B2B Prospecting: The Difference Between a Filter and a Signal

Two teams build a list of 500 accounts in the same week. The first filters for industry, headcount band and geography, and gets a list that will return almost exactly the same 500 companies next quarter. The second filters for companies that posted a specific job in the last thirty days, and gets a list that will be completely different next quarter.

Both lists are 500 accounts. Only one of them contains a reason to write this week, and that difference is what most prospecting advice skips past on its way to talking about subject lines.

A filter describes a state, a signal describes an event

This distinction is the whole article, so it is worth making precisely.

A filter is a condition that is true for a long time. Two hundred employees, headquartered in Texas, running Shopify, in the logistics sector. Filters answer the question of who could plausibly buy, and they are stable by design. Stability is what makes them useful for defining a market and useless for deciding who to contact on a Tuesday.

A signal is something that happened, and it has a date attached. A company posted a role, replaced a VP, announced a funding round, added a technology, opened a location, launched a product line. Signals answer a different question: who has a reason to reply now.

You need both. The filter tells you an account is worth talking to at all. The signal tells you this week is a better week than an arbitrary one. Teams get into trouble when they build a list from filters alone and then try to manufacture urgency in the copy, which produces the whole genre of cold email that opens by asserting a problem the sender has no evidence the recipient has.

Filter-built listDescribes a durable state

• Returns nearly the same accounts every quarter
• No reason attached to contacting anyone this week
• Everyone in your category can build the identical list
• Message has to assert a problem rather than reference one
• Never expires, so it gets reused until it is exhausted

Signal-built listDescribes a dated event

• Completely different set each period
• The reason to write is the reason the account is on the list
• Depends on where you look and how often
• Message can reference something the recipient did
• Expires, which forces the list to be rebuilt

Four questions that separate a signal from noise

Not everything with a timestamp is worth building a list around. Run any candidate signal through these before it earns a place in the motion.

Is this signal worth building on

• ✓Yes: Can you observe it yourself, or are you buying it from someone who also sold it to your competitors
• ✓Yes: Does it name a person, or only an account
• ✓Yes: Does it imply a problem you actually solve, or just activity
• ✓Yes: Can you act on it inside the window where it is still true
• ✓Yes: Would your message read differently without it

The last question is the sharpest one and it takes ten seconds. Write the message the signal justifies, then delete the signal reference and read it again. If the message still works, the signal contributed nothing and you have added cost for no reply. That test kills more candidate signals than the other four combined.

Question two matters more than teams expect. Account-level signals tell you a company is doing something and leave you guessing which of forty people to write to. Person-level signals name someone. A job posting for a demand generation manager tells you the demand generation function exists, is under-resourced, and probably reports to someone you can find. A topic surge across an account tells you somebody in a building of four hundred people read some articles.

The signals that are cheap and observable

These need effort rather than budget, which is exactly why they stay useful. A signal anyone can buy gets bought by everyone in your category.

Hiring. A posted role is a budgeted, approved, publicly declared problem. It names the function, frequently names the tools in the requirements, and carries a date. It also tells you something about direction: three roles in the same function is expansion, one replacement role is churn, and the two justify completely different messages. Read the company's own careers page as well as the job boards, because the careers page is usually more current.

Leadership change. A new head of a function rewrites their stack, their agency roster and their process in the first two quarters, because that is what they were hired to do. This signal is unusually good because it comes with a built-in window and a named person, and because incumbency is at its weakest.

Technology adoption or removal. Visible in page source, in job requirement lists, and in public integration directories. Adoption tells you a budget line exists. Removal tells you the previous answer failed, which is a far stronger signal and much less commonly used.

Expansion. A new office, a new market page, a new product line, a pricing page that changed. Each implies operational strain somewhere, and the strain is usually predictable from what you sell.

Public statements of intent. Conference speaker lists, podcast appearances, a post about a problem, a funding announcement on the company's own newsroom. Somebody stood up and said what they are working on. Taking them at their word is legitimate and it is not surveillance.

The signals that cost money

Paid signals are not worse. They are worse value when bought before a motion exists to consume them, which is the usual sequence.

Third-party intent feeds tell you an account is consuming more content on a topic than its own baseline. That is genuinely useful as a prioritisation layer and it is account-level, refreshes on a schedule, and gets sold to your competitors as well. What it can and cannot predict, and what the providers charge, is covered properly in the B2B intent data guide.

Website visitor identification tells you an anonymous visit came from a company. It fires on your own team, on bots, and on people with no purchase intent whatsoever, so suppression lists are mandatory rather than advisable.

Technographic databases at scale, and enriched firmographic or hiring APIs, are worth paying for once you have proven the signal works when you gathered it by hand. The connection method and refresh rate matter as much as the data, because a nightly pull is by definition a day stale, and some signals do not survive a day. The landscape by connection type is in intent signal APIs for outbound.

The sequence that works is to prove the play manually on twenty accounts, then buy the API that removes the manual step. The sequence that wastes money is to buy the data first and then look for a play, which produces a dashboard nobody opens and a renewal conversation nobody enjoys.

Noise dressed as intent

Five things get sold or self-reported as buying signals and mostly are not.

Anything that resolves to firmographics. "Companies in your ICP showing activity" often decomposes to companies in your ICP. If removing the signal leaves the same list, it was a filter with a marketing name.

Aggregate topic surges with no person and a weekly refresh. Useful for prioritising an account list you already have. Weak as a reason to contact a specific individual this week.

Engagement on your own social posts. A like is an act of politeness far more often than it is an act of evaluation. Treating a reaction as a hand raise produces awkward outreach and burns a warm audience.

Aged content downloads. A whitepaper download from nine months ago is a filter now. The event happened, the window closed, and the person may have changed jobs.

Any signal you cannot act on inside its window. This is the general case of all of the above. A signal you respond to three weeks late is a filter that cost you money, and a slow response is by far the most common way good signal data produces bad results.

Building the list from the signal

The mechanical order matters, because doing it backwards produces the filter-built list with extra steps.

1. Step 1Pick one signal and write the message first

   Write the message that only makes sense if the signal is true. If it reads fine without the signal, choose a different signal.

2. Step 2Set the window

   Decide how many days after the event the message still makes sense, and build the expiry into the list rather than into somebody's memory.

3. Step 3Find the person the signal implicates

   The signal usually points at a specific function. Write to that person, not to whichever title your standard filter returns.

4. Step 4Apply firmographics as an exclusion pass

   Run the filters last and only to remove: wrong size, wrong region, existing customer, live opportunity, competitor.

5. Step 5Send once, then let the list expire

   One message per campaign. The reason to write existed in a particular window, and a follow-up after it closes has no reason attached.

Step five is house doctrine and it earns its place here specifically. We run one message per campaign with no thread replies and no bumps, and a signal-built list is the clearest argument for that policy. The whole justification for the message was something that happened on a date. Two weeks later the hiring manager has been hired, the new VP has picked their vendor, and a bump lands underneath an ignored message with nothing new to say. Rebuild the list from this period's signals instead. That is a better use of the same effort and it is the reason signal-led prospecting produces work rather than saving it.

Referencing the signal without being creepy

There is a clean line and it is about whether the company published the thing. A job posting, a funding announcement, a conference talk, a new product page: the company put those in public deliberately, and referencing them is ordinary attentiveness. A page visit, an anonymous session, a content consumption score: the person did not choose to tell you, and saying so out loud converts a useful timing signal into a reason to block your domain.

The practical rule is to let unpublished signals decide timing and let published ones appear in the copy. Write to the problem the signal implies rather than to the signal itself, and the message reads as relevance rather than as monitoring.

Where the ICP still does the heavy lifting

None of this replaces knowing who you sell to. A signal narrows a market you have already defined, and applied to an undefined market it just produces a smaller pile of the wrong accounts. If the profile is a personality description rather than a set of filters that returns a count, the exclusion pass in step four has nothing to run on. Building one that actually changes the target list is covered in how to build an ICP, and the boundary between a lead worth passing to sales and one worth nurturing is in MQL versus SQL.

The short version

Filters describe a durable state and signals describe a dated event, and only the second gives you a reason to write this week. Test a candidate signal on five questions, and lean hardest on the last one: delete the signal reference from your draft, and if the message still reads fine, the signal added nothing. Hiring, leadership change, technology adoption and removal, expansion and public statements of intent are observable without a budget, which is what keeps them valuable. Paid feeds are worth buying after a manual version of the play has worked, never before. Build the list by writing the message first, setting an expiry window, finding the person the signal implicates, applying firmographics last as an exclusion pass, and sending once.

RevenueFlow builds signal-led campaigns and is paid on attended meetings that meet criteria agreed in writing before launch. You can see what a campaign would look like for your market.