Before Changing Bids: How to Pull a Google Ads JSON File Your CRM Can Match
For paid teams doing a closed-month review: what IDs and splits to export, how to fix click-time vs conversion-time and time zone, when to use CSV, Sheets or the API for JSON, and the narrow CRM check to run before bid changes.
Brand promotion: this article introduces Guangsuan's Google Ads management service for paid teams. It is not an independent review.
Direct answer for a closed-period review: write down the row identity, the calendar rule and the conversion rule before you download. This guide fits a lead-generation account where sales qualifies enquiries and revenue closes later. It does not fit a reach-only campaign where a click is not expected to create an enquiry.
Lock the calendar and conversion rule first
An export is only as good as the range and time zone recorded with it. Google Ads stamps rows on the account clock, which can differ from the CRM clock. A form saved late on the last evening can sit in October on one side and November on the other, with neither side mistaken. Record the range plus the account zone on the file name or first tab.
Then fix how conversions are placed in time. One rule files the event on the date the conversion happened. The other files it on the date of the click that led to it. A 28 October click that becomes a 3 November enquiry therefore lands in different months under the two rules. Pick one rule for both sides of the comparison and keep the attribution window beside it.
Open conversion settings and note which actions are primary and which are secondary. Only primary actions feed the Conversions total used for bidding. Page views, button taps and other micro steps stay outside that total and need their own labels. A settings change also rewrites history, so keep counting choice and attribution model with the file. The download shows only what Google could record. Sales-confirmed stages remain in your own system.
Give each row an identity the CRM can meet
Names are edited, numbers persist. Carry the customer number through the file, show the numeric campaign number beside the campaign label, and keep each split that defines uniqueness, including date, device, network slice, place slice and conversion action where used. A campaign total and a campaign-by-day-by-device table need different keys. Mixing grains makes valid rows overwrite each other.
Pull only the grain the opening question needs. On the Campaigns, Ad groups, Keywords, Search terms or Ads view, set dates and columns first, then use the download control. When day, device, place or conversion splits must appear together, build the view in Reports instead. For a repeated or large pull, query with GAQL through the API rather than clicking.
Name the file with account, report, range, timing rule and pull moment. Leave the raw download untouched and work in a copy, so a retry replaces the same slice instead of adding a duplicate.
Choose CSV, Sheets or the API path for the task at hand
The Google Ads interface does not offer a JSON download.
The download menu lists Excel CSV, TSV, PDF, XLSX, XML, and Google Sheets.
XML is the closest native option, but it is not JSON.
For background on fields and the download menu, see Guangsuan guide to exporting Google Ads data to JSON.
Match format to the next action. Pick CSV for scripts and storage, XLSX for human review, and Sheets when several people must read the same live sheet. When opening CSV, use the text import wizard, force UTF-8, confirm the separator, turn spend and clicks from text into numbers, and test one date row before trusting month-day order. Build your own rates in the sheet because a shipped average can hide the slice that needs work.
Getting that data as JSON means calling the REST interface with a GAQL query, which returns only the fields named in the query.
The other owned route is to stage data in a warehouse and emit JSON there for a monthly match. That adds a project to look after, so it fits a routine better than a single file. Converting a CSV or XML download to JSON on your own machine is allowed for a one-off, but keep that script versioned with an owner and test it on a small hand-checked slice.
Cost in an API row arrives in micros in account currency. Divide once by one million in a noted place and never add raw totals across currencies.
import json
rows = [{"campaign.id": "881204", "campaign.name": "Search-Partner-Check", "metrics.clicks": 264, "metrics.conversions": 9, "metrics.cost_micros": 19850000}, {"campaign.id": "881207", "campaign.name": "Lifecycle-Assist-Test", "metrics.clicks": 517, "metrics.conversions": 6, "metrics.cost_micros": 34200000}]
print(json.dumps(rows, indent=2))The snippet above only shows field shape for parsing. Numbers are invented teaching data, not account data.
Run a one-action cross-check before bid moves
Limit the first pass to one primary action, one campaign group and one closed four-week window plus the first three days of the next month to catch late qualifications. Align timing rule, range and zone on both extracts. Keep ad-sourced rows apart from organic and referral rows. Check counting behaviour and repeated uploads before calling a remainder unexplained. Leave any leftover gap marked as open.
As a hypothetical teaching illustration with invented figures, not a client outcome: a file logs 58 ad-recorded form fills. Sales accepts 19 as qualified under a written definition, 14 dated inside the window and 5 dated just after it, with 4 reaching order. Each total answers a separate stage. When only a small share ever qualifies, inspect query intent and page fit before moving bids.
Guangsuan's Google Ads management covers search terms, ads, landing pages and conversion tracking for real enquiry demand. Scope, fees and handoff sit in a contract, with no fixed enquiry promise. Details are on the Guangsuan Google Ads management service.
FAQ
Which date should month-end use? The date that answers the opening question, applied on both sides with window and zone noted. Do not blend the two timing rules.
Does a file that opens cleanly match cleanly? No. Readable layout says nothing about grain, units, timing rule or which actions were summed.
Should a missing term be read as zero demand? Not yet. Quiet rows can drop from a filtered view, so confirm filters and grain first.
See every redirect as a revenue event.
Book a live walkthrough of the MoeLink attribution dashboard. We'll route a real link through your stack in the call.