Make the case for what to build, from messy customer interviews.
Nala analyses every interview the same way and counts people, not mentions. Each claim in your case links to the words behind it, so anyone can check it.
Rebuilding the month-end report by hand came up with the most people, 8 of 10.
01What they talked about
People rebuild the same report from exported spreadsheets every month, and the rebuild is the job they want gone.
Nala’s summaryCheck every step, including Nala’s.
The same questions, every passage
Every passage Nala reads gets the same four questions: is it evidence, what kind (experienced, routine or hypothetical), how did they feel, which theme? Nothing is skimmed, and any passage in a theme traces back to its transcript.
People, not mentions
Ten mentions from one customer is still one customer. Every count shows its n, and a point only one person made stays a lead to test, never a finding.
Claims you can check
Every claim in a case keeps its quotes, so anyone reading it can open the words behind it.
Weigh each theme against everyone you heard from.
Every theme shows how many people raised it, out of everyone you heard from, and how many didn’t. A count tells you how far a pattern reaches; your research questions decide what matters. Open any count to the quotes behind it.
Rebuilding the month-end report by hand came up with 8 of 10 people.
“Every month I export four CSVs and glue them into the same deck. It’s the same deck.”
“I rebuild it from scratch because I don’t trust last month’s links.”
“I don’t trust a number unless I can see where it came from.”
“Finance asks where it came from and I genuinely can’t say.”
“Quarter end is when it all falls apart.”
“I want the board pack done by Tuesday, not Friday night.”
Read each interview, person by person.
A theme shows how far a pattern reaches; a snapshot shows what it meant to one person. Nala writes each person up as opportunities and insights, every line citing the passages it came from, and the transcript opens right there, so you quote a real customer in the meeting, not a paraphrase.
Write up a decision, with the evidence attached.
Type # and a theme drops in with its count. If a line has no source, or claims more than the study shows (“most users” from 3 of 10), Nala flags it in the margin. Share it as a link or as Markdown, and every claim opens to its quotes.
Do we build saved month-end views before a spreadsheet sync?
By you · edited just now
What we heard
Most of the people we spoke to rebuild the month-end report by hand Rebuilding reports by hand8 of 10.
“Every month I export four CSVs and glue them into the same deck. It’s the same deck.”P3 · 14:32
“I rebuild it from scratch because I don’t trust last month’s links.”P1 · 08:40
All 8 quotesNothing in the study backs this line yet. The nearest theme:
Rebuilding reports by hand · 8 of 10 people
They don’t trust last month’s links, so they start again from the exports every time.3 quotes from 3 people may back this⌘⇧E
Check a chatbot’s summary against the interviews.
A chatbot’s summary reads well and can’t be checked: “users overwhelmingly want…” with no one behind it. Check both against the interviews and see which claims hold.
How Nala reads an interviewUsers overwhelmingly want automated reporting.No source Most participants expressed frustration with manual processes, and many highlighted trust in data as a key concern.How many? Prioritising a spreadsheet sync would likely drive adoption.Who said so?
Nothing to open. No quote, no person, no count behind any of it.
Rebuilding the month-end report by hand
8 of 10 people“Every month I export four CSVs and glue them into the same deck.” P3 · 14:32
Doubting any number without a source
6 of 10 · 2 disagree“I don’t trust a number unless I can see where it came from.” P7 · 08:05
Asked for a spreadsheet sync
2 of 10“A Sheets export would be nice, I guess, but it’s not the problem.” P4 · 31:40
Ask about your interviews from Claude or ChatGPT.
Connect Claude or ChatGPT over MCP. Your AI reads the counts and quotes and drafts suggestions; Nala checks every number, and only a person publishes.
Connect your AI
Your team’s AI connects over MCP, with a key for each member.
Your AI asks, Nala answers
Questions go out and the counts come back with their quotes, each with its link.
Suggestions, never edits
Whatever your AI writes lands as a suggestion. An editor keeps it or dismisses it.
Every claim checked
Nala checks each number against the interviews. A claim with no quotes behind it says so.
Pay for the interviews Nala reads.
Read every study yourself. Share the case as a link.
VAT included. Billed monthly, cancel any time.
- 5 interviews a month, unused ones roll over up to 15
- Brief, People snapshots and cases, unlimited studies
- Share links
- Your own AI over MCP
- No comments, viewers or guests
Write the case together, comment by comment.
For 3 editors. VAT included, billed monthly.
- 10 interviews per editor, pooled: 30 a month on 3
- Comments beside the part they question, and asks by name
- Unlimited viewers, and guests who comment from a share link
- Everything in Solo, with an AI key per member
- +£9 a month for each extra editor
Solo plus 2 packs: £19.99 a month for 12 interviews.5 come with the plan, packs of 5 are £5, and the 3 left over stay in your balance.
An interview is one customer in one session; the interviewer never counts. Each 10,000 words one person says counts as one, up to three. Re-runs are free. Need more? 5 interviews for £5.
Before you start.
Does an AI write my findings?
No. Every judgment is a fixed question with a typed answer (is this evidence, what kind, which theme), and the counts are computed in the database. A model may word a theme’s name or summary, citing its quotes, and you can switch that off. The argument is always yours.
How does this compare to doing thematic analysis myself?
Nala does the systematic part: every passage coded with the same questions, gathered into candidate themes and counted by person. You do the analysis: combine themes, move quotes between them, group sub-themes, and decide what they mean for your question. How many people said something is one input, never the verdict.
Will I get the same answer twice?
Yes. Every judgment is cached and each study keeps its model version, so the same interviews give the same themes and counts, run after run.
Why count people, not mentions?
Ten mentions from one customer is still one customer. Every number shows its n, like 6 of 10.
What counts as an interview?
One customer in one session. The interviewer never counts, a focus group counts each customer, and each 10,000 words one person says counts as one, up to three. Re-runs are free.
What can I bring?
Transcripts (.txt, .md, .docx, .vtt, .srt), calls from Fathom or Granola, or a pasted transcript. You pick which speakers are customers.
Can I use my own Claude or ChatGPT?
Yes, on both plans, over MCP. Your AI reads counts and quotes with their sources and drafts suggestions; Nala checks every number. Only a person publishes.
Where does my data live?
In London. Names are masked by default, original files are deleted after 30 days, and you can download or delete everything at any time.
Why is there no free trial?
The screens on this page are the real interface. Pay for your first month, and if Nala isn’t for you, ask for a refund within 14 days.
Who is Nala for?
Product people in small teams (PMs, designers, researchers and founders) with interviews to read and a decision to make this week. If you need a repository across many studies, or the study itself is the deliverable, a research repository suits you better.
Bring this week’s interviews.
Cancel any time, and ask for a refund within 14 days of any payment. Your data stays in London.