Why your conversion tracking is probably wrong

The most expensive mistake in paid media isn't a bad campaign structure or weak creative. It's optimising confidently against numbers that were never right.

When tracking is broken, the platform doesn't tell you. It just gets very good at finding more of whatever it thinks a conversion is.

Here's what I check first on any account.

1. The same conversion counted twice

Symptom: platform conversions are roughly double what your back end shows.

Common causes: a conversion event firing both through Google Tag Manager and a hardcoded tag; a thank-you page that fires on reload; an event firing on both "add to cart" and "purchase" while both are set as primary conversions.

Fix: list every configured conversion action and confirm exactly which are marked primary. Most accounts have several counting toward optimisation when only one should.

2. Counting every conversion instead of one

In Google Ads, a lead form set to count "every" conversion means one person submitting three times counts as three. For lead generation, that's wrong. Set it to "one".

For e-commerce purchases, "every" is correct. Getting these backwards is common and quietly distorts everything.

3. Attribution windows nobody chose

Default windows differ across platforms. If Meta is on a 7-day click and Google is on a 30-day click, you're comparing two different definitions of a conversion and concluding one channel is better.

Pick windows deliberately based on your actual sales cycle, and know they're different when you compare.

4. Conversions that aren't conversions

I regularly find accounts optimising toward page views, time on site, or "contact page visit". The platform will faithfully deliver people who visit the contact page and never fill anything in.

Optimise toward the thing you actually want. If the volume is too low for the algorithm to learn, use a genuine intermediate step — a real form start, not a page view — and reconcile it against real outcomes.

5. Consent mode and blocked tags

A meaningful share of traffic blocks tracking. If your consent setup isn't configured properly, conversions vanish and performance looks worse than reality — which leads people to cut campaigns that were actually working.

6. No reconciliation against reality

This is the big one. Once a month, put platform-reported conversions next to what your CRM, invoicing system or Shopify back end says.

They will not match. That's expected. What matters is knowing the size and direction of the gap so you can factor it in.

A practical order of operations

1. Fix duplicate firing
2. Set count settings correctly per conversion type
3. Choose attribution windows deliberately
4. Confirm you're optimising toward real outcomes
5. Verify consent handling
6. Reconcile monthly against your own numbers

Do these in order. There's no point tuning bids on numbers that are double-counted.

Why this matters more with automated bidding

Smart Bidding, Advantage+ and TikTok's automated bidding all learn from the conversion signal you feed them. Feed a broken signal and they optimise hard toward the wrong thing — and they do it faster and more efficiently than a human ever could.

Automation makes accurate tracking more important, not less. The better the algorithm, the more expensive your measurement errors become.

If you're setting tracking up from scratch, start with setting up GA4 conversion tracking properly.