Two screens show different numbers for the same thing

Why: They are different rates. Win rate on the Data Audit tab is measured on deals: won over won plus lost, with open deals in neither side. The rates in the contact tables are measured on people who reached customer stage, over everyone carrying that value. A conversion rate to customer is almost always the smaller number.

Fix: Read the column heading. A deal-grain figure says "win"; a contact-grain figure says "convert". Compare like with like, or quote both and say which is which.

Why: The CRM & Sources tab reduces submissions to one row per person and matches them by email against your CRM contacts, so its Unified Profiles figure counts people. A record count counts rows, and duplicate emails are separate rows.

Fix: Read the Duplicate emails tile on the Data Audit tab. Counts there treat each record separately, so per-person figures elsewhere fold those repeats together.

Why: Focus filters choose which records get enriched, normalised, scored and reported on. Reports honour them. The Data Audit describes the whole synced store, so it reconciles against what you see in your tool.

Fix: The Data Audit tab says so when they are on: "Focus filters are active in Intelligence settings — reports honor them, while this audit describes the full synced store." Press Focus filters on the Intelligence Studio screen and either clear them or apply the same narrowing to what you are comparing against.

Why: The range applies to the Executive Read, Data Audit, Analysis and Suggested Plan tabs. Field Intelligence and Report Templates always read the whole store, because a field-by-field profile describes the field rather than a period.

Fix: Compare ranged figures with ranged figures only. The scope line above the tabs prints the period in words, so you can see at a glance which one you are reading.

Why: Pipedrive Leads live in the same store as deals. Any analysis about deals excludes them and reports how many it set aside. Analyses about the top of the funnel keep them, because that is where a lead belongs.

Fix: Read the caveat naming the number set aside. A lead is counted in the funnel analyses and left out of every opportunity count, pipeline value and outcome.

Why: Above an analysis's cap MarqIQ reads an even sample — one record in every N, by a fixed hash of its id — while a count reads the whole store. The sample spans your whole history, and the same records are read on every visit, so the figures do not shift between two looks.

Fix: Take shares and rates from the sampled analysis and counts from the audit. The population line at the top of the analysis says which of the two you are reading.

Why: The audit is expensive at scale, so it is worked out once per sync and stored. Right after a sync, or after a MarqIQ update that changes how it is calculated, the audit computed before is shown while the new one is worked out behind you. On a very large workspace that takes several turns of the scheduled refresh.

Fix: Read the grey line under the tabs that says when the figures were computed: it gives the date, and no line means the figures are current. Press Recompute in the page header to work the audit out now. The header also prints when the tool was last synced.

Why: The assistant answers over all time unless it is given a period. The data audit it reads takes a period only when you ask for one, and it says which period it used in its answer.

Fix: Ask for the period explicitly — "the deals created between 1 July and 30 September". Where a screen sits behind a plan gate, the assistant applies the same gate: the executive read is refused to it on a plan that does not include it, so an assistant is never a way round one.