On November 2, the first day of Cyber Monday (the Argentine Chamber of Electronic Commerce's sales event, which ran November 2 to 4), an Argentine store made almost seven times the revenue of a normal October day. Google Ads was spending 2% of its usual budget and Meta more than four times its usual. People searching for the store came in anyway, through the organic result. When Cyber Monday ended, spend stayed high and revenue fell 38%.
The calendar set demand, not the budget. Many buyers were not discovering the store but coming back: 64% of those who bought on Android had visited it before. And each platform's reports are not enough to separate the sales an ad generated from the ones it only claimed: Meta reported 3.8 times the purchases our model assigned it.
Both stores are anonymized and every amount is shown as an index or a share. The study covers 40 days, October 1 to November 9, with a first-party pixel on each site and the ad platforms connected. The full methodology is in the next section.
Updated November 11, 2026: corrections and a new Store B section.
- Demand did not leave with the ad: with Google Ads off, paid and organic search replaced each other almost one for one.
- Buyers were coming back: a median of 3 visits and 1.8 days before buying.
- Each platform tells its own part: Meta reported 3.8 times what the model assigned it, Google 0.8 times, and TikTok reported no purchases.
- Spend did not follow the peak down: revenue per peso of ad spend went from 10.4x to 6.4x.
How we measured it: two stores, 40 days
Two Argentine ecommerce stores between October 1 and November 9, 2026, with a first-party pixel on each site and each ad platform's spend connected. The pixel records every visit, every product added to the cart and every purchase, along with the channel the person came from. When someone leaves their email at checkout, we stitch their visits across devices.
Across the two stores we analyzed more than 800,000 visits and close to 9,000 purchases or checkout arrivals.
To split each sale across channels we used two models: Markov, which estimates what would be lost if a channel did not exist, and last non-direct click. Neither measures how many sales an ad caused. The detail is in how each model works.
Were Cyber Monday buyers new?
Many were not discovering the store: they were coming back. At Store A, 64% of Cyber Monday buyers on Android had already visited the site before the event. On iOS we identified at least 46%: Safari deletes the pixel cookie after 7 days, as documented by WebKit, so a visit from two weeks earlier can show up as new and that figure is underestimated.
Someone sees a product in late October, comes back two days later, builds a cart and waits for Cyber Monday. When they buy, the report gives the sale to the last channel they touched. But the decision started earlier. At Store A, the typical Cyber Monday buyer made 3 visits before buying.
Buying during the event involved more consideration time. We use the second half of October as the comparison because the pixel started measuring on September 30: in early October nobody had history and almost everyone looked new.
Was what people bought on Cyber Monday already in the cart?
Partly, and less than people usually think. 19% of the products bought during Cyber Monday had been added to the cart on an earlier day. On the other hand, of the people who added something to the cart in October, only 1.8% ended up buying during the official Cyber Monday.
Store A · people who added to cart in each period: whether they bought, and when
Among the people who discovered the store during the pre-sale, another signal showed up. Of those who added a product to the cart, 30% came back to the store during Cyber Monday. Of those who added nothing, 11% came back. These are people who returned, not necessarily buyers. Still, the gap shows that earlier behavior separates levels of interest: someone who browsed products, built a cart and came back days later left a very different signal from someone who visited once.
If the sale happened during Cyber Monday but the decision started earlier, which campaigns and content built that interest? The ROAS on the day of the purchase does not answer it.
How many of your event buyers had already visited you, and how many days passed between the first visit and the purchase.
What happened when ads switched off?
Sales kept going. Between October 31 and November 2, Store A went through ad interruptions: Meta Ads was off for two days and Google Ads for almost three. While visits from Google Ads fell 90%, visits from Google organic grew 4 and 8 times, depending on the day. When Google Ads came back on November 3, organic traffic dropped again.
People searching for the store did not stop coming in: they used a different door. Google campaigns that buy searches for the brand's own name send people to the home page, and someone who already knows the store and searches for it may click the ad at the top even with the organic result right below. If we add up the home page visits that came from Google, through the ad or the organic result, the total followed event demand, not whether the ad was running. In October, 14 out of every 100 of those visits came through organic; on November 1 and 2, with the ad off, 98 out of 100 did. We described it in paying Google Ads for clicks you already had for free.
| Day | Home visits | Via organic |
|---|---|---|
| Oct 15 to 28 | 100 | 14% |
| Oct 31 | 220 | 48% |
| Nov 1, no ad | 177 | 98% |
| Nov 2, no ad | 340 | 98% |
| Nov 3, ad back | 311 | 64% |
| Nov 4 | 244 | 27% |
Store A. Visits that landed on the home page from Google, through an ad or the organic result, with the October 15 to 28 daily average equal to 100.
The Meta cut showed another side. On the two days without Meta, visits from that platform fell 95% compared with the two previous days, and still the store made 2.9 and 3.3 times the revenue of a normal October day. Those days there was an email send that multiplied the channel's usual visits by 19, and owned social brought 3.6 times its usual traffic on Sunday. We do not know what would have happened with Meta on. We know the store kept selling and other channels carried the movement.
On November 2 ads were not fully off: Google Ads was still at 2% of normal, but Meta came back and spent 4.4 times an October day. Total spend was 2.7 times the usual and revenue 6.8 times. Sales that closed on an ad multiplied by 2.9; sales that closed through channels that are not paid per click, by 8.9.
| Channel (last non-direct) | October | November 2 |
|---|---|---|
| Organic search | 8% | 32% |
| Meta Ads | 17% | 18% |
| Owned social | 13% | 18% |
| 8% | 15% | |
| Google Ads | 29% | 3% |
Store A. Each channel's share of sales by last non-direct channel. Meta kept almost the same share as in October: it grew at the pace of the total.
What we can conclude: with the branded ad off, paid and organic search replaced each other almost one for one: organic went from 14 to 98 out of every 100 home page visits from Google.
What we cannot conclude: how many sales keeping Google Ads on would have added, because the cut overlapped with the event peak and was not a controlled experiment.
Separate brand campaigns from generic ones, and test pausing the brand campaign by region or time slot before deciding.
Did Meta make people search for the store on Google?
It is the obvious question: if Meta came back the same day organic search took off, maybe people saw the ads and then searched for the store on Google. We checked it two ways and neither supports it.
The first is the pixel. Of the people who came in through organic search on November 2, only 8% had clicked a Meta ad in the previous 30 days, compared with 4% in October, and six in ten had never visited the store. But the pixel only sees clicks: someone who saw an ad without tapping it and later searched for the brand shows up as a new organic visit.
To cover that gap we looked at Search Console, which counts how many times the store appeared on Google when someone searched for its name, whether or not they clicked. The logic is simple: if Meta ads led people to search for the brand, days with more Meta spend should have more searches, and days with Meta off, fewer.
| Day | What was happening | Branded searches | Meta spend |
|---|---|---|---|
| Oct 30 | Presale | 2.7x | 2.6x |
| Oct 31 | Presale, Meta off | 2.9x | Off |
| Nov 1 | Presale, Meta off | 3.2x | Off |
| Nov 2 | Cyber Monday starts, Meta returns | 5.2x | 4.4x |
| Nov 3 | Cyber Monday | 4.0x | 6.5x |
| Nov 4 | Cyber Monday | 3.4x | 4.7x |
| Nov 5 to 9 | After Cyber Monday | 2.7x to 1.6x | 4.0x to 5.6x |
Store A. Searches: daily Google impressions for searches that include the store's name, from Search Console. Spend: daily Meta Ads spend. Both columns compare each day with an average day from October 1 to 28: 1x is a normal October day and 3.2x is a little over three times that.
The table reads in three moments:
- With Meta off, searches did not drop. On October 31 and November 1, in the middle of the presale and with no Meta ads, the store showed up in almost three times as many searches as on a normal October day (2.9x and 3.2x). Slightly more than on October 30, when Meta was on.
- More spend did not bring more searches. November 3 was the day of highest Meta spend (6.5x), and searches fell from the day before, from 5.2x to 4.0x.
- When Cyber Monday ended, searches dropped and spend did not. From November 5 to 9 Meta kept spending between 4 and 5.6 times its normal level, and searches fell to 1.6x, close to October.
In October, with no event, the same thing happened: the days with more Meta spend did not have more branded searches, neither that day nor the next. The only day we cannot separate is November 2, when the start of Cyber Monday and Meta's return overlapped.
In the observed data, branded searches followed the event calendar and we found no relationship with Meta spend, neither the same day nor the next. Cyber Monday brought the demand, and since Google's branded ad was not there, those people came in through the organic result. That does not mean Meta added nothing: it added what shows in its own journeys, 18% of the record day's sales. It does not rule out slower effects either, but the hidden effect people usually assume does not show up.
How much of the demand attributed to ads was created by them, and how much already existed? They are two very different things, and an attribution report alone cannot separate them.
Cross daily Meta spend with branded impressions in Search Console, same day and next day.
Was the pre-sale more efficient?
We cannot say so. Store A launched its own pre-sale on Thursday, October 29 at 1 p.m., and those four days had the best ratio of total revenue to ad spend of the six weeks: 17.2 pesos of revenue per peso of ad spend. But two of those four days were the days of the Meta cut. With full ads, Thursday and Friday of the pre-sale came in at 9.9x, practically the same as the official Cyber Monday (10.4x). The 17.2x comes from the days with ads almost switched off.
Store A · total revenue per peso of ad spend
What is interesting is not that the pre-sale performed better, but that the store sold a lot on days with very little advertising. One hypothesis is that the pre-sale captured demand that had been building up during October. It fits what we saw in the buyers: many already knew the store before the event.
What we can conclude: with full ads, the pre-sale performed almost the same as Cyber Monday: 9.9 against 10.4 pesos of revenue per peso of ad spend.
What we cannot conclude: that the pre-sale was more efficient. The 17.2x comes from the two days with Meta off, and we do not know how much would have sold on those days with ads on.
Why does Meta claim 3.8 times the purchases our model assigns it?
Because each system answers a different question. During Cyber Monday, Meta claimed almost four times the purchases it got when we split credit across every channel each buyer went through: 3.8 times what the Markov model assigned it. That does not mean Meta exaggerates. Google Ads reported fewer purchases than Markov assigned it, and TikTok reported none, even though it showed up in journeys that ended in a sale. Meta, Google and TikTok use their own rules and windows, and one sale can have several protagonists. We already explained why Google Ads and Meta Ads claim the same sale.
2.9ROAS per Markov during Cyber Monday. 69% of spend and 19% of attributed revenue.
3.8ROAS per Markov during Cyber Monday. 30% of spend and 11% of attributed revenue.
3.7ROAS per Markov during Cyber Monday. 1% of spend and 0.3% of attributed revenue.
There is another reason not to read this as a verdict on Meta. Its share of spend (69%) and of attributed revenue (19%) cannot be compared as if they were profitability: one is spend and the other is split credit. And among the most frequent two-step journeys are people who arrived from Meta and came back later through owned social or direct. Meta may sit at the start of journeys that close in other channels. Meta's dashboard is still useful: it tells one part of a story that includes other channels, decisions and moments.
The chart below looks at something else: the people who arrived for the first time each day from Meta or Google, and what they bought in the next 7 days against what was spent that day. It measures the ability to bring in new buyers, not retargeting or incrementality.
During Cyber Monday it brought almost four times more new people per day than in October, and bringing in each one cost between 33% and 55% more. 1.4% of those who arrived in October ended up buying, and 1.0% of those who arrived during Cyber Monday, who had fewer days to do it. Of those who arrived after Cyber Monday, 0.7% had bought by November 9.
The clicks arrived: the pixel recorded 8 out of 10. Of the people who discovered Store A through TikTok, 13% came back, against 29% of those who came through Meta or Google, and 0.2% bought. If we give those people full credit for everything they bought over the 40 days, revenue was 2.4 times TikTok spend. The 3.7 on the card above comes from a different calculation: the credit Markov assigns TikTok across Cyber Monday purchases, including those by people who already knew the store.
Compare what each platform reports with a journey-based model, without taking either as causal truth.
How much does each platform claim at your store? Compare what Meta, Google and TikTok report with your buyers' real journeys.
How many purchases arrived without going through an ad?
41% of Cyber Monday purchases had no ad in the journey we could observe. In another 22%, advertising showed up at the start and the last visit came from search, email, owned social or direct. The most frequent path during Cyber Monday was coming in from Google organic and buying.
Store A · most frequent journeys to purchase, Cyber Monday
Store B · most frequent journeys to "Pay", Cyber Monday
When we split the credit for each sale across the whole journey, owned channels also gained prominence. Under Markov, organic search went from 8% of Store A revenue before the event to 18% during Cyber Monday, while Google Ads went from 25% to 11%. Under last non-direct click the relationship holds, 24% against 12%, so the signal does not depend on a single model. It does not prove that SEO produced more incremental sales than Google Ads: it means that, within the observed journeys, organic search got more credit.
Store A · share of revenue by channel (Markov)
See the data: Markov vs last non-direct, both stores
| Channel | Before | Pre-sale | Cyber | After |
|---|---|---|---|---|
| Store A · Direct | 29% · 28% | 20% · 18% | 20% · 15% | 21% · 17% |
| Store A · Meta Ads | 16% · 16% | 17% · 16% | 19% · 16% | 28% · 28% |
| Store A · Organic search | 8% · 8% | 12% · 14% | 18% · 24% | 4% · 6% |
| Store A · Organic social | 12% · 11% | 15% · 14% | 16% · 17% | 12% · 12% |
| Store A · Email | 8% · 8% | 16% · 19% | 13% · 14% | 11% · 12% |
| Store A · Google Ads | 25% · 27% | 17% · 18% | 11% · 12% | 21% · 23% |
| Store A · TikTok Ads | 0.7% · 0.4% | 0.6% · 0.9% | 0.3% · 0.2% | 0.6% · 0.8% |
| Store B · Organic social | 20% · 20% | 30% · 30% | 27% · 27% | 26% · 25% |
| Store B · Google Ads | 19% · 18% | 17% · 16% | 18% · 17% | 20% · 18% |
| Store B · Organic search | 15% · 15% | 11% · 11% | 17% · 16% | 12% · 13% |
| Store B · Meta Ads | 15% · 13% | 22% · 21% | 12% · 11% | 17% · 15% |
First value: Markov. Second: last non-direct channel in the previous 30 days. For Store A it is the share of revenue; for Store B, the share of people who reach checkout. Rows do not add up to 100 because minor channels and sales with no recorded journey are left out.
We did not see channels working in isolation: we saw people moving between them. A channel can look less important if you only look at the purchases it claims directly, because conversion can happen in a different channel from the one where the relationship with the brand started.
What share of your sales had no ad anywhere in the journey.
The other store: when Meta traffic fell, intent rose
At Store B, Cyber Monday visits fell 55% from October and the people who clicked "Pay" rose 56%. Almost all of the drop was Meta traffic: those visits fell 61%, while visits from every other channel grew 17%.
Store B · visits (October = 100)
Store B · people who reach checkout (October = 100)
That traffic was large, but it converted little. In October, 92% of visits came from Meta and only 12 out of every 10,000 of those people reached checkout. During Cyber Monday, out of every 100 people who came in through organic search, 12 reached checkout; through Meta, fewer than one.
Store B, Cyber Monday: out of every 100 visitors who came through each channel, how many clicked "Pay".
We do not have this store's Meta spend, so we do not know whether the drop came from a budget cut, a campaign change or something else. We cannot separate the effect of the event discounts either. What it does show is that a source can fill the site with visits without creating opportunities: judged by clicks, CPC or sessions, that traffic looks like a success.
Buying also works differently from Store A: it is more impulsive. Only 1 in 5 people who reached checkout during Cyber Monday had visited the store before, and most did it on the same day they arrived.
Checkout arrival rate by channel, not just CPC and sessions.
What changed in the order?
At Store A, Cyber Monday changed what people put in the order. The published price did not move: the discount was applied at checkout and, at the median, people paid 15% less than the list price. Units per order rose 32% and the average order fell 13% compared with October. More orders and more units did not mean a higher ticket. We do not have costs or margins to know how profitability evolved, but we do know that growth came with a different cart mix.
Timing changed too. In the first hour of November 2, about seven times more orders came in than on a normal midnight: the event concentrated demand in a very short window.
What happened to spend when Cyber Monday ended?
It stayed just as high while demand came down. From November 5 to 9 promotions stayed on and daily spend stayed at the same level as during the official Cyber Monday. Daily revenue fell 38% from the official days, though it was still almost four times October's. What changed most was efficiency: revenue per peso of ad spend went from 10.4x during Cyber Monday to 6.4x, below even October (7.5x), and ad cost per order rose more than 50% compared with the official days.
Searches for the store on Google did drop: they were half of what they had been on the official days. Already during Cyber Monday, bringing in a new person from Meta had cost between 33% and 55% more than in October, and from Google Ads between 2 and 2.7 times more. We do not know what the optimal budget would have been, because that requires measuring incrementality and margin. What we saw is that the end of the demand peak did not coincide with a drop in spend.
An alert on revenue per unit of ad spend for the day after the event, with a threshold set in advance.
What to check before the next sales event
At these two stores, the calendar set demand, not the budget, and no single report was enough to explain what happened. That does not mean it repeats in every ecommerce store, or that SEO is worth more than Google Ads, or that cutting ads improves sales. It does leave three questions worth asking before the next event.
- How much demand was built before the purchase?A sale recorded during Cyber Monday can have several days of history: earlier visits, carts and brand searches.
- How much additional value does advertising create?Taking part in a sale and causing a sale that would not have happened are different things. The ad cuts during this Cyber Monday made that visible. With Google's branded ad off, people searching for the store came in through organic anyway.
- What do we stop seeing when we measure each channel separately?People do not buy along the boundaries of dashboards: they start in one channel and finish in another.
Before the next event:
- Measure how many event buyers already knew you.
- Separate spend on branded and generic searches.
- Compare what each platform reports with a journey-based model.
- Decide in advance when to lower spend after the peak.
CyberMonday lasted three days. We watched forty. Much of what was interesting happened outside those three days: in earlier visits, open carts, searches, emails and decisions that were not yet purchases. The day of the sale is easy to spot. The journey that made it possible is much harder to understand. That is what we tried to observe with Dashcrab Attribution: not to find a winner among Meta, Google, TikTok or owned channels, but to understand how spend, people's behavior and a business's real results connect.
Do your ads generate sales, or just claim them? Run this same analysis on your own data: full journeys, attribution models side by side and what happens outside each platform.
Keep reading. We ran the same exercise with a Mercado Libre seller, where there is no first-party pixel and the platform that sells the ads also reports how much they sold. We cover it in did advertising explain Mercado Libre sales on Cyber Monday 2026?
Glossary
- Attribution: the rule used to split the credit for a sale across the channels the buyer went through.
- Markov: a model that looks at every journey, with and without a purchase, and gives each channel the share of sales that would be lost if that channel did not exist.
- Last non-direct click: gives the whole sale to the last channel the person came in through, ignoring direct visits.
- Incrementality: the sales that would not have happened without the ad. It is measured with experiments, not with an attribution model.
- Revenue per peso of ad spend: site-wide revenue divided by total ad spend. It is not a platform's ROAS, which only counts the sales that platform claims.
- Cohort: the group of people who arrived for the first time in the same period, followed over time.