Prompt library

26 ready-to-use prompts for asking an AI assistant about your marketing attribution data, grouped by use case.

Once SourceLoop is connected at https://app.sourceloop.ai/api/mcp, the quality of the answer depends almost entirely on the question. These are the ones worth keeping. Copy any of them, swap the names and dates, and they work as written.

Five habits that change the answers

Name the window. "Last month" resolves in your workspace's timezone. "Recently" does not resolve at all, and the assistant will pick something.
Ask for the caveat. Adding "and tell me how confident you are given the volume" turns a confident wrong answer into a useful uncertain one.
Say which number you mean. Leads, conversions and closed revenue tell different stories. An unqualified "performance" leaves the choice to the model.
Chain the checks. "Check data health first, then …" costs one extra tool call and removes the most common cause of a wrong conclusion.
Ask for the definition. "State what each number counts" makes the assistant quote our definitions rather than assume ROAS is profit.

The weekly review

What to ask on a Monday, in the order that makes the answers trustworthy. Start with whether the data is sound, then read the numbers.

Check SourceLoop's data health, then give me last week versus the week before: conversions, revenue, spend and cost per lead, split by channel.
What changed most between last week and the four-week average? Only mention things that moved more than 15%, and say whether the change is big enough to be real given the volume.
Summarise last week for a non-marketing audience in five bullets. No jargon, and state what each number counts.
Which campaigns produced leads last week that already have deals attached, and which produced leads with nothing behind them?

Diagnosing a drop

The sequence that gets to a cause rather than a restatement of the symptom.

Our cost per lead went up last week. Work out why: break spend and conversions down by channel, then by campaign inside whichever channel moved most, and tell me whether it was a spend increase or a conversion decrease.
Conversions fell on Tuesday. Before blaming marketing, check attribution coverage and data health for that day and tell me if the drop is real or a tracking gap.
Compare this month's traffic mix with last month's by source and device. Has the mix shifted in a way that would explain a lower conversion rate on its own?
Show me the funnel from first touch to closed won for this quarter and last, and tell me which single step accounts for most of the difference.

Budget decisions

Questions that end in a decision. Ask for the caveat as well as the number, because a confident answer over thin data is worse than an uncertain one.

If I had 20% more budget next month, where would you put it? Use closed revenue rather than lead volume, and tell me how confident you are given the sample size.
Which campaigns have spent more than 2,000 this quarter and produced no closed revenue? Show the spend and the number of leads for each.
Compare cost per acquisition and lifetime value by channel, and tell me the payback period for each. Say which channels are profitable and which are only cheap.
What would our cost per lead look like if I cut the worst-performing 20% of campaigns by spend? Be explicit about what that estimate assumes.

Pipeline and revenue

Marketing questions answered in sales terms, which is the version a founder or a CFO will act on.

Which marketing sources produced the deals we closed this quarter? Use the deal value, not the lead count.
Show me the full journey behind our three largest closed deals this year. What did they have in common?
Which open deals over 10,000 came from paid channels, and what was the first touch for each?
What is our lead-to-close rate by channel this year, and how long does each take to close on average?

Agencies and multiple sites

One connection can cover every client website, so the loop happens in the assistant rather than in a script you maintain.

List every website I have access to, then give me last month's conversions, spend and cost per lead for each, as one table sorted by spend.
Across all my clients, which ones had a worse cost per lead last month than the month before? For each, name the channel most responsible.
Draft a short monthly update for Acme: what worked, what did not, what I am changing next month. Use their numbers and state the date range you used.
Which of my clients have attribution coverage below 80%? Those are the ones whose reports I should caveat.

Data quality

Worth running before anyone builds a decision on a number, and worth running after any change to your site.

Check data health and attribution coverage for the last 30 days. List anything that would make a report misleading, in order of how much revenue it affects.
How much of last month's revenue is attributed to a source, and for the part that is not, what are the reasons?
Are any of our forms or integrations sending fewer conversions than they did a month ago?

Writing outcomes back

These change your data, and in the case of deals your CRM. They need a connection with write permission, and it is worth asking the assistant to confirm before it acts.

The Acme lead closed at 12,000. Record that in SourceLoop, and tell me what it does to paid search's cost per acquisition.
Find contacts from last month's webinar campaign with no outcome recorded, list them, and ask me before changing anything.
Move the Acme deal to Negotiation and confirm the change reached the CRM.

If an answer looks wrong

Ask the assistant to show its working: "which tools did you call, what window did you use, and what does that metric count?" Every response carries its own window and definitions, so it can answer that precisely. Most surprising numbers turn out to be a different date range or a different metric than the one you had in mind, and the rest are usually an attribution gap, which data provenance explains.

For what each tool can do, see the tools reference.