Snap changed the count, not the campaign

What Snap actually changed, and what it did not#

The fox does not argue with the wind. It notes which way the wind is walking, then moves accordingly. Something similar is being asked of app teams this week, because Snap has opened Unified Attribution to every app advertiser globally, and the sensible response is neither celebration nor suspicion. It is a careful read of what the change touches, and mobile attribution is the whole of what it touches.

Unified Attribution is a campaign-level setting inside the App Promotion objective. Turn it on and Snap stops counting conversions its own way and starts counting them your measurement partner's way, adopting that partner's events, attribution windows and settings. Snap describes the product as one that accounts for multiple touch points across the buyer journey, aligning with mobile measurement partners so advertisers can judge campaigns in real time.

The rollout is the news. Marketing Dive and Social Media Today both report the setting moving from a limited beta, which began in May, to every advertiser worldwide using AppsFlyer or Adjust, with Singular, Kochava and Branch still in progress, which makes mobile attribution the one lever that moved. Social Samosa frames it as a fix for fragmentation between platform metrics and cross-channel ones.

So the honest summary is narrow and worth saying plainly. Snap has not changed its auction, its delivery, its creative ranking or its audience reach. It has changed the ruler. Mobile attribution on Snap now measures with your instrument instead of Snap's, and everything downstream of that ruler will move on the day you flip it. That is the scent worth following through the rest of this trail.

Mobile attribution is the ruler, not the wall#

Here is the distinction that decides whether this helps you or hurts you. If your reported cost per install falls the week you switch, two explanations fit. Either the campaign got better, or the counting got kinder. Only one of those pays rent, and the mobile attribution change guarantees the second while promising nothing about the first. A fox that mistakes a moved fence for a moved field goes hungry in a well-measured meadow.

Why a better count is not a better campaign#

Attribution in marketing has a stubborn habit of being mistaken for causation, and mobile attribution carries that habit further than most, and the best evidence against that habit comes from inside a platform rather than outside one. Gordon, Moakler and Zettelmeyer analysed 663 large-scale experiments at Facebook to ask whether observational methods can recover what a randomised experiment finds. The answer was not encouraging.

Their randomised trials put median lift at 29% for upper-funnel outcomes, 18% for mid-funnel and 5% for lower-funnel. The same data run through a strong observational method returned 83%, 58% and 24%. The published conclusion is blunt: despite large-scale experiments and rich user-level data, the authors state they are unable to reliably estimate an ad campaign's causal effect by these means.

Read that chart carefully before the next planning meeting. It does not say measurement is worthless, and it does not say your measurement partner is wrong. It says the gap between what a mobile attribution rule credits and what an advert actually caused is large, systematic and pointed in one direction. Switching whose counting rule you use moves you along that gap. It does not close it.

Which is why the sober framing for Unified Attribution is a reporting alignment, not a performance product. Unified mobile attribution makes Snap legible next to your other channels. Legibility is genuinely valuable. It is simply not the same thing as incremental revenue, and a marketing attribution software decision has never once, on its own, sold a single subscription. No mobile attribution setting has ever persuaded a person to download anything.

Where mobile attribution numbers disagree, by design#

That is a builder's complaint, posted while this rollout was landing, and it points at the part nobody puts in a launch post. Mobile attribution plumbing is fiddly, and the fiddliness is where the discrepancies breed in the undergrowth. A partner integration is not a switch. It is a set of agreements about what counts, for how long, and on whose clock.

Snap's integration is a useful example of why two honest systems produce two honest numbers. Adjust documents that Snapchat does not use its link URLs, so the partner instead sends them every app session our SDK reports, and Snapchat answers with a device identifier and a timestamp. That is a fundamentally different matching mechanism from a click-through link, and different mechanisms find different users.

Retention windows differ too. Adjust notes that it, and every other Snapchat measurement partner, does not store Snapchat attribution data for longer than 150 days. So a long consideration cycle can fall off one system's ledger while remaining on another's. None of that is a fault. It is the ordinary friction of two clocks in one room, and mature mobile attribution practice budgets for it.

The practical consequence is that you should expect Snap's own reporting and your appsflyer attribution numbers to sit apart even after you unify them, because unifying the mobile attribution setting does not unify the mechanism. Plan for a delta, document it, and stop treating each month's mobile attribution variance as a mystery to be solved in the thicket.

The setting you cannot unpick#

This is the mobile attribution detail that turns a reporting choice into a planning decision, and it is the one most likely to be missed. The attribution method is selected when the campaign is created, under the App Promotion objective. PPC Land reports that it cannot be switched off once a campaign exists, and that an advertiser wanting to change method must build a replacement campaign instead.

Think about what that does to a like-for-like comparison. Your existing campaigns keep Snap's counting. Any new campaign you build with Unified Attribution keeps your partner's counting. Put both mobile attribution populations in the same spreadsheet and the column headers will lie to you, quietly and consistently, for as long as the two run side by side.

Snap has published beta results, and they are worth reading precisely because of how they are framed. PPC Land's write-up lists a gaming advertiser at 26% lower cost per acquisition against SKAN-optimised campaigns, Mohegan Sun at 89.8% higher return on ad spend and 77% lower cost per install by week four, Deblock at 43% more installs, and Codeway at 18% lower cost per install at 1.8 times the scale.

Notice what the framing does not claim. Every one of those figures is a comparison of attributed outcomes under changed accounting, which is exactly how a careful reader should hold them. The trade write-ups and the martech coverage repeat the numbers faithfully; none of them can turn an attributed mobile attribution delta into a causal one, and neither can you. That is the hedgerow every performance claim eventually has to get through.

Five mobile attribution checks before you flip the switch#

None of this argues against switching. Aligning Snap to the same measurement spine as your other channels is straightforwardly sensible, and the quiet quarry here is comparability, not a quick win, and better mobile attribution genuinely delivers it. A patient prowl beats a quick pounce: it needs a clear head and a written mobile attribution record, so next quarter's questions have answers.

The fifth step is the one that gets dropped, and dropping it is how a mobile attribution rollout ends up being reported as a performance improvement. A holdout is unglamorous, slightly expensive and the only thing on this whole trail that measures the wall rather than the tape. It is how you outfox your own dashboard.

Teams that run regulated or awkward categories have a further reason to be tidy here. When a compliance question arrives about what an advert did, a documented mobile attribution change with a dated baseline is a defensible answer. An undocumented one is a shrug with a spreadsheet attached, and folkfox has watched that shrug cost real money.

Frequently asked questions#