Run an analysis
There is no Run button. An analysis runs when you open its tab, and the result is kept and handed back until the data behind it changes. This article covers choosing a tool and an analysis, setting the period, re-reading after a fix, and how long the slow ones take. If you want to know how to read what comes back, start with Read your first analysis.
Before you start
Permissions required
Nothing on the Analysis tab writes to your tool, so a read-only connection runs everything here.
Run a sync first if you want today’s records in the figures. Open Intelligence Studio in the left-hand menu, then Data sources, find the tool and press Sync.
Choose a tool and an analysis
Open Insights in the left-hand menu, then Analysis.
The page opens on the tool MarqIQ treats as your system of record — the one holding deals, else the one holding the most people — or on the tool you chose last time. Every connected tool with data is a pill along the top; pick another to read it, and the address carries your choice so a link you send opens on the same tool.
Click the analysis you want.
The ten are Overview, Source truth, Field discipline, Forecast accuracy, Audience health, Lead scoring, Funnel, Form fills, Email and Automations. Overview summarises the other nine and adds no finding of its own.
Set the period with the Created filter above the tabs.
Pick a preset, or click Custom, fill in the two dates and click Apply. Click Clear to go back to all time.
- All time
- Last 30 days
- Last 90 days
- This quarter
- Last quarter
- This year
- Last 12 months
Read the sentence under the filter before you read anything else.
It names the period and the counts it holds, such as “People created 1 Jul 2026 – 30 Sep 2026 · 12,400 of 77,662; deals created in the same period · 176 of 715”. The range travels in the page address, so a link you send opens on the same population.
Wait for the read to finish.
The tab loads only when you open it, so the first one you open is the one that waits.
Tip
Open Overview first when you want everything. It runs every analysis that applies to the tool in one pass, and its loading line says so: “Running every analysis for this tool. The first time takes the longest…”
How long each one takes
Two analyses read your tool one record at a time, because a field’s change history is one call per record. They say so while they load.
| Analysis | What it reads | What the screen tells you |
|---|---|---|
| Field discipline | The change history of every field it can test. | “The first read takes up to a minute.” |
| Forecast accuracy | What each closed deal’s forecast said before it closed. | “The first read takes up to a minute.” |
| The results of every send in the window. | No stated time. It reads your tool live. | |
| Automations | Every workflow, smart campaign, automation, journey or flow. | No stated time. It reads your tool live. |
| Form fills | Fills your tracking script captured, and on HubSpot and Marketo your tool’s own forms. | No stated time. The tool-forms half is kept for a day. |
| Source truth, Audience health, Lead scoring and Funnel | Records already synced into your workspace. | No stated time. They do not call your tool. |
Where the screen states no time, none is given here.
Note
A very large store is read as an even sample spread across the whole history rather than the most recent records, and the header says the sample size. The most recent records would make a score look worse than it is and a funnel leakier, because recent leads have had the least time to buy. Form fills is the exception: it reads the most recent 60,000 captured fills and says so on the card when the cap is reached.
Where results are kept
- The four record-based analyses — Source truth, Audience health, Lead scoring and Funnel — build on one read of your synced records, stored with the time of the sync it was read after. A completed sync makes that stored read an earlier computation: it stays on screen, marked with the date it was computed, until the new one is ready. So a sync is what makes the figures move, a little after it finishes.
- Every stored result is keyed on the date range as well, so a filtered result is never handed back for a different period.
- Field discipline, Forecast accuracy and the tool-forms half of Form fills keep their reading of your tool for a day, and the header says the date: “As HubSpot held it on 24 Sep 2026. Kept for a day.”
- Email and Automations keep their reading for five minutes, and the header gives the date they read: “As HubSpot reported them on 24 Sep 2026.”
Re-run after a fix
The analyses that read your tool
Field discipline, Forecast accuracy, Automations, Email and the tool-forms half of Form fills call your tool directly, so they re-read on their own button.
Open the analysis you want to re-check.
Click Read the latest from your tool.
On Email the same link reads Read the latest results from your tool. The button reads Reading… while it works, and the page dims rather than emptying, so you keep the figures you were looking at.
The four analyses that read synced records
Source truth, Audience health, Lead scoring and Funnel read records already in your workspace. They have no re-read button, so a sync is what moves them.
Open Intelligence Studio in the left-hand menu, then Data sources.
Press Sync next to the tool.
Reload the tool’s page when the sync finishes.
What *Recompute* does
Recompute sits on the Executive Read and Data Audit pages, beside their export, and rebuilds both. It does not re-read the analyses. Use Read the latest from your tool on the analysis itself, or run a sync.
The three analyses that read the date range differently
The Created range is your tool’s own created date, applied inside the query. A record with no created date falls outside every range. Three analyses answer a different question, which is what lets you make two of them comparable on purpose.
| Analysis | What the range means there |
|---|---|
| Automations | Nothing. An automation is not a record with a created date, so every one is shown whatever range you set. The filter still appears above the tabs and nothing on the page says it is being ignored here. |
| A send window, not a created window. With no range picked it covers the last 90 days. | |
| Form fills (script-captured half) | The date the fill was submitted. The card reads “fills submitted” and then your period. |
| Everything else | The date the person or deal was created in your tool. |
Take care
To compare Email with a record-based analysis, set the same dates on both and remember you are comparing sends in that window with people created in it. They are different populations by design, not a fault.
If it did not work
The tab spins and then says the analysis could not be loaded
Why: The read from your tool failed — an expired token, a rate limit, or a request that did not come back.
Fix: Click Read the latest from your tool to try again. If it keeps failing, open Connected Tools under Capture & Instrument and check the connection has not asked to be reconnected. Fix a connection that stopped walks through getting a fresh key.
You fixed something in your tool and the numbers have not moved
Why: You are on one of the three analyses that keeps its tool reading for a day, or on one of the four that reads records already synced.
Fix: Click Read the latest from your tool on Field discipline, Forecast accuracy and Form fills. On Source truth, Audience health, Lead scoring and Funnel, run a sync on Data sources and reload.
You clicked Recompute and the analysis is unchanged
Why: Recompute rebuilds the Executive Read and the Data Audit. The analyses each keep their own saved reading and are not part of what it rebuilds.
Fix: Go to the analysis and click Read the latest from your tool. For Source truth, Audience health, Lead scoring and Funnel, run a sync instead.
You set a date range on Automations and nothing changed
Why: An automation has no created date, so the Created filter has no effect on that tab. The filter still renders there, and no line on the page tells you it is being ignored.
Fix: Nothing. Read the counts as covering every automation in the tool. The card’s own note explains that each count runs from the date that automation was created.
Every figure dropped after you set a date range
Why: The range is the created date, and a record with no created date falls outside any range at all.
Fix: Click Clear on the Created filter. Compare the count in the sentence under it with the total, such as “12,400 of 77,662”, before you conclude records are missing.
The counts are lower than your tool reports and the header mentions a sample
Why: Your store is larger than the analysis’s cap. Most analyses then read an even sample spread across the whole population. Form fills is the exception: it reads the most recent 60,000 captured fills instead.
Fix: Read the population line: shares are reliable, and the counts are the sample’s. To bring the whole population under the cap, set a shorter Created range and read the periods one at a time.
An analysis you expected is muted in the row of ten
Why: The row shows all ten for every tool. One the tool cannot feed sits there muted with a dashed border rather than being left out.
Fix: Click it and it gives the reason in a sentence. Where another connected tool can feed it, pick that tool along the top.
Next steps
Related articles
Was this article helpful?