The Buying Signal Testing Playbook
Works with
Implementation kit
- The signal shortlist, with how crowded each common signal is
- The signal card, one per signal, with its retirement rule
- The test plan, signed before the first message goes out
- The pre-send check and the readout
- A worked example, a filled card for one signal
Your team acts on buying signals, and nobody can say which ones mean an account is ready. This playbook shortlists two or three, writes down what must still be true for each, tests them against accounts that do not show them, and lets a rule signed in advance decide which to keep.
The full method is below. A shortlist, one card per signal, a check before every batch, and a test sized to the volume you actually send.
A message about a moment that has passed tells the buyer you did not look.
The failure this playbook fixes
How it works
01Shortlist two or three signals
Few signals, each one visible from outside the account. List every signal you could act on, then rate each on three things. Can you see it before the conversation, how many teams already act on it, and what does checking it cost?
Most teams chase the same few signals, like content engagement and website visits. Put at least one on the list that fewer teams watch, such as tech changes or public records specific to your market.
Watch for signals you can only see on the call. Those are discovery questions and stay off the list.
02Write the fact that must still be true
One card per signal. The event you see, the one fact that would make it meaningless if it stopped being true, where to check that fact, and how long the signal stays fresh.
Write the opening line the signal makes possible, and the rule for when the signal retires. Write the retirement rule now, while nobody has a stake in the signal.
03Split the list and sign the rule
The signal arm gets accounts that show the signal. The comparison arm gets accounts from the same band that do not, with your usual message. Keep the sender, channel, days and follow-ups the same.
Plan on 100 accounts per arm. Before the first message, sign how sure you need to be to keep the signal. We use a 90% chance that the signal arm really replies more than the comparison.
Watch for calling it early. A small test can still be read, it just gives a wide range, and a lucky week read as a sure thing is how a hunch becomes a strategy.
04Check every fact before every batch
Before a batch sends, check each account's fact at its source. Send only when it is still true and inside its fresh window. "Cannot tell" is its own answer, and the account is held.
Then read the rendered message, the one the tool will actually send. When we read every message before one load, 8 of 129 contacts weren't employees of the company the message was written for.
05Send both arms and count
Count replies per account contacted and meetings booked, for each arm. Keep the pre-send tallies too, because they show how often the signal goes stale or cannot be checked.
06Read it against the signed rule
Look your counts up in the test plan's table. 9 replies of 100 against 3 of 100 is about a 96% chance the signal arm really replies more, which clears 90, so keep. 6 against 3 is about 84%, more likely than not but under the line, so fix the message once or retire. A gap with something else that could explain it means hold for another period.
Report the chance and the range together. That 9 against 3 read puts the real gap anywhere from about 1 to 8 times, so the direction is known and the size is not.
07Scale what earned its place, test the next
A kept signal scales with the pre-send check still on every batch, and gets read again in a quarter. A retired signal makes room for the next one on the shortlist.
A signal that works today can stop working once more teams act on it.
Check the signal before the send, and let your own numbers decide which ones stay.
The implementation kit above packages the shortlist, the signal card, the signed test plan, the pre-send check, the readout and a worked example.
