Marketing Attribution: Why WhatsApp Sales Vanish From Your Reports

Why do first-touch, last-touch and multi-touch give different answers for the exact same sale?

Attribution isn't a measurement problem before it's a definitional one. Picture one real shopper: they see a Facebook ad on Monday, get a WhatsApp broadcast about a sale on Thursday, click a follow-up email link on Saturday and buy the same day. Every euro of that sale is real and undisputed. Which channel gets credit for it is not — it's a modeling choice, and different models hand the entire, or partial, credit to completely different steps in that same sequence.

First-touch attribution gives 100% of the credit to the first interaction — the Facebook ad, in this example — on the theory that nothing else happens without that first spark of awareness. Last-touch does the opposite: 100% to whatever the customer touched right before buying — the email link, here — on the theory that whatever closed the sale is what actually mattered. Linear attribution splits credit evenly across every touch; time-decay weights the more recent touches more heavily; position-based (or "U-shaped") attribution gives most of the credit to the first and last touches and splits a smaller remainder across whatever happened in between. None of these is objectively correct — they're accounting conventions for the same underlying event, chosen for how they'll be used, not because one uniquely reflects what actually persuaded the customer.

• ✓First-touch — 100% of the credit to the first interaction a contact ever had with the business.
• ✓Last-touch — 100% of the credit to whichever interaction happened immediately before the sale.
• ✓Linear — equal credit split across every touch in the sequence, regardless of order.
• ✓Time-decay — more credit the closer a touch sat to the actual purchase.
• ✓Position-based ("U-shaped") — most of the credit to the first and last touches, a smaller share to everything in between.
• ✓Multi-touch / data-driven — credit weighted by a model trained on which touch combinations actually preceded a conversion, rather than a fixed rule.

Why does a link survive a click, but a WhatsApp chat usually doesn't?

The tracking parameter almost every marketing tool relies on — utm_source, utm_medium, utm_campaign and the rest — traces back to a specific product: Urchin Software Corporation's web analytics tool, whose "Urchin Traffic Monitor" gave the format its name. Google acquired Urchin in April 2005, for a reported $30 million, and relaunched the product free that November as Google Analytics — at which point UTM parameters went from one vendor's tagging convention to the closest thing web marketing has to a universal standard, adopted today by essentially every analytics and marketing platform regardless of any connection to Google.

The reason it works at all is mechanical, and it's worth stating plainly because it's the whole story: a UTM-tagged link is a URL. A browser loads it, the destination server reads whatever parameters rode along after the question mark, and something on that server writes them down — against a landing page visit, a form fill, eventually an order. Email and SMS links fit this model perfectly, because the thing being clicked is always a webpage. A WhatsApp chat, opened either by tapping a wa.me link or by simply messaging a saved number, isn't a webpage at all — it's a conversation thread that opens inside an app. There's no browser in that path and no address bar, so there's nowhere for a query parameter to ride along to. The click can be every bit as real and as commercially decisive as an email click. It just doesn't leave the same kind of paper trail by default.

So how does Meta actually solve this for its own ads — and where does that fix stop working?

Meta built a specific, separate mechanism for exactly the gap above, but only for one entry point into WhatsApp: a Click-to-WhatsApp (CTWA) ad running on Facebook or Instagram. The moment someone taps such an ad, Meta generates a click identifier — ctwa_clid — and attaches it to the referral object carried on that customer's first incoming WhatsApp message. It arrives on the business side inside the ordinary WhatsApp webhook payload, alongside which ad, and which piece of ad creative, the person actually tapped.

Getting credit for the resulting sale isn't automatic from there, though. Per Meta's own WhatsApp Business Platform developer documentation, that same ctwa_clid has to be captured off the referral object and reported back to Meta — inside a user_data.ctwa_clid field, via the Conversions API, attached to whatever lead or purchase event the conversation eventually produces. Skip that round trip, and the conversation still happens and the sale still closes; it simply shows up in Meta's own ad reporting as if the click that started it never occurred, which is exactly the kind of gap that quietly makes a working ad campaign look like it isn't.

How big is that blind spot, really, in a market like Israel?

This isn't a theoretical edge case here the way it might be elsewhere. According to the Israel Internet Association's own Internet Index, WhatsApp is used by 99% of Israeli adults — ahead of every other platform the index tracks, including YouTube (98%), Facebook (90%), Instagram (82%) and TikTok (59% overall usage, 34% daily). It's not a close race: the single channel most Israeli businesses already lean on hardest for customer conversation is also, structurally, the one the rest of this piece has just walked through as the hardest to natively credit for a sale.

That combination is what makes the mechanical gap in the previous two sections worth taking seriously rather than filing under "edge case." A business running its comparison of "which channel actually converts" on WhatsApp, email and SMS side by side isn't comparing three channels with an equally intact paper trail — it's comparing two channels whose links can be tagged and read like any webpage, against one whose dominant form of contact, for almost every Israeli business, carries no click trail of any kind by default.

Does the "customers need dozens of touchpoints" idea even apply to a WhatsApp order?

"Buyers need 27 touchpoints before converting" circulates constantly, and unlike some of the numbers that make the marketing-blog rounds, this one does trace to a real, named, dated source: Forrester's own 2021 B2B Buying Study, more than 950 respondents worldwide, found the average number of vendor interactions had jumped to 27 — a mix Forrester itself has broken down as roughly 15 digital and 12 human — up from 17 shortly before the pandemic. The number is real. What it describes is a committee-driven enterprise software or services purchase, decided over months by multiple stakeholders who each need convincing separately. That's about as far as you can get from a WhatsApp order for a birthday cake, and the figure gets repeated in exactly that mismatched context constantly, with the "B2B software buyers" part quietly dropped.

Google's own research into ordinary consumer purchase behavior — a study run with The Behavioural Architects, published in July 2020 and built on a genuinely large experiment covering 310,000 real purchase scenarios across financial services, consumer goods, retail, travel and utilities, on top of literature reviews and shopping-observation studies — reaches a real but differently shaped conclusion. It found people loop, unpredictably and repeatedly, between an expansive "exploration" phase and a narrowing "evaluation" phase, cycling through that pair as many times as they personally need before deciding. What that research does not do, worth stating plainly, is publish a single headline "average number of touchpoints" the way the Forrester figure does for its own, narrower population — despite how often research in this space gets summarized as if a clean, universal touchpoint count exists behind it.

Do the people who study this for a living actually trust their own attribution numbers?

Not especially, and it's worth knowing that going in — it recalibrates how much confidence to put in any one channel comparison, including your own. An EMARKETER survey run in partnership with Snap, fielded among 282 senior-level US marketers in 2024, found that 78.4% of them still use last-click attribution alongside standard web analytics to judge how a channel is actually performing. Asked directly whether they're confident last-click reasonably reflects a channel's real, long-term impact on the business, only 21.5% said yes — and 74.5% said they're either already moving away from last-click or actively want to.

A related, if not identical, measurement problem shows up one level down the funnel. Branch's 2025 State of App Growth Report, surveying more than 700 director-to-C-level app-marketing leaders worldwide, found 82% reporting only partial clarity matching a mobile app install back to the channel that actually drove it, with just 18% calling themselves "very confident" in their attribution data — this despite most of those leaders having dedicated attribution vendors built specifically to solve that exact problem. It's a different layer of the funnel than WhatsApp-versus-email, but the shape is the same: even well-resourced measurement, aimed squarely at this problem, still runs into it.

So what should a business actually do about it?

None of the above is a reason to stop measuring — it's a reason to measure each channel according to what it can actually tell you, instead of comparing all of them as if they carried an equally intact trail by default.

1. 1Tag every link you control — email, SMS, landing pages, popups — with consistent UTM parameters. These are the channels where the mechanism above works by default; not using it is leaving a free, already-standard tracking layer unused.
2. 2If you're running Click-to-WhatsApp ads specifically, confirm with whoever manages them that ctwa_clid is actually being captured and reported back through the Conversions API. Skip that step and the campaign's real WhatsApp-driven sales will keep reporting as zero, regardless of how well the ads are actually working.
3. 3For organic WhatsApp — by far the larger volume for most Israeli businesses — stop expecting message-level attribution and lean on contact-level truth instead: since every order and every message to a given contact lives on that same person's record in a connected e-commerce and CRM platform, "how much this contact has bought since we started messaging them" is a far more honest number than trying to pin one specific message as the cause of one specific sale.
4. 4Don't import a borrowed industry number — 27 touchpoints, a specific lift percentage, an open-rate folklore figure — as if it describes your own customers. The Forrester figure above is precise, real, and describes enterprise B2B buyers for a reason; it says nothing reliable about a retail WhatsApp order.
5. 5Re-run any "channel A beats channel B" comparison with real skepticism whenever the two channels being compared don't have an equally intact tracking trail — in a WhatsApp-first market like Israel, that mismatch is closer to the default case than the exception.

Frequently asked questions

Why can't I track a WhatsApp sale the same way I track an email or SMS sale?

Because an email or SMS link opens a real webpage, which can read tracking parameters (UTM tags) out of the URL and record them. A WhatsApp chat — whether opened via a wa.me link, a saved number, or an ad — opens a conversation thread inside an app instead, with no webpage and no address bar for a parameter to survive in. The sale is just as real; it simply doesn't leave the same kind of trail by default.

What is ctwa_clid, and do I need to worry about it?

It's a click identifier Meta generates the moment someone taps a Click-to-WhatsApp ad, attached to the referral object of their first incoming message. Per Meta's own developer documentation, a business has to report that identifier back to Meta's Conversions API (as user_data.ctwa_clid) on the resulting lead or purchase event, or the sale won't be credited to the ad that caused it. It only applies if you're running Click-to-WhatsApp ads specifically — it does nothing for a wa.me link, a printed phone number, or an organic message.

Is it true that customers need dozens of touchpoints before buying?

The specific "27 touchpoints" figure traces to a real Forrester survey of B2B software and service purchase decision-makers — a slow, committee-driven buying process, not a retail or WhatsApp order. Google's own separate research on consumer purchase behavior confirms journeys loop unpredictably between researching and deciding, but doesn't publish a single universal touchpoint count the way that figure implies exists.

Should I trust last-click attribution?

Use it, but don't treat it as gospel. A 2024 EMARKETER survey run with Snap found 78.4% of senior marketers still rely on last-click to judge channel performance, yet only 21.5% are confident it reflects a channel's real, long-term impact, and nearly three-quarters are moving away from it or want to. Even people whose job is measurement don't fully trust the model most businesses default to.

What can an Israeli business actually do differently?

Tag every link-based channel (email, SMS, landing pages) with UTMs since the mechanism already works there for free; wire ctwa_clid through if running Click-to-WhatsApp ads; and for organic WhatsApp — the dominant channel for almost every Israeli business, per the Israel Internet Association's own Internet Index — stop expecting message-level attribution and track revenue at the contact level instead, since every order and every message to that contact already lives on the same record.