See what one field actually holds
The Field Intelligence page profiles every field on every object your tool has: how full each one is, what values it holds, when it was last filled, whether it is still being filled today, and — on HubSpot — what kind of thing last wrote it. Use it when a number looks wrong and you need to know whether the field behind it can bear the weight. If you want the judged verdict on your source field rather than the raw profile, read Can you trust the field you report channels from? instead.
Before you start
- One finished sync. The profile reads MarqIQ’s synced copy of your records, not a live call to the tool.
- For field labels and types rather than raw keys, a sync that saved the tool’s field catalog. Without one the profile still runs and shows the key.
- For “who fills it in”, HubSpot, with a token that can read property history.
Every field, not a chosen list
There is no fixed list of fields MarqIQ knows about. Every property present on your records is profiled, and a role — lead source, stage, amount, owner and the rest — is worked out from the field’s own name and the values it holds together. That is why anything reported here can be reproduced by filtering your own tool.
Open the profile
Open Insights in the left-hand menu, then Field Intelligence.
The same profiles are on the Integration Audit Report at /audit, which Starter workspaces can reach.
Pick the tool along the top.
Choose the object along the top.
Each of the tool’s own objects is profiled separately and named the way the tool names it: Contacts, Deals and Companies on HubSpot; Leads, Contacts, Opportunities and Accounts on Salesforce; Persons, Leads, Deals and Organizations on Pipedrive.
Open All fields.
Type the field’s name or its API key into Find a field.
The search matches both, so “lead source” and “LeadSource” find the same row.
Tick show technical fields if the field does not appear.
Bookkeeping keys the tool writes for itself are hidden by default. They are still profiled and still searchable once shown.
What the row tells you
| Column | What it is | Measured on |
|---|---|---|
| Field | The label the tool shows, with the API key underneath when the two differ | — |
| Used for | The role this field plays, when it has one — Lead source, Stage, Amount, Owner and so on. A bookkeeping key reads “technical” | The field’s name and its values together, not the name alone |
| Holds | What kind of thing is in it: text, category, date, number, yes / no, email, link, ID or empty | The values themselves |
| How complete | The share of the profiled records carrying a value | Every record read for this object. A blank counts as empty, and so does a record holding the literal text “null” |
| Different values | How many distinct values it holds. A trailing + means there are more | Up to 400 distinct values are tracked per field; past that the + appears |
| Last filled | The creation date of the newest record carrying a value | The tool’s own created date on the record |
| Most common values | The three commonest, with counts. Hover for six | The records read |
Read the scope line before any percentage
Above the sections is one line giving how many records the profile was built on, how many are in the store, and the created dates the read spans. Every percentage on the page is a share of the first of those numbers. A fill rate with no denominator beside it is half a fact.
How many records are read
Deals and companies are small; contacts carry far more data each, so they have a tighter cap.
| Object | Records read at most |
|---|---|
| Deals | 20,000 |
| Companies | 10,000 |
| People — contacts, leads, prospects, subscribers, members | 6,000 |
A store no larger than its cap is read whole. Above the cap, an even spread of the whole store is read instead — one record in every so many, picked by a fixed rule so the same records are read on every visit and the figures do not move between two looks at the same data.
Note
That matters more than it sounds. Reading “the newest 6,000” would make a field your team started filling last spring look fully filled and a field they stopped filling look empty. An even spread describes the store, which is what a fill rate is meant to describe.
Is it still being filled, or only on old records?
A field at 62% can be a field filled on everything for two years and then abandoned. That is why the profile keeps two fill rates as well as the overall one.
- The newest fifth of the records read, by the tool’s created date, gives the Filled on the newest records figure.
- The oldest four fifths give the comparison beside it.
- Both need at least ten of the records read to carry a created date. Below that neither figure can be worked out, and both are blank rather than guessed.
- Where an attribution field — source, medium, campaign, content, term or landing page — sits at 50% or more on the older records and at least 40 points lower on the newest, that is raised as a finding in its own right, tagged easy to miss, because nothing on the screen looks broken.
Who fills it in
This is the one that finds an integration quietly overwriting your source field. HubSpot keeps the kind of thing that wrote each value — a person typing, a form, a workflow, an import, an API, the Salesforce sync, a bulk edit, a calculation — and MarqIQ reads it back as a share.
It appears as a Who fills it in line on the Fields we rely on table, reading something like “64% a workflow, 22% typed in by a user, 14% an import”.
| What it needs | The limit, exactly |
|---|---|
| The tool | HubSpot only. No other tool exposes per-value write history in a form this can read |
| The records | The 100 newest records of that object. A history read per record per field is the expensive shape of this API |
| The fields | Up to 15, and only the fields a role hangs off — the ones your report’s figures depend on |
| The field names | HubSpot’s own lower-case internal names. A column MarqIQ folded in under its own spelling cannot be asked for |
| Freshness | Kept for the day against those exact records and fields. A new record in the sample, a different field, or tomorrow, and it is read again |
Every figure states that it came from the 100 newest records. Where the token cannot read history, the page says so in Report notes and shows no provenance rather than a guess. Five findings come straight out of this read: a field only people ever type into, a field only an import ever filled, a field several hands write with no rule about who wins, a field whose writer has stopped, and — under Working well — a field a system reliably sets.
Take care
A 100-record sample of your newest records answers “what writes this field today”. It does not answer “what wrote this field in 2023”. For a field’s history across the whole store, the Field discipline analysis is the screen that reads change histories rather than current values.
The field MarqIQ reads for each role, and changing it
Fields we rely on lists each thing the report needs to know about a record and names the one field it reads for it. Expanding a row gives the evidence behind the pick.
- Field name in the tool — the API key, so you can find it in your own admin.
- Last filled on a record created — with the first date beside it.
- Filled on the newest records — with the older-record figure beside it.
- Who fills it in — on HubSpot, as above.
- Most common values — the five commonest.
- Other fields with a similar meaning — the alternatives, best first. Nothing is hidden because it lost.
- Why this one — the reason it won, which is a mix of how confidently it matched, how full it is and how recently it was filled.
Open the row for the role that is reading the wrong field.
Pick the right field under Change it.
The list holds every field that has data on this object, fullest first. Not tracked here is the honest answer when nothing on this object plays that role, and it is remembered as a decision rather than a gap.
Wait for the page to re-pull.
Your choice is saved against this tool and this object, and every figure on the audit is recomputed from it — the profile, the data audit and the suggested plan together.
Tip
Your choice beats what MarqIQ inferred, permanently, until you change it back. Leaving a role unset is not the same as choosing Not tracked here: the first lets the engine keep deciding, the second tells it to stop.
Check any finding in your own tool
Findings sit under What we found, in four bands: Fix first, Worth a look, Good to know and Working well. A report that cannot be reproduced in the source system is an opinion, so every finding carries the way to reproduce it.
Click See how to check this yourself under the finding.
Read the filter it gives you.
Build that filter in your tool.
Compare the count against Expected count.
It prints as a count and a denominator — “412 of 5,980” — so a mismatch tells you which of the two is off.
Check the values listed under Values involved against what you see.
Note
A Fix first finding on this tab also puts a red dot on the tab itself, so you can see one is waiting without opening it.
If it did not work
The tab says nothing to flag on this object, but you know a field is thin
Why: Findings fire on thresholds. A field can be thin without crossing one, and the profile is the place that shows it either way.
Fix: Open All fields and read the field’s own row. How complete and Last filled answer the question directly, whether or not a finding fired.
A field you use every day is missing from All fields
Why: Either the field is empty on every record read, or it is classed as a technical key, or the sync did not bring it through.
Fix: Tick show technical fields first. If it is still absent, check whether the field is being synced at all — a field the sync never pulled cannot be profiled.
A field shows its API key instead of its proper name
Why: The tool’s field catalog has not been saved by a sync for this object yet, so there are no labels to show.
Fix: Run a sync from Data sources. Labels, types and whether each field is custom or standard arrive with the catalog.
The Different values count ends in a plus sign
Why: Up to 400 distinct values are tracked per field. Past that, the count is “400+” and the commonest values shown are still the commonest.
Fix: Read it as “more than 400 distinct values”, which on a picklist field is itself the finding: a field with hundreds of distinct values is free text, whatever it was meant to be.
No “who fills it in” line anywhere
Why: Write provenance is HubSpot only, and needs a token that can read property history.
Fix: On HubSpot, read Report notes — a refused history read is named there with HubSpot’s own status. On every other tool the line is absent by design.
The profile has not changed after you fixed records in the tool
Why: The profile is stored and reused until the object’s records change or you change a field-role choice. It follows the newest write to those records, so a correction that has not synced yet is not in it.
Fix: Run a sync, then reload the tab. On the Integration Audit Report, Re-pull does the same thing.
A fill rate here disagrees with the one on the Integration Audit Report
Why: They measure different populations on purpose. This tab profiles each of the tool’s own objects separately within the caps above; the audit’s field completeness scans every synced contact as one object.
Fix: Read the denominator each screen prints. Where the store is under the cap and the tool has one people-object, the two agree.
Next steps
Related articles
Was this article helpful?