♾️ Mining sales & CS data for ideas

You’re reading The Content Loop — a 5-minute read on how B2B SaaS marketers can use original research and product-led content as a growth lever.

Did someone forward this email to you?

Every time content teams have to come up with an idea, I’ve seen them run a keyword search.

I’m not against leaning in on SEO tools—especially if SEO is a huge priority for your strategy. But even then, those seed ideas should come from real conversations you’re having with your audience.

Every content asset you create should be rooted in the fact that your audience needs that information. Otherwise, what’s the point?

But most teams treat these sales and CS data like they’re locked in a vault somewhere.

One hill I’ll die on is that we’re not tapping into our first-party data enough these days.

We have critical audience data locked away in sales calls, demos, and CS tickets that could fill your content calendar for months. That's why I spend each month going through these calls to make sure we’re always on course for speaking to the right problems.

Here’s my process:

Step 1: Listen to the calls and review tickets

I listen to sales calls—regardless of the number.

Listen to at least 15 to 20 conversations if you’re building something from scratch. Or ask to sit in on live conversations with sales and CS folks. You can also review customer support tickets or email chains to which you might get access.

I highly recommend this because, as much as we’d love to automate these processes, there’s nothing like actively listening to the people you’re marketing to.

The nuance you pick up from tone, hesitation, excitement, frustration—you can’t get that from a transcript or summary.

Don’t just skim call summaries or rely on your CRM notes. The magic is in the actual conversations—the pauses, the “ums,” the moments when someone gets genuinely excited about a feature. You miss all of that in a summary.

Step 2: Document recurring patterns

I jot down the most common themes and take notes on things like:

• Role of the person

• Type of company (firmographics, demographics, team specifics)

• Current state (what they’re doing now)

• Current workflow

• Workflow-specific challenges

• What they want to achieve (dream state)

• What they’ve tried so far

• Potential objections

• Buying triggers (why you and why now)

• Who they’re evaluating against you (consideration set)

• Specific quotes for each category (Voice of the customer)

Don’t skimp out on the quotes. In fact, if you’re running it through an AI prompt like the one below, you’ll get the exact phrases your audience uses. You can use that on your website, ads, and cold outbound campaigns to show your audience that you “get” them.

When someone says, “We’re drowning in spreadsheets” or “Our current tool makes me want to pull my hair out,” that’s your next headline.

I’m not looking for specific ideas here. But I want to categorize them to make sure I have the most prevalent themes and category, product, and audience-specific insights in one place first.

AI prompt to extract VoC:

<SYSTEM>
You extract Voice-of-Customer (VoC) from sales/CS transcripts into tab-separated rows for Google Sheets. Use ONLY what is explicitly stated or clearly implied by the interviewee (customer/ICP). No assumptions. Preserve exact phrasing.
</SYSTEM>

<TAXONOMY>
Allowed categories (use EXACT spelling; one category per row):
 
- Current state (what they're doing now)
- Current workflow
- Workflow-specific challenges
- What they want to achieve (dream state)
- What they've tried so far
- Potential objections
- Buying triggers (why you and why now)
- Consideration set (who they're evaluating against you)
</TAXONOMY>

<OUTPUT_QUOTA>
- TARGET rows per category: MIN=5, MAX=15.
- If there are more than 15 relevant quotes for a category, keep the 15 strongest/clearest.
- If there are fewer than 5 full-sentence quotes for a category:
  1) Use **micro-splitting**: from any long utterance (≈18+ words), extract 2–4 **non-overlapping** exact sub-spans (clauses/phrases) that each stand alone semantically. Keep them verbatim.
  2) Use **cross-categorization**: if a quote cleanly maps to multiple categories, create separate rows (one per category).
  3) If still below 5, continue micro-splitting different, non-overlapping spans from the strongest available utterances until MIN=5 is met.
- Never fabricate content. Never paraphrase inside the VoC field.
</OUTPUT_QUOTA>

<OUTPUT_RULES>
- Format: **TSV** with real TAB characters (U+0009) between fields.
- Print a **single fenced code block** with language tag `vbnet` and NOTHING else before or after it.
- Header row (exactly this, with tabs):
Company	Interviewee / Role	Category	VoC (direct quote)	Timestamp	Takeaways	Intended use case
- Columns:
  • Company — leave blank if unknown. 
  • Interviewee / Role — e.g., “IT Manager, Manufacturing”; leave blank if unknown. 
  • Category — one from the allowed list.
  • VoC (direct quote) — exact words from the interviewee; for micro-splits, keep ellipses only if they appear in the transcript. Do not add your own “…”.
  • Timestamp — mm:ss or hh:mm:ss if available near the quote; else blank. For micro-splits from the same utterance, reuse the same timestamp.
  • Takeaways — brief, objective insight grounded in the quote (1 sentence; no speculation).
  • Intended use case — short suggestion like “Landing page proof”, “Objection-handler post”, “Case study angle”, “Ad copy”, “Sales enablement slide”.
- Cleaning:
  • Do NOT wrap fields in quotes.
  • Replace any tabs/newlines inside fields with a single space.
  • Keep punctuation as-is.
</OUTPUT_RULES>

<PROCESS>
1) Sweep the transcript once to collect candidate utterances per category.
2) De-duplicate within each category (drop rows with ≥85% wording overlap).
3) Apply micro-splitting and cross-categorization as needed to reach 5–15 rows per category.
4) Output the complete TSV table for all categories in one block.
</PROCESS>

<TASK>
Produce rows per the rules above. Prioritize clarity and raw customer phrasing. Only include interviewee (customer/ICP) content, not the seller’s.
</TASK>

<CONTEXT>
Company: [optional]
Interviewee / Role: [optional]
</CONTEXT>

<TRANSCRIPT>
[PASTE FULL TRANSCRIPT HERE]
</TRANSCRIPT>

<FINAL_RENDERING>
Render ONLY one fenced code block tagged `vbnet` containing the TSV table.
</FINAL_RENDERING>

Step 3: Categorize everything in a spreadsheet

Next, I organize all this intel in a spreadsheet. Ideally, it should come out in a way that you get the following:

• Individual’s name

• Designation

• Quotes under each category

Here’s an example:

Now, you have a whole sheet with the exact phrases your audience uses, categorized by their dreams, fears, frustrations, and objections.

Download this data as a CSV, and you can use it to run different analyses.

Step 4: Turn insights into content ideas

I use that spreadsheet to generate content ideas for the calendar. I usually use the CSV files to do a few things:

• Understand the most recurring themes

• Understand why folks are coming to us

• Understand their perspective on the problem

You can use this file to revamp your website, create ad copy, OR long-form content assets.

Here’s how I usually find content ideas:

• Which themes emerge the most

• Sort based on frequency

• Check if it’s something you actually solve with the product

• Prioritize those themes/titles for production in the next 90 days

One conversation theme can become 4-5 content pieces across multiple channels.

Let’s say you notice that most of your recent prospects or customers are wondering how long implementation takes or struggling through the implementation process. Now, you know that you could create pieces on:

• How to realistically adopt the product in XYZ scenario

• How different customers spearheaded adoption

• Common mistakes your CS team sees during implementation

• X things teams must have in place to adopt your product

Always tap into data you own first to make sure you act on the things that matter to YOUR audience—not what some keyword tool thinks might matter to someone, somewhere, maybe.

P.S. Liked the issue? Share it with someone who could benefit from it.

That’s all for today!

As always, if you have any questions or feedback, hit “Reply” and let me know. 😄