We Tested Whether Our Hiring Signals Predict Stock Moves — They Don't, Yet
October 29, 2026 · TalNexis Research
TalNexis tracks hiring patterns across nearly 100 tech and tech-adjacent companies. The obvious next question: do those patterns show up in the stock price before the market notices? We backtested two of our hiring signals against real stock performance to find out. The honest answer: not yet, not at a level we'd trust. Here's how we tested it and what we found.
How we tested it
We took every signal of two types — AI Hiring Surge and Hiring Velocity — detected between March and October 2026, kept only the signals at publicly traded companies (53 tickers), and simulated buying the stock on the day the signal fired and selling 7, 30, or 90 days later.
To know if a result is real, you have to compare it to something. We checked each trade against two benchmarks:
- The S&P 500 (SPY) — the broad market
- The tech sector (XLK) — the fairer comparison, since most signal companies are already tech stocks
A signal that only beats the S&P but not the tech sector isn't finding anything — it's just restating that tech had a good year.
What each signal actually measures:
- AI Hiring Surge fires when a company posts at least 5 AI/ML roles in a week, and that's at least double the prior week's count (with at least 3 more roles, not just a 2-to-4 blip). It's a sudden, short-term spike — one strong week against the week before it.
- Hiring Velocity fires when a job category grows for 3 straight weeks in a row (each week higher than the last, starting from at least 2–3 roles and reaching 5+ in the current week). It's a sustained trend, not a single spike.
One more rule: independent trades only. Some companies fire the same signal many times in a few months, with their 90-day price windows overlapping each other. We only count a new trade for a company once its previous 90-day window has fully closed — so every trade in this analysis is a genuinely separate, independent observation, not the same stock move counted several times.
One more rule: exclude data-pipeline artifacts. A company whose job-board source breaks (a moved ATS URL, a misconfigured feed) goes silent for a stretch, then dumps a backlog of postings in a single day once it's fixed. That backlog dump can look exactly like a real hiring surge to a signal detector, so we exclude any signal within 21 days of a company's source resuming after a 10+ day silent stretch. This removed 26 of 3,533 signals in our broader dataset (concentrated in Walmart and Replit), including one signal that had been feeding this analysis.
What we found
Source: TalNexis signals data, March–October 2026
Neither signal beat the tech sector by a statistically meaningful margin at any holding period. The 90-day Hiring Velocity result is the closest thing to a real finding — a +4.85 percentage point average edge — but it's well short of the standard significance threshold (p = 0.10, not p < 0.05), and the win rate (46%) is actually below a coin flip. The honest read: there's a hint worth watching here, not a strategy worth trading on yet.
| Signal | Hold | Independent trades | Avg excess vs. tech sector | Beats tech sector? | Statistically real? |
|---|---|---|---|---|---|
| AI Hiring Surge | 7 days | 100 | +0.01% | No (44%) | No |
| AI Hiring Surge | 30 days | 70 | −0.85% | No (47%) | No |
| AI Hiring Surge | 90 days | 33 | +2.26% | No (42%) | No |
| Hiring Velocity | 7 days | 345 | −0.42% | No (48%) | No |
| Hiring Velocity | 30 days | 147 | +1.05% | No (48%) | No |
| Hiring Velocity | 90 days | 57 | +4.85% | No (46%) | No (p = 0.10) |
What this isn't
This is a research finding, not a trading recommendation. A few honest limits:
- One time window. All of this comes from eight months of data (March–October 2026), during a strong run for tech stocks. We haven't tested it against a down market or a different year.
- No out-of-sample test. We found this pattern and checked it on the same data — we haven't yet confirmed it holds on signals detected after this analysis.
- Small per-company samples. Requiring independent, non-overlapping trades means most companies only contribute one or two data points each over eight months. The portfolio-level average is meaningful; any individual company's result is not.
- No trading costs. Real trades have fees, slippage, and taxes, none of which are modeled here.
None of this is investment advice.
The part a stock-only strategy can't see
A lot of the most striking hiring activity in our data is happening at companies you can't buy: OpenAI, Anthropic, Stripe, Databricks, Anduril, SpaceX, and dozens of other AI-native companies have no public stock. Any strategy built purely on tradeable signals misses exactly the companies most likely to be moving first.
That's the real argument for TalNexis as a dataset, not just a backtest: the hiring signal exists at these companies today, years before some of them go public — and whoever is watching that signal now has a head start when they do.
We're tracking this forward on live signals as they come in, and we'll publish an update once we have enough fresh data to know if it holds. If you want to see the hiring signals behind this research as they happen, explore TalNexis signals.
Appendix: the full company list
All 55 publicly traded companies in our tracked universe, with their Hiring Velocity performance using independent, non-overlapping 90-day trades. Average excess return is measured against XLK (tech sector). Companies with no entry had no Hiring Velocity signal, or not enough elapsed time for a full 90-day window yet.
| Company | Ticker | Independent trades | Avg return | Avg excess vs. tech sector |
|---|---|---|---|---|
| Okta | OKTA | 1 | +57.85% | +61.65% |
| CrowdStrike | CRWD | 2 | +56.76% | +46.08% |
| Elastic | ESTC | 2 | +48.23% | +36.26% |
| Snowflake | SNOW | 2 | +47.95% | +36.68% |
| GitLab | GTLB | 1 | +46.52% | +35.97% |
| Coinbase | COIN | 1 | +16.99% | +9.12% |
| MongoDB | MDB | 1 | +27.50% | +19.16% |
| Cloudflare | NET | 2 | +27.07% | +15.38% |
| Workday | WDAY | 2 | +24.48% | +13.68% |
| Regeneron | REGN | 1 | +24.15% | +21.16% |
| Intel | INTC | 2 | +20.67% | +11.19% |
| Salesforce | CRM | 2 | +18.33% | +7.52% |
| Pfizer | PFE | 1 | +18.04% | +9.60% |
| Mastercard | MA | 1 | +16.48% | +17.47% |
| Verizon | VZ | 1 | +14.15% | +15.14% |
| Visa | V | 1 | +13.90% | +14.90% |
| AT&T | T | 1 | +13.52% | +5.88% |
| Target | TGT | 1 | +12.82% | +6.86% |
| Comcast | CMCSA | 1 | +9.09% | +10.08% |
| CarMax | KMX | 1 | +6.82% | −0.81% |
| Nvidia | NVDA | 1 | +6.35% | −9.47% |
| RDDT | 1 | +6.21% | −3.78% | |
| Lyft | LYFT | 1 | +6.12% | −4.43% |
| Block | XYZ | 2 | +5.80% | −7.40% |
| Zoom | ZM | 1 | +4.12% | −2.36% |
| Instacart | CART | 1 | +3.03% | −5.31% |
| Samsara | IOT | 1 | +2.65% | −1.89% |
| Toast | TOST | 2 | +2.54% | −8.74% |
| CoreWeave | CRWV | 1 | +1.91% | −9.57% |
| Adobe | ADBE | 2 | +1.17% | −10.34% |
| Disney | DIS | 1 | +0.98% | +2.31% |
| Robinhood | HOOD | 1 | +0.09% | −8.94% |
| Datadog | DDOG | 1 | −0.46% | +0.58% |
| Twilio | TWLO | 1 | −4.18% | +0.66% |
| Broadcom | AVGO | 1 | −4.43% | −10.34% |
| UPS | UPS | 1 | −5.10% | −2.97% |
| Micron | MU | 1 | −5.49% | −11.47% |
| Home Depot | HD | 1 | −6.03% | −5.04% |
| FedEx | FDX | 1 | −9.68% | −11.20% |
| Walmart | WMT | 1 | −11.85% | −18.34% |
| Lennar | LEN | 1 | −12.15% | −8.34% |
| CVS Health | CVS | 1 | −12.99% | −14.51% |
| Lowe's | LOW | 1 | −14.04% | −19.08% |
| FIS | FIS | 2 | −14.16% | −26.23% |
| Caterpillar | CAT | 1 | −17.88% | −16.89% |
| GE Vernova | GEV | 1 | −17.97% | −15.84% |
| ASML | ASML | 0 | — | — |
| Airbnb | ABNB | 0 | — | — |
| Broadridge | BR | 0 | — | — |
| CDW | CDW | 0 | — | — |
| Ciena | CIEN | 0 | — | — |
| Fiserv | FI | 0 | — | — |
| PayPal | PYPL | 0 | — | — |
| Rubrik | RBRK | 0 | — | — |
| Vertex Pharmaceuticals | VRTX | 0 | — | — |
Note the spread: Okta (+62 points) and Caterpillar (−17 points) sit at opposite ends. Most companies only have 1–2 independent trades in this window — that's the real sample size per company, and it's small. The portfolio-level average is the only number here worth taking seriously; no single company's row is a reliable prediction.