteresis — stops a name hovering at the line from churning in and out every day and quietly bleeding the book with transaction costs.

Choosing the signal that drives it

The same universe produces very different strategies depending on which signal you weight. A headline composite score builds a balanced, quality-tilted book; a pure momentum sub-score builds a faster, higher-turnover one; an adoption or attention signal captures different edges again. Part of building an alternative data strategy is discovering which signal, on which universe, at which thresholds, has actually paid — which is a job for rigorous backtesting, not intuition.

Why this beats trading the number

Reported fundamentals are backward-looking and already priced in by the time they land. A signal-driven strategy trades the leading indicator instead of the lagging one, and does it systematically — the same rule applied to every name, every rebalance, with no story-telling. That discipline is the entire point: it makes the edge measurable and repeatable rather than a series of good calls you cannot reproduce.

See it working

Prismetric's Strategy Lab builds exactly these books from live scores — thematic portfolios sized by score, triggered by thresholds, backtested on real prices and marked to live market data. Read how the strategies work, or start with the frontier data sources powering the newest signals.