Ad Hoc Reporting and Ad Hoc Analysis Tools That Query Your Database in Plain English
Ask the one-off question. Agentsql reads your live schema, writes the SQL for your engine, runs it read-only and shows you the query. No dataset to publish first.
PostgreSQL · MySQL · Snowflake · BigQuery · read-only · Last updated August 2026
Ask your data a question:
Writing SQL… Running (read‑only)… SQL Agentsql wrote
▋
Click a question. Agentsql writes the SQL, runs it read-only, and answers.
Direct answer
Ad hoc reporting is building a report to answer one specific question on demand, instead of reading a dashboard somebody designed in advance. Ad hoc analysis is the same idea applied to exploring the data rather than publishing the result. The part every guide skips is that in the major BI platforms the ad hoc stops at the edge of a dataset someone already modeled. Tableau documents that connecting to data is a Creator capability, and that an Explorer can build a new view in the browser but cannot save it. Microsoft documents that creating a report from a Power BI semantic model requires Build permission, granted separately from Read. Agentsql removes that step: it connects read-only to PostgreSQL, MySQL, Snowflake or BigQuery, reads your live schema, writes SQL in that engine's own dialect, runs it and shows you the query. A genuinely new question does not need a new modeling ticket.
›_ ad hoc reporting tools, the part nobody prices
Check which seat is allowed to build before you compare features.
Capability columns read from Tableau's and Microsoft's own documentation on 29 August 2026. Prices re-verified the same day.
Every vendor sells ad hoc reporting as self-service: business users answer their own questions, IT stops being the bottleneck. The licence documentation tells a narrower story. In Tableau, the capability to connect to data belongs to the Creator role; Explorer and Viewer cannot reach the interface that connects to a new or existing data source. A plain Explorer can open a published workbook in Web Edit and build a new view, and then has nowhere to save it. In Power BI, Microsoft is explicit that "creating a report based on a semantic model" requires Build permission on that model, and that removing Build leaves a user who "can still see the report, but can no longer edit the report or export underlying data".
Neither of those is a criticism of the products. They are reasonable governance designs. But they mean the phrase "self-service ad hoc reporting" describes something specific: ad hoc inside a dataset that someone modeled first, performed by a seat that was licensed to author. If your question needs a table nobody has published yet, you are not in the self-service path at all. You are in a queue.
| Seat | Can build a new report | Can connect to unpublished data | Can save what it built | Published US price |
|---|---|---|---|---|
| Tableau Viewer | No | No | No | Tableau publishes no role price |
| Tableau Explorer | Yes, in Web Edit | No | No, changes are lost | Tableau publishes no role price |
| Tableau Explorer (can publish) | Yes, from published data sources | No | Workbooks only, not data sources | Tableau publishes no role price |
| Tableau Creator | Yes | Yes | Yes | Tableau publishes no role price |
| Power BI Free | Only in My workspace | In Desktop, cannot share the result | Personal workspace only | $0 |
| Power BI Pro | Yes, with Build permission | Yes, via Desktop | Yes | $14.00 per user per month, billed yearly |
| Power BI Premium Per User | Yes, with Build permission | Yes, via Desktop | Yes | $24.00 per user per month, billed yearly |
| Qlik Cloud Standard | Yes | Yes, within the data allowance | Yes | $825 per month billed annually, users not metered |
| Agentsql | Yes, by typing the question | Any table in the connected schema | Yes | See pricing |
Tableau stopped publishing Creator, Explorer and Viewer prices; the figures widely quoted for those roles come from resellers, not from Tableau, so this table does not repeat them as if they were official. Power BI Pro at $14.00 and Premium Per User at $24.00 are Microsoft's own numbers and are quoted by Microsoft only as billed yearly. Qlik's three published tiers were $300, $825 and $2,750 per month billed annually on 29 August 2026, and the phrase "No cost for additional users" appears on Qlik's pricing page verbatim. Full workings on our Power BI pricing and Tableau pricing pages.
›_ what ad hoc reporting actually costs
The people asking the questions are the cheapest part of the bill.
Arithmetic on rates verified 29 August 2026. No estimates.
Take a 40-person company where 8 people genuinely need to run their own reports. The instinct is that ad hoc reporting costs 8 licences. It does not, because on a per-seat platform the other 32 people need a licence to read the result. At Power BI Pro that is 40 seats at $14.00, so $560 a month, of which the 8 people doing the actual ad hoc analysis account for $112. Roughly 80% of the bill buys read access.
The documented escape from per-viewer licensing is capacity. Microsoft Fabric capacity meters run at $0.18 per CU-hour in westus2 on the Azure retail price API, so an F64 is 64 x $0.18 = $11.52 an hour, which is $8,409.60 a month pay-as-you-go. That is 15 times the seat bill for a company this size. Capacity is the right answer at several hundred readers and the wrong answer at forty, which is worth knowing before a vendor conversation rather than after one.
Seats for the askers
$112
8 people x $14.00 a month. The part of the spend that actually buys ad hoc capability.
Seats for the readers
$448
32 people x $14.00 a month, so someone can open the report the other eight produced.
F64 capacity instead
$8,409.60
Per month, pay-as-you-go, at the $0.18 per CU-hour rate. 15x the seat bill at 40 people.
›_ canned reports vs ad hoc reports
Two different supply models for two different kinds of question.
Most teams need both. The backlog appears when ad hoc demand is served by canned supply.
| Standard or canned report | Ad hoc report | |
|---|---|---|
| Who decides the question | A data team or vendor decided it in advance, often months ago. | The person who needs the answer decides it at the moment they need it. |
| When it is built | Once, then scheduled. Changing it is a change request. | On demand, in minutes, and often thrown away afterward. |
| How long it lives | Indefinitely, which is how dashboards accumulate. | As long as the question does. Most are read once. |
| Cost of the next one | Another ticket, another sprint, another review. | The cost of asking. This is the whole point of the category. |
| Failure mode | It answers last quarter's question accurately. | It answers a slightly different question than you meant, silently. |
The last row is the one worth sitting with. A canned report fails loudly and obviously, by being out of date. An ad hoc report fails quietly, by answering a question adjacent to the one you asked. "Revenue last month" could mean orders placed, orders shipped or payments captured, and could include or exclude refunds, tax and canceled orders. The SQL is valid either way and the chart looks equally confident. That is the argument for a tool that shows you the query rather than only the number, and it is a stronger argument for ad hoc work than for anything scheduled, because nobody reviewed this one in advance.
›_ how to create an ad hoc report here
Connect once, then ask. There is no dataset to publish in between.
The step that traditional ad hoc reporting tools require first, and that this skips, is the modeling step.
01
Connect read-only
Point Agentsql at PostgreSQL, MySQL, Snowflake or BigQuery with a least-privilege user that holds SELECT and nothing else. A replica is the usual choice for exploratory work.
02
It reads the schema
Tables, columns, types and relationships are introspected on connection. This is what replaces the published dataset, and it covers every table you granted access to, not a curated subset.
03
Ask the one-off question
Type it the way you would say it to an analyst. SQL is generated for your engine's own dialect and run against live data, so the answer is current rather than as fresh as the last extract.
04
Read the SQL, then act
You get a chart, a table and a one-line answer with the exact query beside it. Check the WHERE clause, copy it, save it, or hand it to an analyst who can confirm the logic in seconds.
›_ comparing ad hoc reporting software
Six questions worth asking a vendor, in this order.
The first two decide whether the rest of the evaluation matters.
Does a new question need a new data model?
This is the question that separates the category. If the tool can only report on datasets someone published first, then every genuinely new question is a modeling ticket wearing a self-service badge.
Which seat is allowed to build, not just read?
Read a licence table before a feature table. In Tableau, connecting to data is a Creator capability. In Power BI, creating a report on a semantic model needs Build permission, which is granted and revoked separately from Read.
Can the builder save what they built?
A Tableau Explorer can open a workbook in Web Edit and build a new view, and then cannot save it. That is a real ceiling that no pricing page mentions and every evaluation should test.
Does everyone need a seat to read the answer?
Per-seat platforms charge for the audience, not the author. The eight people running ad hoc reports are usually the cheapest part of the bill.
Can you see the query behind the number?
An ad hoc report is by definition unreviewed. Nobody signed off on it last quarter, because it did not exist last quarter. Being able to read the SQL is what makes it safe to act on.
Is the connection read-only?
Ad hoc means people are exploring. Ask what privileges the credential holds. Read-only should be enforced by the database grant, not by a setting in the tool.
›_ who runs ad hoc reports
The heaviest users are rarely the data team.
Finance
Ad hoc financial reporting clusters around close: an unexplained variance, an auditor's request, a board question about one cohort. None of it matches the standard monthly pack, and all of it has a deadline.
Analysts
Ad hoc reporting requests are most of the queue. Handing people a way to run their own query, with the SQL visible, turns producing an answer into reviewing one. More for analysts.
Ops teams
Questions that arrive during an incident are ad hoc by definition, and the dashboard built last quarter will not have the cut you need at 2am. More for ops.
Product managers
Feature questions are one-offs by nature, which makes them the worst possible fit for a pre-built report and the best fit for asking directly. More for product.
Worth being plain about the limit, since the marketing in this category rarely is. Answering your own ad hoc questions does not remove the need for a data team. It removes the routine half of their inbox. Deciding what an active customer means, fixing the pipeline that made yesterday's number wrong and building the model that everything else depends on are still jobs for a person, and they are the jobs worth an analyst's week. What changes is that "can you pull last quarter by region" stops being one of them.
›_ frequently asked
Ad hoc reporting questions, answered.
What is ad hoc reporting?
Ad hoc reporting is creating a report to answer one specific question, at the moment the question comes up, rather than reading a report that was designed and scheduled in advance. Ad hoc is Latin for "for this". The report usually exists to settle a single decision and is often not kept afterward.
What are ad hoc reports used for?
They are used for the questions a dashboard was never built to answer: why a number moved last week, which accounts a pricing change would affect, whether a spike is one customer or many. Anything investigative is ad hoc by nature, because you cannot pre-build a report for a question you have not had yet.
How do I create an ad hoc report?
In a traditional BI tool: get a licence that permits authoring, get Build or Connect permission on a dataset someone has already published, then drag fields into a view. With a tool that reads your live schema, you connect the database once and then type the question. The difference is whether a new question needs a new dataset.
Canned reports vs ad hoc reports: what is the difference?
A canned report is defined once by a data team and run repeatedly, so it is consistent and always slightly out of date with the business. An ad hoc report is defined by the person asking, at the time of asking, and usually used once. Most organizations need both. The mistake is trying to serve ad hoc demand with canned supply, which is what a dashboard backlog is.
What is ad hoc reporting in Power BI?
In Power BI it means building a new report on an existing semantic model. Microsoft documents that creating a report from a semantic model requires Build permission on that model, which is granted separately from Read, and that users without a Pro or Premium Per User licence can only author against models in their own My workspace. So it is self-service within a model, not against your raw database.
How do I create ad hoc reports in Excel?
Most people do it by exporting a CSV and building a pivot table, which works and is why so much reporting still lives in Excel. The costs are that the extract is stale the moment it lands, the logic sits in one person's file, and nobody can audit how a number was produced. A query you can re-run and read solves all three.
What is ad hoc reporting in accounting?
In finance teams it usually means the unscheduled pulls around close: a variance nobody expected, a request from an auditor, a board question about one customer cohort. These arrive with a deadline and never match the standard monthly pack, which is why finance is often the heaviest user of ad hoc analysis in a company.
What are ad hoc reporting tools?
They are tools that let someone build a report without a developer. The category covers full BI platforms like Tableau, Power BI and Qlik, embedded report builders inside SaaS products, SQL clients, and newer natural-language tools that write the query for you. They differ mainly in what has to exist before you can ask.
How do I handle ad hoc reporting requests?
Two things reduce the queue. First, publish the definitions, so people stop asking the analyst what counts as an active customer. Second, give requesters a way to run their own query against real data with the SQL visible, so the analyst reviews an answer instead of producing it. Triage alone does not work, because the volume is a symptom.
Do I need a data warehouse for ad hoc reporting?
Not necessarily. If your data lives in one production database, a read-only replica is usually enough and is far faster to set up. A warehouse earns its cost when you need to join several systems, keep history the source system overwrites, or protect production from analytical load.
Is ad hoc reporting safe to run against a production database?
It is if the connection is read-only at the database level and, ideally, points at a replica. Grant a least-privilege user with SELECT and nothing else, so no query anyone writes can modify data. Running exploratory work against a primary database without a replica is the more common risk, since ad hoc queries are unpredictable in cost.
Which databases does Agentsql connect to?
Four: PostgreSQL, MySQL, Snowflake and BigQuery. Each gets SQL generated for its own dialect rather than a generic string, which matters more than it sounds, since these engines disagree on date truncation, string concatenation and identifier quoting.
›_ how it works
›_ connect your database
›_ what the seats cost
Ask the one-off question.
Connect read-only to Postgres, MySQL, Snowflake or BigQuery and get the answer with the SQL beside it. No dataset to publish first.