A B2B SaaS demand-signal stack is a small set of leading indicators that, read together, tell you whether demand is rising or falling before it reaches pipeline. It layers four signals: share of search, AI visibility (how often AI assistants mention you), branded search volume, and intent signals. The reason to build one is simple. MQLs are a lagging signal. They tell you demand existed weeks ago, after the moment you could have shaped it. A demand-signal stack moves your attention earlier, to where demand is still forming.
I’m Andrii Byzov, a fractional CMO for B2B tech. Most teams I work with measure demand only after it converts into a form fill, then wonder why the number moves without warning. This is a framework for an early-warning demand system: what goes in it, how the layers fit, and how to read it without fooling yourself.
Key takeaways
- A demand-signal stack reads leading indicators together to see demand forming before MQLs do.
- The four layers: share of search, AI visibility, branded search, and intent signals.
- MQLs are lagging. They confirm demand after the window to influence it has closed.
- Each signal is noisy alone. Confidence comes when several point the same way.
- Treat it as directional: consistent methodology, trends over snapshots, honest hedging.
Why MQLs are not enough
The MQL is a late-funnel artifact. Someone has already found you, formed an opinion, and decided to raise a hand. The demand that produced that action formed earlier, often somewhere you never measured: a search trend, a peer recommendation, an AI answer that named you. By the time it lands as an MQL, the trend is already set.
This matters more now because a growing share of B2B research happens before a buyer reaches your site. They search, they ask AI assistants, they read peer discussion, and only then do they fill a form, if they fill one at all. If your only demand instrument is the MQL, you are measuring the last step of a journey whose direction was decided much earlier. The fix is not to abandon MQLs. It is to add leading signals that move first.
The four layers of the stack
Think of the stack as layers, each measuring demand a little earlier in its life and a little wider than the last.
| Layer | What it measures | Why it leads |
|---|---|---|
| Share of search | Your slice of category search interest vs competitors | Mind-share proxy, documented leader of market share |
| AI visibility | How often AI assistants name or recommend you | Where buyers increasingly research first |
| Branded search | Volume of people searching your name directly | Demand that has already crystallized on you |
| Intent signals | Accounts researching your category or topics | Net-new demand forming at named accounts |
Layer 1: Share of search
Share of search is the proportion of category search interest that goes to your brand versus competitors. Les Binet’s work popularized it as a proxy for mind share and a leading indicator of future market share. For a B2B SaaS team, the practical version is to track branded search trend for you and your main rivals against total category search interest. When your share of that interest rises, mind share is usually growing. When it falls while a competitor’s climbs, that is an early warning worth acting on, often well before pipeline reflects it. Free tools like Google Trends give a directional read; the trend matters more than any single week.
Layer 2: AI visibility
The second layer measures presence where classic search no longer captures the whole picture: inside AI assistants. AI visibility, the metric I describe as share of model, tracks how often ChatGPT, Perplexity, Gemini, and Google’s AI Overviews name or recommend you for buyer-style questions. As more research starts with an AI assistant, being named there is the new being shortlisted. This layer answers a question share of search cannot: when a buyer asks AI which vendors to consider, are you in the answer? If you want a baseline read on this, an AI visibility score is the entry point, and the growth work sits in AI search optimization.
Layer 3: Branded search
Branded search volume is demand that has already crystallized on your name. Someone searching your brand directly has moved past category awareness to specific interest. This is a higher-intent, slightly later signal than share of search, which makes it a useful confirmation layer. When branded search rises after a campaign, a content push, or growing AI visibility, it is evidence that upstream demand is converting into direct interest in you. When it flattens while category demand grows, your share of attention is slipping.
Layer 4: Intent signals
The fourth layer is account-level intent: signals that specific companies are researching your category or the problems you solve. Third-party intent data is probabilistic and varies in quality, so treat it as one input, not gospel. Read alongside the other three layers, it adds a named-account dimension the others lack: not just that demand is forming, but where. The honest caveat is that intent data is the noisiest layer of the four. Use it to prioritize outreach and content, not to make confident claims.
How the layers fit as one system
The power is not in any single layer; it is in reading them together. Share of search and AI visibility are your earliest, widest signals of mind share across two channels. Branded search confirms whether that mind share is converting into direct interest. Intent signals tell you which accounts are moving. A healthy demand picture shows several layers pointing the same way: rising share of search, growing AI presence, climbing branded search, and intent firing at target accounts.
Divergence is the most useful pattern. If share of search and AI visibility climb but branded search stays flat, awareness is not converting into specific interest, and your positioning or mid-funnel content is the suspect. If branded search holds while category share slips, you are coasting on existing demand while losing the next wave. These reads arrive earlier than any MQL trend, which is the entire point. This is the leading-indicator complement to the lagging measures in my B2B SaaS marketing metrics breakdown.
The honest caveats
Every layer in this stack is directional, not exact. Search trends wobble week to week. AI outputs shift between runs and model updates, so AI visibility has to be measured with a fixed prompt set, a steady cadence, and repeated runs reported as a range. Intent data is probabilistic. None of these is a precise score, and anyone selling them as one is overstating.
So read the stack as a trend system. Watch direction over time, look for several signals agreeing, and hold each loosely on its own. Used that way, it gives you an honest early warning that demand is shifting while you can still respond. Treated as a dashboard of exact numbers, it will give you false confidence. The goal is to see the wave forming, not to measure it to the decimal.
FAQ
What is a demand-signal stack? A small set of leading indicators that, read together, show whether demand for your category and brand is rising or falling before it reaches pipeline. It layers share of search, AI visibility, branded search, and intent signals to spot demand forming earlier than MQLs can.
Why look beyond MQLs for demand? MQLs are a lagging, late-funnel signal. The demand that produces a form fill formed weeks or months earlier, often inside search and AI assistants you never measured. Leading signals catch the trend while you can still influence it.
How do share of search and AI visibility fit together? Share of search tracks your slice of category search interest, a documented leading indicator of market share. AI visibility tracks how often AI assistants name you. One measures classic search, the other measures where buyers increasingly research first; together they show whether mind share holds across both.
Is the demand-signal stack reliable? It is directional, not exact. Each signal is noisy alone, so the value comes from reading them together over time and watching whether several agree. Treated as a trend system with consistent methodology it is honest; treated as precise scores it will mislead. If you want help building one, I’m on LinkedIn.