What paid ads reporting platforms do: four different jobs
All-in-one paid ads reporting platforms do one of four jobs: move your ad data, show it, do both, or re-decide which ad earned the sale. Then settle each platform's attribution window and how much history you keep, because those two choices move your totals more than any logo.
| Job | What it does | Examples | How it is priced |
|---|---|---|---|
| Move data (connector, ETL) | Copies your ad data into Sheets or a warehouse | Fivetran, Airbyte, Supermetrics, Funnel | Rows (Fivetran); credits or capacity (Airbyte); fixed (Supermetrics); connector tiers (Funnel) |
| Show data (dashboard, BI) | Charts the data you already connected | Looker Studio, Power BI, DashThis | Google-built connectors free; DashThis by dashboards and sources |
| Move and show (all-in-one) | Connects your sources and reports in one tool | Improvado, Whatagraph, Databox | Usage (Improvado); per account (Whatagraph); flat or account packs (Databox) |
| Re-decide credit (attribution) | Reassigns credit from your site and order data | Northbeam, GA4 attribution reports | Northbeam: data volume, measured in pageviews |
If your dashboard disappoints you, check which job you bought: your dashboard cannot show history your connector never stored. Our analytics and insights team starts every reporting brief with that map.
Why your platform totals never match
Your totals differ because each platform counts your sale by its own documented rules: its own credit model, its own window, and its own date. Google expects its own two products to disagree.
"Discrepancies, often up to 20%, are expected due to different attribution models"
Google Ads Help, "Understand your conversion tracking data", comparing Google Ads with Google Analytics. Read on 19 November 2026.
| Platform | Documented setting | What it does to your totals |
|---|---|---|
| Google Ads | Data-driven model is the default for most conversion actions | Spreads credit using your past data for that action |
| Google Ads | 30-day click default (Search, Display); view-through 1 day; by click time | Your later sales drop out; your months shift |
| LinkedIn Campaign Manager | Default 90-day click, 90-day view | Claims your sales long after an ad view |
| Meta Insights API | Mirrors the Ads Manager setting since 10 June 2025 | Your older connector pulls may not match newer ones |
| Meta Insights API | 7-day and 28-day view not returned from 12 January 2026 | Your long view-through history stops |
| TikTok Ads Manager | Click 1 or 7 days; view off or 1 day | Your slow buyers fall outside the window |
| Microsoft Advertising | 1 minute to 90 days; view-through up to 30 days | Set per goal, so check each of yours |
| Apple, iOS apps | ATT consent; AdAttributionKit value thresholds | Your app conversions arrive partial or without values |
Nothing there is broken. Read each column as that platform's claim, and let your order system or an incrementality test settle the real count. In Google Ads, you can also report conversions by when they occurred.
Which of your totals will differ, and why?
Tick your platforms, pick your referee and enter your usual days from first ad click to purchase. You get the documented reasons your totals differ and the one setting to align first.
3 reasons your totals will differ:
- Each platform credits its own ads, so one sale can sit in every column and the columns add up to more than your sales.
- Google Ads reports by click time with a data-driven model for most actions, so a late-month click counts in that month even if the order lands in the next.
- Meta's API has mirrored each ad set's attribution setting since 10 June 2025 and dropped 7-day and 28-day view on 12 January 2026, so older stored rows may not match.
Align first: in Google Ads, add the conversion-time columns, so your months line up with the order system.
Re-run it after any conversion edit. For GA4, our GA4 board-grade reporting cadence guide covers the monthly review.
How much ad history the platforms keep for you
Since 1 June 2026, the Google Ads API keeps your daily and weekly data for 37 months, and Meta keeps your totals for up to 37 months. Your monthly Google Ads data stays for 11 years. For daily year-on-year comparisons past that, you need your own stored copy.
Your retention limits as a table
| Data | Kept for | Since |
|---|---|---|
| Google Ads API, daily, hourly and weekly | 37 months | 1 June 2026 |
| Google Ads, monthly, quarterly and yearly | 11 years | 1 June 2026 |
| Meta Insights API, total values | Up to 37 months | Meta blog, October 2025 |
| Meta, unique-count breakdowns | 13 months | Meta blog, October 2025 |
| Meta, frequency breakdowns | 6 months | Meta blog, October 2025 |
| GA4 explorations, standard property | 2 or 14 months | Your setting |
| GA4 explorations, Large and XL property | 2 months | Fixed |
The policy also covers Google's BigQuery Data Transfer Service, so a warehouse started next year cannot backfill aged-out daily rows. Our guide to moving LinkedIn Ads data into PostgreSQL shows one way to keep your own copy.
Choosing a vendor when your customer data is regulated
If you are a bank, insurer or fintech, ask where your data is processed before you compare features. Of the 11 vendor sites we read on 19 November 2026, only Fivetran lists Asia-Pacific hosting, including Singapore and Sydney. Ask the rest.
| # | You ask the vendor | Why it matters |
|---|---|---|
| 1 | Which attribution setting does each connector pull for me? | Your Meta API numbers mirror Ads Manager since June 2025 |
| 2 | How far back can you backfill, and do you keep it? | Your Google Ads granular API data stops at 37 months |
| 3 | Where is data processed and stored? Can I choose? | Only Fivetran's pages list Asia-Pacific regions |
| 4 | Can I read the SOC 2 or ISO 27001 report? | You need the audit, not the website claim |
| 5 | Which sub-processors touch the data, and where? | Supermetrics says some may process outside the EU/EEA |
| 6 | What unit does your connector count use? | You cannot compare connectors, sources and platforms |
| 7 | What drives price: sources, accounts, rows or pageviews? | Your bill grows with that number |
| 8 | Can I export raw rows when I leave? | Your history should outlive your contract |
| Vendor | Count, in its own unit | Security claim stated | Region stated |
|---|---|---|---|
| Supermetrics | No total on home page | SOC 2 Type II audits; ISO 27001 | EU processing (AWS, GCP) |
| Funnel | 600+ connectors (home page) | SOC 2 Type 2; ISO 27001 | EU, US or "globally"; US or EU on select plans |
| Adverity | 600+ platforms | SOC 2 Type 2; ISO/IEC 27001 | Not stated |
| Improvado | 1,400+ data sources | SOC 2 Type II report | Not stated |
| Airbyte | 700+ sources | Not stated on pages read | Your VPC or Airbyte Cloud |
| Coupler.io | 400+ sources | SOC 2; ISO 27001 | Not stated |
| Windsor.ai | 350+ sources | SOC 2 Type II audit | Not stated |
| Databox | 130+ tools (107 one-click) | SOC 2 | Not stated |
| Whatagraph | 60+ native sources | Not stated | Not stated |
| DashThis | 30+ platforms | Not stated | Not stated |
| Fivetran | No count on pages read | Lists its cloud providers' certifications | Singapore, Sydney, Tokyo, Mumbai among others; cloud and region by plan |
| Market | What the guidance says | Source |
|---|---|---|
| Singapore | Transfer only with protection "comparable to the protection under the PDPA" | PDPC Advisory Guidelines, para 19.3 (May 2022) |
| Malaysia | Law "substantially similar" to the Act, or adequate protection | PDP Guideline No. 3/2025, para 4.1 |
| Canada (PIPEDA) | Transfers allowed; "The transferring organization is accountable" | OPC guidelines, January 2009 |
For Australia, read the OAIC's APP 8 guidance with your privacy officer; we could not open it on 19 November 2026. Our 20 privacy checks for marketing platforms cover the contract side.
Which kind of reporting tool do you need?
Answer seven quick questions. The picker names one kind of tool and the first question to ask, never a vendor: the job decides your shortlist.
You need: a connector (ETL) into your own warehouse or Sheets, with a dashboard on top, from a vendor that states a hosting region you can accept.
Why: you want 48 months of history and Google Ads granular API data stops at 37, so start storing your own copy now.
Ask first: which attribution setting does each connector pull, and how far back can it backfill? Also ask where data is stored: few vendor sites state Asia-Pacific hosting; Fivetran lists Singapore and Sydney, so ask the rest.
Leapbuzz is building its own Marketing Operating System (MOS), a layer that reads spend and outcomes across every ad platform. It is in build: we use it to run engagements and ship it to anchor clients first, as part of the engagement where it is ready. Read this page knowing that; our performance marketing page describes MOS.
Four things to settle before you pick a logo
- Write down every conversion action with its model and window, platform by platform, in one sheet.
- Name your referee: your order system or CRM, never a platform column.
- Start storing daily rows you own this month, before Google's 37-month limit on granular data eats them.
- Send the eight vendor questions above and keep the written answers.
Put your reconciled number on the board report; the board KPI scorecard shows the format. For a second read of your setup, talk to Leapbuzz.
