Affirm → BigQuery · Conversion tracking
Connecting Affirm to BigQuery
Affirm is longer-term financed BNPL, typically on higher-ticket items, which adds a credit decision between the checkout and the sale. BigQuery has no attribution model, no deduplication rules and no opinion about your data — which makes it the only destination on this site where every number is exactly as good as the query that produced it. The join between them succeeds or fails on one thing: whether whatever join key you decide on reaches the Affirm charge before the conversion happens, because BigQuery can only match a conversion it can tie back to a click it recognizes.
✓ 30-minute call · No commitment · You leave with a written remediation plan even if we’re not the right fit.
Written guarantee
95%+ conversion accuracy on Affirm → BigQuery
Measured against your own order records at handover. Miss it and we keep working at no cost until it’s met.
Built on — and certified for — the platforms that matter
Stape
Pro Partner
Meta
Tech Provider
Tag Manager
TikTok
Events API
Microsoft
UET Partner
CAPI Partner
CAPI Partner
Built on the same partner stack used by enterprise DTC brands.
By the numbers
No fake testimonials.Just the math.
70+
Platforms supported
Shopify · Woo · Magento · BigCommerce · 65 more
8+ / 10
Meta EMQ guaranteed
in writing — keep working free until hit
2–4d
Build to live
accurate data flowing within 7 days
$0
Monthly fees, ever
one-time fixed price · you own the stack
400d
Cookie lifetime
max RFC 6265bis allows · 57× Safari ITP
9
Ad channels covered
Meta · Google · TikTok · MS · LI · Pin · Snap · Reddit · X
95%+
Tracking accuracy
verified vs Stripe · GA4 · ad-platform reports
30 min
Free founder call
no SDR · no pitch deck · written remediation plan
Every figure above is either a commitment we put in writing or a countable fact. What your own setup is losing is not on this list, because we haven’t measured it yet — that happens on the call, against your own order records.
Real clients · real videos · real outcomes
Don’t take our word for it. Watch the receipts.
Founders, marketers, and agency owners describing the before-and-after of their Affirm → BigQuery tracking rebuild — in their own words, on camera.
Why this join usually fails
Whatever join key you decide on never reaches Affirm
The single most common cause. The shopper leaves for Affirm's flow, is assessed, and returns only if approved. Both the redirect and the decline create paths where nothing fires on your site.
A credit decision inside the checkout
Declines are real volume that reached checkout and produced nothing. Counting checkout starts rather than captured charges overstates performance by the decline rate.
The warehouse quietly disagreeing with every platform
This is not a bug — it is the point. Your warehouse counts orders; Google counts click-attributed conversions in its window; Meta counts differently again. When the four numbers differ, the warehouse is usually right about revenue and wrong about attribution, and knowing which question you are answering is the whole skill.
Nothing reconciles afterwards
An upload that succeeds is not an upload that matched. BigQuery will not tell you anything is wrong. Verify by querying the table directly against your source system's own totals, and check for duplicate rows on the join key before trusting any aggregate.
Architecture
How Affirm and BigQuery actually connect
Two systems that know nothing about each other. Everything below exists to give them one shared reference.
Step 1
Affirm holds the truth
Affirm runs a real credit decision, so a shopper who reaches checkout may be declined. The order only exists if the financing was approved and captured.
Step 2
BigQuery needs a click to match
BigQuery needs a schema and consistent types. The failure mode is not rejection, it is silent divergence: a column that arrives as a string in one load and an integer in the next, or a timestamp in the wrong zone, which produces answers that look plausible and are not.
Step 3
The identifier is the bridge
Whatever join key you decide on, captured at the visit and carried into the Affirm charge. Without it there is nothing to join on and no tool fixes that.
Step 4
The event source decides reliability
Fire on capture after approval, and tag the payment method — Affirm orders skew to higher ticket values that will distort a blended average.
Step 5
Deduplication at the destination
Nothing deduplicates for you. Streaming inserts are at-least-once, so a retried load can create duplicate rows, and the only protection is a deduplication step you write and actually run.
Diagnose it honestly
Working vs. broken, side by side
Both states look identical from the BigQuery dashboard, which is why this runs for months before anyone notices.
Normal — not a fault
Whatever join key you decide on present on the order
Visible on the record itself in Affirm, not merely captured somewhere on the site.
Sent from your store's order webhook, on Affirm charge capture
Server-side, so it does not depend on the customer's browser still being there.
Deduplicated correctly
Nothing deduplicates for you. Streaming inserts are at-least-once, so a retried load can create duplicate rows, and the only protection is a deduplication step you write and actually run.
Reconciles with the source
BigQuery's count and Affirm's records agree within a few percent over a closed window.
Actually broken
Whatever join key you decide on missing on most records
The shopper leaves for Affirm's flow, is assessed, and returns only if approved. Both the redirect and the decline create paths where nothing fires on your site.
Sent from the browser
Anyone who closes the tab, blocks the script or converts without a page load is invisible. Declines are real volume that reached checkout and produced nothing. Counting checkout starts rather than captured charges overstates performance by the decline rate.
Counted twice, or not at all
This is not a bug — it is the point. Your warehouse counts orders; Google counts click-attributed conversions in its window; Meta counts differently again. When the four numbers differ, the warehouse is usually right about revenue and wrong about attribution, and knowing which question you are answering is the whole skill.
Nobody has reconciled it
The dashboard shows a number and no one has compared it to Affirm. A wrong number that nobody checks is indistinguishable from a right one.
Diagnostic sequence
Diagnostic order
Each step rules out a class of cause. Run them in order and the last one usually has a single explanation standing.
- 1
Count orders carrying whatever join key you decide on
The ceiling on everything else. Do this before debugging anything downstream.
- 2
Follow one real conversion end to end
From the ad click through Affirm and into BigQuery.
- 3
Compare Affirm checkouts started, approved and captured
The decline rate is a genuine cost of this payment method and it belongs in the report.
- 4
Reconcile warehouse order counts against each platform for one closed window
The gaps localize the problem: a gap against all platforms is a collection issue, a gap against one is that platform's window or matching.
- 5
Read the match rate, not the success message
BigQuery will not tell you anything is wrong. Verify by querying the table directly against your source system's own totals, and check for duplicate rows on the join key before trusting any aggregate.
Source-of-truth validation
How it gets proven
Against Affirm's own records, not against either platform's dashboard.
- Whatever join key you decide on confirmed present on a live order.
- A real conversion followed from click through Affirm to BigQuery.
- Warehouse order counts and revenue reconciled against the source system for a fully closed window, with duplicates checked on the join key.
- An approved Affirm order confirmed to report and a declined one confirmed not to.
- Reconciled over a fully closed window, matched per record rather than on totals.
- Re-checked 30 days later.
Straight talk
When you don't need this
Worth reading before paying anyone, including us.
- If BigQuery already reports conversions that reconcile against Affirm's records within a few percent, this join is working and there is nothing to buy.
- Measure identifier coverage yourself first — pull your last hundred orders and count how many carry whatever join key you decide on. Ten minutes, free, and it tells you whether any of this is worth doing.
- If your average order value is low, Affirm is a poor fit for the payment method and the integration follows that.
- If you have one store and one ad platform, a warehouse is infrastructure you will maintain for a reconciliation you could do in a spreadsheet.
Who this is for
Honest about who we’re a fit for.
We don’t take every project. Here’s how to tell if we’re right for you.
You’re a great fit if…
- You run Affirm → BigQuery (or are migrating to it)
- You spend $5K+/month on paid ads (Meta, Google, TikTok, Microsoft, LinkedIn)
- Your reported ROAS doesn’t match what your bank account is saying
- You want to own your tracking stack — no monthly SaaS fees forever
- You have 2–4 days to do this once and never think about it again
Skip us if…
- You spend less than $500/month on paid ads (browser pixel is fine)
- You want a SaaS dashboard with monthly fees forever (Triple Whale, Hyros)
- You need attribution modeling — we fix the data foundation, not multi-touch reports
- You want a 6-month enterprise consulting engagement (we ship in 4 days)
- You expect us to manage your ad accounts (we don’t — that’s a different service)
If you’re still unsure,book the call anyway— we’ll tell you straight whether we can help.
What working looks like
Whatever join key you decide on on every order
Server-sideCaptured before the conversion and stored where it survives.
Conversions arriving in BigQuery
8+ EMQThe row lands in a table you own, joinable to every other table you own, with no window and no expiry.
Reconciled against the source
DedupedAffirm's own records are the reference, not either dashboard.
What gets built
Identifier capture into Affirm
Capture the identifier before checkout and store it on the order, so approved financing is attributable regardless of the return journey.
Event source: your store's order webhook, on Affirm charge capture
Fire on capture after approval, and tag the payment method — Affirm orders skew to higher ticket values that will distort a blended average.
Dispatch to BigQuery
Streaming inserts or scheduled batch loads from your order system, with the raw event retained alongside whatever you derive from it.
Reversals where they exist
Refunds and cancelled financing both exist and should reduce the reported figure.
Pick the package that fits your ad spend
Fixed-fee, no surprises. If your ad spend is meaningful,server-side tracking is what we recommend— it’s the only setup that survives iOS, Safari ITP and ad blockers.
Browser-side starter
Wires the join and verifies it on a live conversion. Honest about its ceiling where the source can only be read from the browser.
- The join built and tested end to end
- Verified on a real conversion
- Reconciled against the source's own records
- Written handover
1st-party server-side
Save $18K–$120K over 5 years vs Triple Whale / Hyros
The join built server-side from the source's own records, so it does not depend on a browser being present when the conversion happens.
- Events sourced from the system of record, not the browser
- Identifier captured and persisted end to end
- Deduplicated at the destination
- Reversals wired where the source supports them
- 95%+ accuracy guaranteed in writing
Full 1st-party stack
Every destination fed from one place, plus monitoring so a broken join surfaces in days rather than at quarter end.
- Everything in 1st-party server-side
- All paid destinations from one container
- CRM and offline loop
- Refunds and cancellations reversed everywhere
- Monitoring per destination
Not sure which tier fits?Book a 30-min scoping call— we’ll tell you straight, no upsell pressure.
How it works
Built in 2–4 days. Accurate data within 7.
We deliver the entire 1st-party server-side stack in 2–4 business days. You start collecting accurate, deduped conversion data the moment it goes live — and within 7 days every signal is flowing cleanly to Meta, Google, TikTok and the rest.
Kickoff + audit
30-min call. We map your current pixels, GTM, dataLayer, consent setup, and ad-platform integrations. You get a written report of every leak.
Build the stack
Server-side GTM container deployed on your subdomain. Meta CAPI · Google EC · TikTok Events API · Microsoft UET wired. Identity match + dedup configured.
QA + go-live
Live-fire test purchases across every funnel step. Meta EMQ verified at 8+. Tag Assistant + Pixel Helper passes. Side-by-side report delivered.
Verify accuracy
Your data flows clean. Within 7 days every signal — Meta, Google, TikTok, Microsoft — is fully accurate. Notion runbook + Loom walkthrough handoff.
Got questions?
The honestanswers.
No buzzwords, no upsell. If we don’t know, we say so.
01Why aren't my Affirm conversions showing in BigQuery?
Almost always because whatever join key you decide on never reached the Affirm charge. The shopper leaves for Affirm's flow, is assessed, and returns only if approved. Both the redirect and the decline create paths where nothing fires on your site. Measure coverage before debugging the dispatch.
02What identifier does BigQuery need?
There is no vendor identifier here. You choose the key — order ID, user ID, click identifier, hashed email — and you live with the consequences. That is the whole appeal and the whole risk: nothing is standardised for you and nothing is silently wrong in a vendor's favour either.
03Where should the event come from?
Your store's order webhook, on Affirm charge capture. Fire on capture after approval, and tag the payment method — Affirm orders skew to higher ticket values that will distort a blended average.
04How do I know it worked?
BigQuery will not tell you anything is wrong. Verify by querying the table directly against your source system's own totals, and check for duplicate rows on the join key before trusting any aggregate.
Still have questions?Book a 30-min call— bring them all.
The call
What happens on your free 30-min call
No pitch deck, no SDR, no upsell pressure. Founder-led. Bring your ad-platform screens and yourAffirm → BigQuery setup — we’ll work through it together.
Min 1–5
We get the lay of your store
What you sell, what you spend on ads, who handles dev. Quick context-gathering — no questionnaire upfront.
Min 5–15
We audit your tracking live
Screen-share. We open your Meta Events Manager, GTM, and ad reports. You see exactly which conversions are leaking.
Min 15–25
We map a fix
What needs rebuilding, what stays. The exact Affirm → BigQuery setup we'd ship and what it would recover for you.
Min 25–30
Decide. Or don’t.
If we're a fit, we send a 1-page proposal in writing. If not, you walk away with a free remediation plan you can hand to anyone.
Get Affirm and BigQuery joined properly
Book a free 30-minute call. We measure your identifier coverage live on your own account, and you leave with a written plan — including the option that you don't need us.
Free · 30-min consultation · No commitment · Replies in under 2 hours · Available worldwide via WhatsApp
95% tracking accuracy guarantee
Free 30-min consultation · from $400 · 1st-party server-side
95% guarantee · from $400
