Designing signals that actually qualify

A signal is only as good as the action it triggers. Too broad and you drown in noise. Too narrow and you miss the buyer who didn't fit your template. Most "intent data" platforms ship with hundreds of signals turned on by default, which is roughly equivalent to having an alarm that goes off every time anyone opens a door anywhere in the building.

The teams getting the most out of Orcha treat signals like product features: defined, versioned, and measured against outcomes. Start by writing the buying moment in plain English, then translate it into the data points your pipeline actually has access to.

Start with the buying moment, not the data

The mistake most teams make is starting with the data they have access to and working backwards. "We have firmographic data, technographic data, and intent data, so let's build a signal off those." The output is a signal that fires whenever a company in your ICP visits a category page, which describes roughly half the internet.

Instead, start with the moment. Write a sentence that describes the exact instant a buyer goes from "not in market" to "actively evaluating." For most B2B products, this moment is triggered by a change: a hire, a funding round, a tooling switch, a strategic announcement, a regulatory deadline, an executive transition. The signal is the change, not the steady state.

Then ask: what is the smallest set of data points that would let me detect that change with high precision? The answer is almost never "intent score above 80."

Three signal patterns that work

After looking at hundreds of customer signal configurations, three patterns consistently outperform the rest. Each one starts with a change, layers a confirming data point, and ends with a clear next action.

The funded-and-hiring signal

A company raises a Series B. Within 30 days they post three or more sales-related job openings. This combination is a near-perfect predictor that they are about to expand a go-to-market motion and will be evaluating tooling to support it.

The naive version of this signal fires on the funding event alone, which gets you in front of every vendor pitching every newly funded company in your space. The refined version waits for the hiring confirmation, which cuts the volume by 80% and roughly triples the response rate. The message writes itself: you are not pitching cold, you are pitching at the exact moment they have decided to scale the function you support.

The tech-stack-shift signal

A prospect removes a competing product from their site, job posts, or public stack. Within 60 days, they will likely be evaluating replacements. This is the highest-converting signal we see, but it is also the easiest to mishandle. Reps see "they removed X" and immediately pitch a swap, which lands as opportunistic and slightly creepy.

The version that works leads with curiosity, not assumption. Reference the change as context, ask a real question about what they are trying to solve, and let the conversation reveal whether the swap is real. The signal got you the meeting. The conversation closes the deal.

The executive-change signal

A new VP of Sales, RevOps leader, or CRO joins the company. New executives have a 90-day window in which they are explicitly looking for tools and processes to mark as their own. After day 90, the window closes and the inertia of "what we already have" takes over.

The signal is simple. The hard part is the timing. A message that arrives in week one looks like spam. A message that arrives in week six lands in the middle of their planning cycle. Most teams default to "fire immediately on the LinkedIn update," which is exactly the wrong answer.

Versioning your signals

The teams that win at this treat signals like code. Each signal has a name, a definition, an owner, and a measured conversion rate. When the rate drops, somebody investigates whether the signal stopped predicting the moment, or whether the message that pairs with it stopped landing.

Most teams skip this step. They turn on a signal, send a few sequences, and never measure whether the signal is actually predictive of pipeline. Six months later they have a signal library full of dead weight, and they assume the problem is the AI writing the emails.

It is not the AI. It is the signals nobody pruned.

The test

For each signal in your library, answer two questions. First: if this signal fires, can a rep write a meaningfully different message than they would write to a generic ICP-fit account? Second: in the last 90 days, what was the meeting-booked rate on this signal versus your baseline? If the answer to the first is "not really" or the answer to the second is "the same," delete the signal.

A small set of signals that genuinely predict the buying moment beats a giant library of signals that fire on everything. The volume feels good. The pipeline does not.