Hyros, GA4 or the ad platform: which attribution number do you actually run the business on?
Ask three tools how much revenue paid search produced last month and you will get three numbers. Not slightly different. Different enough to change a decision. Google Ads says one thing, GA4GA4 Google Analytics 4, the current version of Google's free web analytics. Reports visits, traffic sources and conversions using its own attribution rules. says less, and a third-party tracker like HyrosHyros A third-party ad-tracking tool that follows a customer across clicks, opt-ins and purchases with its own tracking script, so you get one attribution view that isn't run by an ad platform. says something else again, often more.
None of them are lying. They are answering different questions. The mistake is treating them as competing estimates of one truth instead of three instruments with different blind spots.
What each one can and cannot see
The ad platform (Google Ads, Meta Ads ManagerAds Manager Meta's interface for building campaigns and reading their results. Its numbers reflect Meta's attribution rules, not your bank account.) sees its own clicks and impressionsImpressions The number of times an ad was shown. Ten impressions can be ten people once or one person ten times., and whatever conversionsConversion The action you want someone to take after seeing an ad: a purchase, a form fill, a booked call. Each platform counts conversions by its own rules, which is why two dashboards rarely agree. you send it. It cannot see the other platform’s ads.
It grades its own homework, and its defaults include view-throughView-through A conversion credited to an ad the person saw but never clicked. Ad platforms count these; your analytics and CRM cannot see them. and multi-day click windows that your accountant would not accept. Its number is best read as “conversions this platform touched”, not “conversions this platform caused”.
GA4 sees sessionsSession One visit to a website, from arrival to leaving or going idle. Analytics tools count sessions and assign each one a source. on your site and the source that started each one. Its default model is data-driven attributionAttribution The rules for deciding which ad, click or channel gets credit for a sale or lead. Every tool uses different rules, so the same sale can be credited to Google in one report and Meta in another., which “distributes credit for the key event based on data for each key event”, learning from converting and non-converting paths (Google Analytics Help, get started with attribution).
It is the only one of the three that sees organic, email and direct alongside paid. Its blind spots are anything that happens after the site: a sale closed on a call three weeks later, a refund, a rebill. It also loses people across devices and after cookieCookie A small file a website stores in your browser to recognize you on later visits. Ad tracking depends on them, and browsers and privacy laws increasingly limit them. expiry.
Hyros or a similar tracker ties a person to a click through an email or phone number captured on a form, then follows that person through your CRMCRM Customer relationship management software, such as HubSpot or Salesforce, where leads, contacts and deals are recorded. Usually the only system that knows whether a lead was real. and payment processor. Its strength is exactly where GA4 is weak: revenue that lands late, offline or on a subscription.
Its blind spot is the front of the funnelFunnel The path from first seeing an ad to buying, usually drawn as stages that narrow: impression, click, lead, qualified lead, sale. "Higher in the funnel" means earlier and cheaper; "lower" means closer to money.. If someone never gives you an email, the tracker never sees them, so it under-reports top-of-funnel campaigns and over-credits the last click before the form.
Why the numbers diverge
A high-ticket funnel makes the gap obvious. Someone clicks a YouTube ad in January, reads two blog posts from organic search in February, books a call from a branded search ad in March and buys in April.
- Google Ads credits YouTube (view or click, depending on window) and branded search. Both campaigns report the sale.
- GA4’s data-driven model splits credit across YouTube, organic and branded search, and books it in March when the key event fired.
- Hyros credits the click it can tie to the email, usually the March branded search, and books the full sale in April when the money arrived, including any rebills later.
Add up the three and you have three or four sales for the price of one. Compare them and you have a map of where each tool is blind.
The rule I use
Give budget authority to the tool that reconciles to money. For most of my clients that is the tracker, because its revenue figure can be checked against Stripe or the CRM to the dollar. If it cannot be reconciled, it does not get authority, whatever it is called.
Use GA4 for the questions the tracker cannot answer: which content people read before they converted, which landing pagesLanding page The page someone arrives on after clicking an ad. Built for one action, unlike a homepage that tries to do everything. leak, how much organic and email contribute.
Use the ad platform number for what it is good at: relative performance between campaigns on the same platform, and feeding the bidding algorithm. Never use it to compare Google against Meta, and never use it alone to decide total spend.
Then do one thing every month that almost nobody does: put all three numbers in a spreadsheet next to actual revenue and write down the ratios.
Over a few months the ratios stabilize, and you learn that Ads Manager runs at about 1.6x reality and GA4 at about 0.7x on your account. That is the point at which the dashboards become useful, because you know how to read them.
The failure mode
The failure mode is the opposite: picking whichever tool shows the best number this month and quoting it. If the tracker looks bad, cite Ads Manager. If Ads Manager looks bad, cite the tracker.
It feels like flexibility. It is how ad budgets get spent on campaigns that do not work, with a different chart every time to prove that they do.