To improve lead quality in LinkedIn outreach on the day a client complains, do six things in order: read the leads the client called bad, find the criterion that let them in, run a prospect preview, mark fit with the client on a call, change the targeting on a draft copy of the campaign, and remove the wrong leads from the old one. With the GTM Bud connector in Claude or ChatGPT, that is one afternoon of work, and the client sees the new list of people before the day ends.
This article is for an agency account manager with one unhappy client. It is not a general guide to targeting. It starts at the angry email.
The email arrived at 10:12 on a Wednesday. Ines is account manager at an outreach agency. Her client, a firm that sells compliance training to mid-size banks, forwarded three LinkedIn profiles with one line: “These are not our buyers. If this is the quality, we stop at the end of the month.” The three people were a branch manager at a retail bank, an HR business partner at an insurance broker, and a “Head of Compliance” at a 12 person fintech.
First hour: What do the three leads have in common?
Read the three leads before you answer the email. A reply sent at 10:15 defends the campaign. A reply sent at 11:15 explains it.
Ines opened the chat and typed:
- “List the leads that the campaign ‘Aldgate Training | Compliance heads banks UK’ contacts. Find Peter L., Hannah S. and Omar T. and show their title, company and LinkedIn URL.”
- “Show me the current audience description and every criterion of this campaign, with must-have or nice-to-have.”
The lead list returns the name, the job title, the company, the LinkedIn URL and the status of each lead, up to 25 on a page. It has no search by name, so the AI reads pages until it finds the three. That is slow for a large campaign and fine for a few hundred leads.
One limit matters here. A lead in the list does not carry the reason it matched. The result for each criterion exists for people in a prospect preview, not for leads that are already in the campaign. So the next step is done by reading and thinking.
Ines put the two answers side by side:
| Lead | What the client objects to | Criterion that should have stopped it | Was it there? |
|---|---|---|---|
| Branch manager, retail bank | No say over training budget | Role owns compliance or risk | Yes, but nice-to-have |
| HR business partner, broker | Wrong industry and wrong function | Company is a licensed bank | No. Criterion said “financial services” |
| Head of Compliance, 12 person fintech | Too small to buy | 250 or more employees | Yes, must-have. Fintech page listed 300 |
Three leads, three different causes. That is the usual picture, and it is why “tighten the targeting” is not a plan.
- A must-have was written as a nice-to-have. A nice-to-have is a preference and does not reject anyone.
- A criterion was too wide. “Financial services” includes brokers, insurers and payment firms.
- The data was wrong, not the rule. The company page of the fintech claimed a headcount it does not have. No criterion fixes a wrong company page.
At 11:20 Ines answered the client: “You are right about all three. I found two rules that were too loose and one company with wrong public data. I will show you a new list of people at 3 pm on a call. 20 minutes.” No defense, one time.
Is it three leads or the whole campaign?
Check a sample before the call, because the client will ask. Three examples can be three accidents.
“Show me 25 leads that this campaign contacts, with title and company. Mark each one that is not a head of compliance, risk or financial crime, and each one whose company is not a bank.”
The AI marked 7 of 25. Five of the seven had the same fault as Hannah S.: the wrong kind of company. So the campaign has one real problem, the industry criterion, and one smaller one, the role criterion.
Stop the damage while you fix it: “Stop lead finding for the Aldgate campaign.” A stop of lead finding applies at once. The campaign adds no new leads and keeps sending to the leads it has, so the conversations with good leads go on.
Can you change the targeting of a launched campaign from the chat?
No, and it is better to learn that at 11:30 than on the client call. The targeting tool of the connector changes a draft campaign only. For a launched campaign GTM Bud refuses the change and tells the AI to give you the campaign link, because the targeting of a live campaign changes only on gtmbud.com.
That leaves two real paths:
| Path | How | Good when |
|---|---|---|
| Edit on gtmbud.com | Open the campaign link and change the audience in the app | A small change, and you do not need a before and after |
| Draft copy from the chat | Duplicate, change the copy, preview, launch, retire the old one | The client must see and approve the new audience first |
For an unhappy client, take the second path. A draft sends nothing, so you can show the client real people from the new audience before a single message goes out. It also gives you a clean comparison for the next report: old campaign, new campaign, two sets of numbers.
“Duplicate the Aldgate campaign. Name the copy ‘Aldgate Training | Compliance heads banks UK | v2’.”
The copy is a draft. It keeps the targeting, the message sequence, the message settings, the sending accounts, the daily targets and the reply playbook of the original. It does not keep the leads, the replies, the results or the previews. If the copy should send from another LinkedIn account, change that on gtmbud.com.
How do you rewrite the criteria so the same kind of lead is rejected?
Rewrite one criterion per cause, and write each as something a LinkedIn profile, a company page or public web content can show. Ines gave the AI the table from the first hour and asked:
“Update the targeting of the v2 draft. Start from the current criteria. Change three things: role in compliance, risk or financial crime becomes a must-have. Replace ‘financial services’ with ‘the company is a bank or building society, not an insurer, broker or payment firm’, must-have. Keep 250 or more employees as a must-have.”
The update replaces the audience description and all criteria of the draft, so the AI starts from the current ones and sends the full new set. GTM Bud checks each criterion and saves the result. All of Ines’s criteria passed. If a criterion had asked for something that no profile or company page shows, such as “has a training budget for this year”, the check would flag it with a message and a suggested rewrite.
For the third cause, wrong public data, there is no rule to write. Ines asked the AI what evidence could help, and the AI proposed a nice-to-have: “the company appears on the public register of authorised banks”. She kept it as a preference. An honest note for the client: a criterion that reads public data can still be fooled by a wrong company page, and some leads like the fintech will always need a human eye.
Do not over-correct. Every extra must-have rejects people. A person that fails one must-have is not a match. An angry client makes you want six new rules. Three leads justified two.
The 3 pm call: How do you mark fit with the client?
Show people, ask yes or no, and type the client’s own reason. Start the preview ten minutes before the call, because the search takes one to three minutes: “Start a prospect preview for the v2 draft.”
On the call Ines shared her screen with the chat open. The client’s head of sales, Rupert, joined.
“Show the preview of the v2 draft, eight people at a time.”
Each person in the preview has a name, a role, a company, a location and a short reason for the match. Rupert went through sixteen people in twelve minutes. Thirteen were a yes. For the three that were not, Ines typed what he said:
- “Mark Farah K. as not a fit. Reason: deputy money laundering reporting officer at a credit union, too small to buy. Some people like this are acceptable.”
- “Mark Colin B. as not a fit. Reason: works at a bank but in a group audit role, never buys training. Exclude everyone like this.”
- “Mark the thirteen others as a good fit.”
For one person Rupert hesitated. “Open the full profile of Shreya M.” showed her experience, the company description with its headcount, and the result for each criterion. He said yes.
GTM Bud keeps each mark and each reason with the campaign. The reason for Colin B. gave Ines one more line for the role criterion, “not internal audit”, and she updated the draft after the call and started a new preview to check.
Why do this live and not by email? Because a client who said yes to thirteen named people has approved the audience in a way that no criteria document achieves. Write it into the client notes: date, who, how many of how many. A preview is one list per run and each campaign has a daily limit on previews, so do not burn runs on small wording changes before the call.
Last hour: Relaunch and clean up
Three jobs remain, and the order matters.
1. Launch the copy. “Get the launch readiness of the v2 draft, then launch it.” The copy kept the seller profile, the value offers, the sequence, the messages and the sending accounts, so the checklist is usually met. The launch takes two calls. The first shows a preview with the accounts that send and the first message of each channel, and Ines confirmed it in the chat.
2. Remove the wrong leads from the old campaign. “From the old Aldgate campaign, list the leads it contacts whose company is not a bank or whose role is not compliance, risk or financial crime. Show me the list before you remove anyone.”
The AI listed 31. Ines took four off the list, because their titles were unusual and the people were right. Then: “Remove the other 27.” Removing a lead takes it out of the campaign and stops all messages to that lead in that campaign. It is done one lead at a time, so 27 removals take a few minutes. Before a removal, check for a reply: two of the 27 had answered, and Ines closed those conversations by hand first.
3. Decide what happens to the old campaign. Lead finding is already stopped. Ines left the sending on, so that the good leads still get their follow-ups. When the last conversations have gone quiet, she ends it. Ending a campaign is permanent, it needs a preview and a second confirmation in the chat, and the history stays available.
At 16:40 she wrote to the client: “v2 is live with the audience you approved at 3 pm. 27 people who did not fit are removed from the old campaign, and it sends them no more messages. Old campaign: no new leads. First numbers for v2 in Friday’s report.”
What this method cannot do
- It does not explain a past match. You see who a lead is, not why the campaign chose that person. The reasons per criterion exist in the preview only.
- It does not edit a launched campaign in place from the chat. The copy is the path, or the app.
- It does not undo sent messages. The three people in the email already got one. Removal stops the next ones.
- It does not fix wrong public data. A company that misstates its size will pass a size rule.
GTM Bud has no white-label dashboard and no client workspace, so the client never looked at the lead list alone. The client saw three messages go out, and then saw sixteen people on your screen. For an agency that sells outreach under its own name, that is the service: you are the one who reads, decides and shows.
If the complaint repeats for a different segment, the campaign is probably serving two audiences at once. Split it with the method in clone a LinkedIn campaign for five segments. To stop the complaint before it starts, get the client’s approval of real people on day one, as in onboard a new outbound client in one day.
Frequently asked questions about how to improve lead quality in LinkedIn outreach
How many bad leads mean the targeting is wrong?
Three forwarded leads prove nothing by themselves, and they are still a reason to check. Read a sample of 25 leads that the campaign contacts and count how many the client would reject. One or two in 25 is normal noise. Five or more with the same fault means that one criterion is missing or too loose.
Should I stop the whole campaign while I fix the targeting?
Stop lead finding, not sending. A stop of lead finding applies immediately and the campaign keeps its conversations with the leads it already has, so interested people still get answers. Then remove the wrong leads one by one. A full stop also halts the follow-ups to good leads.
Can I change the audience of a campaign that is already launched from the chat?
No. The targeting tool changes a draft only, and for a launched campaign GTM Bud refuses and gives the link to the campaign on gtmbud.com. From the chat, the path is a draft copy: duplicate the campaign, change the audience of the copy, preview it, launch it, and stop lead finding on the old one.
What do I say to the client on the same day?
Three facts and one date. What the three leads had in common, which rule let them in, what you changed, and when the client will see the new list of people. Do not defend the old targeting and do not promise a number of meetings.
Does removing a lead delete the conversation with that person?
Removing a lead takes it out of the campaign and stops all messages to that lead in the campaign. It does not take back a message that was sent. If the person already replied, read the conversation first and answer by hand if a polite close is needed. Do it before you remove the lead.
From an angry email at 10 to an approved audience at 3
You improve lead quality in LinkedIn outreach by finding the exact rule that failed, not by adding rules until the client is quiet. Read the leads, name the cause for each, fix a draft copy, let the client say yes to real people, relaunch, and remove the ones that did not fit. To work this way, add the connector by its address with the steps on the GTM Bud connector page, and keep the Monday check from run client LinkedIn campaigns from one AI chat so that the next problem shows in your table before it shows in a client email.