Every number on this page comes from a live run of the linked play through the Lusha plugin for Claude between September 9 and 10, 2026. Company names are masked to initials on the play pages. No contacts were revealed.
we ran eight buying signals across sample lists of 25 companies each. The signals were accurate every time. The obvious way to read them was wrong every time. Ranking by percentage put a two job change above a thirty job hiring drive. Ranking by article date put a 2023 funding round at the top of a 2026 list. Trusting a category filter returned 48 posts and 9 usable ones. Five checks fix this, and each one costs nothing because it is a step in the prompt, not a credit.
The order below is the order the checks run in. Each one removes accounts the next one would have sent to a rep.
1. Check when the reading was taken, not when you ran it
Signals refresh on different cycles and none of them is “now.” Hiring by location reads weekly; the newest reading in our run was 10 days old. IT spend and headcount read monthly; the newest was 40 days old. Website traffic returned nothing newer than 84 days.
A rep who opens with “we noticed your traffic jumped this month” from a June reading is wrong on the first sentence. Every play in this series now reports the newest reading date in its first line and warns when it is stale.
Find accounts with climbing web traffic →
2. Rank by the size of the change, not the percentage
The same percentage describes two different companies. In the hiring run, one account showed +33% (two jobs against an average of 1.5) and another +81% (66 jobs against 36). In the headcount run, 4% was 35 people at one company and 275 at another. In the traffic run, a third more visits was 75,000 at one company and two million at the next.
Percentages rank the small base first. Absolute change ranks the investment first. Every play prints both and ranks on the second.
Find companies building a team in your region →
3. Wait for a second reading before you believe an outlier
Five companies in the IT spend run returned budget increases between 227% and 264% in a single month, on technology budgets of $170M to $1.25B. A budget of that size does not triple in thirty days. The first version of the prompt ranked by size and would have sent reps to those five first.
The rule that survived: two consecutive readings in the same direction outrank one large one, and any single reading past a threshold is held as VERIFY. The threshold is not symmetrical. Budgets double more often than they halve, so the increase play flags at 100% and the decrease play at 50%.
Find accounts increasing IT spend and work them first →
Find accounts cutting IT spend and lead with consolidation →
4. Filter the filter
The filters are coarser than their names. “London” resolves to England. “Executive Hire” includes board seats, a school newsletter profiling an alumna, and a hire reported four months after it happened. “Funding Round” returned rounds from 2023 and 2025 that a recent article mentioned. Three intent categories on 25 companies returned 48 posts, of which nine matched the topic in any useful sense.
None of this is the signal being wrong. It is the filter being a net. The fix in each play is a second pass in the prompt: match topic words, not categories, compute days in role from the effective date, not the article date; drop board appointments; hold rumors as unconfirmed.
→ Find accounts with a newly hired executive
→ Qualify your account list by financial events
→ Find accounts posting about your problem
5. Stack two weak signals before you call
Headcount decline alone fired on 13 of 25 accounts, and nine of those were normal attrition under 3%. IT spend decrease alone fired on 9. Run together on the same list, six accounts fired on both, and the only two in real contraction were among them.
One signal tells you something moved. Two signals moving the same way tell you why. The stacked list was less than half the size of either list alone and every account on it had a reason attached.
→ Find accounts cutting headcount
→ Find accounts with stacked buying signals
The order is the whole point
Date, then size, then confirmation, then filter, then stack. Reversing it is how signal programs die: a rep gets a list ranked by percentage from a stale reading, calls the account with the 264% budget jump, is told nothing has changed, and stops trusting the feed.
The signal was right. The reading was wrong. Every check above is a step in the prompt, which means the cost of doing it properly is zero credits and the cost of skipping it is the rep’s trust in the whole system.
Three habits that apply to every signal
Run account_usage first. It costs nothing and it tells you the price of the run before you start. Every play in this series stops if credits are below a floor, so a batch never dies halfway.
Keep the accounts with no signal in the table. If 18 of 25 target accounts show nothing, that is information about the list. A table of hits only hides the question a manager should ask.
Tag customers and prospects before you rank. The same signal is a churn warning on a customer and an opening on a prospect. Budget cuts, headcount cuts and leadership changes route to the account owner for customers and to outbound for everyone else. One tag in the prompt keeps a CSM from getting a cold sequence.
What a run costs
One credit per signal event returned, before any filtering. On 25 domains over 90 days, the eight runs cost between 6 credits (website traffic, five hits) and 48 (LinkedIn activity, 48 posts). Weekly signals cost more than monthly ones because there are more readings; the hiring play now caps at two readings over 60 days for that reason. The play pages list the exact cost of each run.
FAQ
How many buying signals does Lusha have?
Twenty six: 24 company signals (budget, headcount, hiring, traffic, seven news types, LinkedIn activity) and two contact signals (job change and promotion). The plays on this page cover the company signals; the contact signals have their own plays under Prospecting.
Which signal is the most reliable?
Hiring by location, because it reads weekly from job postings and the count is a count. Which signal is the most useful depends on what you sell: for anyone selling to sales teams, an employee posting an SDR job ad on LinkedIn is the cleanest trigger we found, and it came from the noisiest signal.
Why not rank by percentage change?
Because the percentage is a ratio and the denominator is doing the work. A company going from 1.5 to 2 job postings is up 33%. A company going from 36 to 66 is up 81%. Only one of those is hiring. The plays print both numbers and rank on the absolute change.
How fresh is the data?
Weekly for hiring signals, monthly for budget and headcount, daily for LinkedIn activity, and variable for news and traffic. Each play reports the newest reading date in its output and warns when it is older than the signal’s cycle would suggest.
What does “no signal” mean?
Not detected, not “nothing happened.” Some companies sit below the coverage threshold for a signal; some events are not yet in coverage. Treat a no signal row as a question about the account, not an answer.
Should I run all 24 signals on every list?
No. Pick the two or three that describe the moment your buyer is in, run them on a list you already have, and stack them. The check signals on target accounts play runs the full set when you need to see everything once; after that, narrow.
Go deeper
- Check buying signals on your target account list — all 24 signals on one list, once
- Build a weekly signal digest on your saved territory — the recurring version of everything above
- Find which signals are actually converting to pipeline — the check that closes the loop
- Find recently funded companies to prospect — the structured alternative to the funding news signal
- Every play in the library — sorted by the problem rather than the feature
One data connection. Works in Claude, ChatGPT, your CRM, or any agent you build. See what Lusha’s plugin covers
