Standing service ten-data-questions-answered-every-month-with-working-and-a-second-check · revised 12 October 2026
Standing service
Get up to ten data questions answered each month, with working and a second check
Send up to ten questions a month with your export. Each answer comes with its query, row counts, a second check by a different method and caveats, or a clear "cannot be answered from this data".
This starts a conversation by email. Nothing is charged, and nothing is monitored, until we have agreed scope, access and terms with you in writing.
The responsibility you hand over
Open-ended "ask the data anything" services fail on plausible wrong numbers, and a retainer with no limit has no way to say what was delivered. A fixed number of questions, each with its working, a second method and caveats, gives you something to check and a clear month's output.
Who it’s for: Owner or manager with a monthly export and a stream of "can you tell me" questions that wait for a busy analyst, or for nobody.
Usually starts when: The same kinds of question come up every month, you have used chat tools and been burned by a plausible wrong number, and you want a person-sized limit on the work.
The result: Each month you send up to ten questions about the export you supply, and every answer comes with its query, row counts, the file checksum, a separately written second method and caveats. If the methods disagree or the data cannot answer, the answer is withheld with the reason. Questions are about what the data shows, never advice or forecasts; supplied reasons are quoted as your statements, not verified causes.
What stays true, and what we do about it
No response-time guarantee is published for this new service. A date for each monthly report is agreed in writing before it starts.
Hours are agreed in writing before the service starts. It runs once a month and is not staffed round the clock.
What must remain true
- Every answer given in the month comes with its query, row counts, file checksum and a second method that agrees
- No answer gives a forecast or recommendation; reasons from your file are identified as your statements, not verified causes
- Both control questions are answered correctly, or the month's report says they were not
What we watch
- The monthly export and its checksum
- The questions and your definitions
- The control questions and your reasons file
When something happens
Scroll the table sideways to read it all.
| When | What we do |
|---|---|
| Your monthly export and questions arrive. | We answer each question with the query, counts and a second method, and send the summary. |
| The two methods disagree. | We withhold that answer, show both results and say what we think differs. |
| A question cannot be answered from the data. | We say so, name the missing field and suggest the export that would answer it. |
| A control question is answered wrongly. | We flag it first and stop giving answers on that dataset until the cause is fixed. |
We do on our own
- Run queries and pivots on the export
- Write the second-method check
- Answer or decline each question with reasons
- Write the summary
We ask you first
- Any question beyond the monthly ten
- Any request for causes, forecasts or advice
- Any change to a rule, setting or page in your own systems
- Any wider access than the exports named here
- Contacting your customers or anyone outside your team
We escalate to you when
- A control question is wrong
- The export shows signs of personal data
- A definition conflicts with an earlier month's
- Two methods keep disagreeing on one question
How you know it held. After the report is frozen, you re-run up to 3 numeric answers, all if fewer, with the supplied queries and matching checksum; with no numeric answer, each question has a stated cannot-answer reason.
How we keep it true
This service is never finished. Each month's report shows whether the things it watches stayed true, and it continues until you end it.
Let me ask my sales spreadsheet ten recurring questions and trust the answers Job Optional
Ten named questions become saved, checked queries on your export, each matching your hand-worked answer on the snapshots we tested. Rephrased questions go to the right query or are refused.
Bought separately, each time it is needed; not part of the monthly fee
The one-off job that turns ten recurring questions into saved queries you can run yourself.
Define our key numbers once so everyone, and any AI tool, gets the same answer Job Optional
Up to ten metrics are defined as code on one schema, with a 30-question set that reconciles to figures you sign off. Five planted questions outside the definitions must be refused.
Bought separately, each time it is needed; not part of the monthly fee
The one-off job that fixes your key definitions so answers agree.
Tell me when orders or refunds suddenly change, and which product or channel moved Job Optional
For up to five daily metrics, a script you run writes an alert when a value leaves its usual seasonal range, naming the products or channels behind it. Figures recompute; no cause is claimed.
Bought separately, each time it is needed; not part of the monthly fee
The one-off alert job, if you want to be told when a number moves.
What is included, and what is not
- Responses to up to ten questions a month: checked answers or an explicit cannot-answer reason
- The second-method results
- The summary of disagreements and unanswerable questions
- The control question result
Included
- Up to ten questions a month about one dataset you export (up to 1 million rows, personal data removed by you)
- For each answer: the query or pivot, the row counts, the file checksum and a second result by a separately written method
- "Cannot be answered from this data" with the reason, where that is the case
- Two control questions a month with answers you already know, which you add and which are not counted in the ten
- A monthly summary of answers, disagreements and open questions
Not included
- More than ten questions a month: extra questions wait for next month
- Forecasts, advice, recommendations or verified causes; supplied reasons are only quoted as your statements
- Live access to your systems or dashboards
- Personal data, user-level logs or special-category data
- Building dashboards, models or pipelines
- Round-the-clock cover or a response-time guarantee
How we know it’s done
Agreed with you before work starts. Each check produces evidence you keep.
After our answers are frozen, you choose up to 3 to re-run, all if fewer, from the query and matching input checksum; all give the same numbers within the agreed rounding. With no numeric answer, every question has an explicit cannot-answer reason; one wrong checksum or unreproducible number fails the whole test.
Evidence: Your numbers beside the answers.
Both control questions match the answers you withheld until our results were frozen. One wrong control answer fails the month's checks, is reported first and stops further answers on the dataset until its cause is fixed; this two-case smoke test makes no general accuracy claim.
Evidence: The control question result.
Every answer given has a second method that agrees, and every answer where the methods disagreed is withheld with both results shown.
Evidence: The second-method table.
Your named reviewer reads every answer and confirms that no forecast or recommendation is given and any reason from your reasons file is quoted as your stated reason, not a verified cause. One unsupported causal statement, forecast or recommendation fails the whole test.
Evidence: Your named reviewer's marks for every answer.
Sign-off. You read each report and re-run the acceptance checks with the files and steps supplied. A month counts as delivered only when those checks pass.
If it fails. If we cannot do a month because an export is missing or has changed, we say so and do not guess. If the service is not working for you, you can end it at the end of any month.
When it fits, and when we stop
It fits when
- A monthly export you can send through a secure route, with personal data removed
- Questions written down, each with the definitions you want used
- A named person who reads the answers and checks the controls
We stop and tell you if
- The export shape changes and no sample is sent
- A control question is answered wrongly: further answers on that dataset stop until the cause is fixed
- Most questions are about causes or forecasts
What could go wrong
The service reads exports and reports, so there is nothing to undo. Ending it leaves your systems exactly as they are; you keep every report.
Scroll the table sideways to read it all.
| Risk | How we handle it |
|---|---|
| A plausible wrong number is delivered. | Each answer needs a second method that agrees, two control questions are checked every month, and you can re-run any answer yourself. |
| The scope grows without limit. | Ten questions a month, extra ones wait, and questions about causes or advice are declined. |
| Exports carry personal data. | You remove personal data before sending, and the provider, its country and retention are named in writing first. |
| Definitions drift between months. | Definitions are written down and any conflict with an earlier month is flagged. |
Each report is checked by a separate automated reviewer before hand-over. A person on your side checks the control questions and decides what to do with each answer. No human supervisor is included unless your agreement names one, and at launch the work is largely automated.
Stays with a person
- You decide what to do with every answer
- You set and change the definitions
Access we would need
- A monthly export of one dataset through a secure route
- No live system access of any kind
Questions
Is an AI answering my questions?
Automated systems do the work, a separately written check confirms each answer, and a separate automated reviewer reads the result. No human analyst from us is included unless your agreement names one.
What if I have eleven questions?
The eleventh waits for next month. Ten is the limit.
Can you tell me why sales fell?
We can quote a reason you supply, marked as your statement rather than a verified cause. The one-off cause job proposes possible causes with supporting data.
How is this different from the ten-questions job?
That job builds saved queries once. This is a monthly service for new questions on fresh exports.
Send an enquiry
Send us
- What the dataset holds and roughly how big it is
- Examples of the questions you ask
- Who would send them
- No logins, keys, customer data or payment details in the first enquiry
Later, once you agree
- The monthly export through a secure route
- Up to ten questions plus two control questions; you keep both control answers from us until our results are frozen
- Your definitions, such as what counts as a customer or a refund
- A reasons file, if you want your stated reasons quoted
You keep your systems and logins; we receive the agreed export and return answers and files, with no warehouse or dashboard access. You remove personal data first; before sharing, the terms state any model service, its country, retention and training status, and when our export copies are deleted.
Email fallback: open your mail app
If website submission is unavailable, review and send the fallback email yourself. An email fallback is not a website receipt. Or write to hello@syntheticindustry.ai with “ten-data-questions-answered-every-month-with-working-and-a-second-check” as the subject.