12 Questions to Ask Your Outbound Data Once You Connect Spear to Claude

TL;DR

Most outbound platforms ship a dashboard and call it analytics. Dashboards answer the questions you already thought to ask. The real wins are buried under questions you didn't know to ask: "why is my new optimized campaign underperforming the old one I shut down?", "who are the 15 prospects I denied as competitors last quarter?", "which trigger actually books meetings, not just generates suggestions?". When your outbound platform connects to Claude via MCP, those questions become possible. Here are twelve worth asking the day you connect.

The dashboard problem

We built Spear's dashboard the way every B2B SaaS company builds a dashboard. KPIs at the top. Funnel chart in the middle. Campaign list at the bottom. Filters everywhere. It looks like the product is doing its job.

‍

But here is what we noticed watching customers (and ourselves) use it. People check the dashboard. They nod at the numbers. They close the tab. The dashboard answers the questions they walked in with. It almost never surfaces the questions they should be asking.

‍

That is the failure mode of every dashboard ever built. The interesting questions are the ones you have not formulated yet. A grid of metrics cannot prompt you to formulate them.

What changes with MCP

The Model Context Protocol lets you connect your tools (Spear, your CRM, your calendar) to an AI assistant like Claude. Spear's MCP server exposes your campaigns, prospects, conversations, and analytics as functions Claude can call directly. No copy-paste. No SQL. No new dashboard.

‍

What you get is not "AI for sales" in the usual sense. You do not type a prompt and watch generic content stream out. You ask a real question about your real data, and the assistant goes and fetches the answer from your live account.

‍

The shift is subtle but big. You stop scanning a dashboard for things that look off. You start asking questions out loud, the way you would ask a smart analyst, and getting answers grounded in your own pipeline.

The 12 questions

We pulled these from a live conversation with Spear's own data. Every one of them returned something I would not have noticed on the dashboard. Use them as a starting point. Once you ask the first three, the next ten will write themselves.

‍

1. Which of my active campaigns are underperforming the ones I shut down?

‍

This is the most painful question to ask and the most valuable. Run it on a year of campaign history and you will find that one of your "optimized" rebuilds is converting worse than the scrappy original you replaced. The winner is usually doing something specific that got lost in the rebuild: a tighter ICP, a sharper opening line, a trigger you stopped using. Find it. Put it back.

‍

2. Where exactly is my funnel leaking?

‍

Every team obsesses about booked meetings. That is the right thing to obsess about, but it hides the diagnostic question. If you are booking less than you should, the leak is somewhere upstream, and you need to know where. Connection acceptance is the foundation: if nobody accepts, nothing downstream happens. Reply rate is the next gate. Booked is the outcome. Ask the assistant to break the funnel down stage by stage, per campaign, and only look at campaigns with at least 100 sent messages (anything less is noise, not signal). The leak is almost always concentrated at one specific stage, and the fix depends entirely on which one.

‍

3. Which triggers actually book meetings, not just generate suggestions?

‍

Most outbound platforms report trigger volume: how many prospects came from a LinkedIn post, a job change, a funding round. Volume is not value. The right question is conversion: of all the prospects sourced from each trigger, what percentage turned into a meeting? You will often find your highest-volume trigger is your lowest-converting one, and a quiet trigger like Podcast or Funding Round outperforms it five to one.

‍

4. Who are the prospects I denied as "competitor" last quarter?

‍

Every denial in your account is free competitive intelligence. The list of companies your team flagged as competitors over the last 90 days is a map of who you are bumping into in the wild. It is more current than any analyst report, and it is sitting in your account waiting to be queried.

‍

5. Of the 91 campaigns in my account, which ones are dead weight?

‍

If you have used your outbound tool for more than six months, your campaign list looks like a graveyard. Test campaigns, untitled campaigns, abandoned experiments. Ask the assistant to surface every campaign with zero activity in the last 90 days. Archive them. The dashboard signal gets ten times stronger.

‍

6. Which of my winning campaigns can I rebuild as an always-on play?

‍

Some campaigns are situational: a conference, a product launch, a one-off list. Others are repeatable patterns that worked once and got forgotten. A "viewed your profile" campaign that booked 35% of replies is not a one-time win. It is a category of play. Find the winners, identify what made them work, and convert the best of them into permanent infrastructure.

‍

7. What did the winning variant of an A/B test do differently?

‍

Most teams run informal A/B tests: same audience, two messages, see which one lands. Then they forget to do the analysis. Ask the assistant to compare the two campaigns side by side: ICP filters, message steps, timing, denial reasons. The diff between a 5% and 18% booked rate is usually three or four specific choices.

‍

8. Was the experiment worth the effort?

‍

Voice messages, video, custom GIFs: the industry pushes a new format every quarter. Most of them sound good in a podcast and underperform in production. Ask the assistant to compare your experimental format against your text-only baseline on the same audience. The honest answer is often "no", and that is fine. The point is to find out before you build a whole workflow around it.

‍

9. Which of my recent existing-network sends actually engaged?

‍

When you message 900 of your existing connections, the reply rate is a noisy aggregate. The interesting question is who replied with intent. Pull the list. Look for patterns: seniority, company stage, industry. That cohort is your real warm audience, and it is usually smaller and more specific than you think.

‍

10. Are my scheduled event campaigns actually ready to launch?

‍

Event campaigns are the easiest to ship and the easiest to forget. You build the ICP, set the dates, and move on. Then the conference starts and nothing is happening because the ICP loaded zero prospects or the campaign never activated. Ask the assistant to audit every scheduled campaign the week before it goes live: is it loaded, is it active, is it healthy.

‍

11. Who do we already know at our target accounts?

‍

Your team's combined LinkedIn network is the most underused asset in your GTM stack. Before you cold-outbound into an account, ask the assistant who already has a first-degree connection there. Warm intro paths convert at multiples of cold outbound. The data is sitting in your account. Almost nobody queries it.

‍

12. What is the one question I have been avoiding?

‍

This is the meta-question. Every founder, every sales leader, every CS manager has a number they do not want to look at. Churned customers from a specific cohort. The campaign you championed that flopped. The trigger you bet on that never worked. Ask it. The whole point of having the data conversational is that the cost of asking a hard question drops to near zero.

The pattern underneath

The twelve questions are not really about Spear. They are about what happens when your tools stop being dashboards and start being interfaces you talk to.

‍

The dashboard is a posture. It says: here is what I think you should care about, arranged in the order I think matters. Most of the time it is wrong, because most of the time the interesting question is not on the dashboard.

‍

A conversational interface flips the posture. It says: what do you actually want to know? That is a more honest framing. It also produces better answers, because the question is yours, the data is live, and the loop between "I wonder" and "here is the answer" is one message long.

‍

We built Spear's MCP server because we wanted to use Spear that way ourselves. Now we do.

Frequently Asked Questions

What do I need to ask these questions?

‍

A Spear account and Claude (or any MCP-compatible AI assistant). You connect Spear as an MCP server inside Claude, then ask the questions in plain English. No setup beyond the OAuth flow.

‍

Will Claude make changes to my account, or just read?

‍

Both, if you let it. Spear's MCP exposes read and write actions: querying analytics, creating campaigns, approving prospects, even sending messages. Start in read-only mode. Once you trust what it is doing, open up write actions.

‍

How is this different from a chatbot in the Spear app?

‍

A chatbot is a UI on top of the same data. MCP is a protocol that lets the AI assistant of your choice (Claude, ChatGPT, or anything else that speaks MCP) talk directly to your Spear account. You are not locked into a particular interface. You bring your own assistant.

‍

What if I do not use Spear?

‍

The questions still apply. Any outbound platform with an MCP server (or an API and a little glue) can be queried this way. The interesting shift is the interface, not the vendor.

‍

Is this just for analytics, or can I actually run outbound this way?

‍

Both. You can ask analytical questions ("which triggers convert best") and execute actions ("approve all 14 of these prospects into the Snowflake Summit campaign"). The second is where it starts to feel less like a query tool and more like a teammate.

‍

Where do I start?

‍

Pick the question on the list that scares you the most. Ask it. See what you learn. The other eleven will suggest themselves.

‍