The Q1 awareness spike was real, unbought, and badly measured.
Amazon's "Customers in Awareness" quadrupled in February 2026 while branded search
went nowhere. Four explanations were on the table. Three are dead, one survives, and the metric at
the centre of the argument is roughly ten times larger than anything real it could be measuring.
Twelve independent measures now say so.
Updated 20 Aug 2026Window: Aug 2025 – Jul 2026Northbeam · Daily Stand · Amazon Brand Analytics
The short version
Something real happened. It wasn't our campaign, and we didn't buy it.
Verdict
February's Amazon surge is real — +48% new customers — but it arrived with
13% fewer paid clicks and flat branded search across three independent
measurements. That rules out our media and rules out our brand campaign. What rose
fastest was head-to-head comparison searching against RYZE (+20%), one month
after its 1,900-store Target launch. So: category demand we captured rather than created. The
"Customers in Awareness" metric points the right direction and overstates the size by roughly
6–15×. It should not be used as evidence for anything.
Amazon new customers Jan → Feb 2026
+48%
6,362 → 9,415. Real orders, real money.
Amazon paid clicks over the same period
−13%
33,625 → 29,280. The lift was not bought.
Branded demand, Jan → Feb three independent measures
+3 / +4 / +7%
Amazon metric, our Google campaign, and 1,133 raw query strings. All flat.
"Customers in awareness" Jan → Mar 2026
+328%
The one number that moved — and the one we can't trust.
The three things worth taking away
Your head of growth is right, for a better reason than the one he gave. He argued
FYF didn't cause the halo because branded search didn't move. True — and confirmed independently on
Google. But the stronger proof is a dose-response test: we spent 2.5× more on
FYF in March and awareness moved +2%; we cut FYF by 99%
in April and awareness hit its peak. That is not a campaign effect.
The awareness metric contradicts its own definition. Amazon defines it as a
trailing-12-month count. A trailing-12-month count cannot quadruple in one month and return to baseline
five months later. August 2025 had more Amazon ad impressions than February 2026 with
one-eighth the awareness. The number is not a stable measure of anything.
The "Q2/Q3 die-down" is mostly us, not the market. We cut Amazon spend 65% from
March to July and impressions 71%. Site visits fell 45% from January. Framing that as a halo fading
lets a real, self-inflicted decline go unexamined.
And your hunch about brand terms was right — two to one. Only
31.3% of brand search uses the canonical "mud wtr". The phonetic
spellings, "mud water" and "mudwater", are 61.3% between them. Nobody
types the backslash. Section 5.
Section 1
The chart, rebuilt so it can't mislead
The original Brand Analytics chart puts two series on two different y-axes. That is the
single most common way a chart creates a relationship that isn't there — you can make any two lines
converge or diverge just by choosing the scales. Here are the same two series, each on its own panel,
followed by everything indexed to a common base so the comparison is honest.
Customers in Awareness Tier C
Searched our brand or viewed our product in search results without purchasing,
trailing 12 months.
Branded Search Customers Tier C
Customers who performed a branded search on Amazon.
Everything indexed to January 2026 = 100 Tier A+ Tier C
Same base, stacked panels sharing an x-axis, because one series moves 5× and the
rest move a few percent — putting them on one scale would flatten the ones that matter into a
single line. Awareness leaves the field. Every measure of actual demand for us stays flat
or falls.
Awareness · index
Real demand · index, same base
Read this before quoting the awareness number
The awareness and branded-search series are digitised off a screenshot of the Brand
Analytics dashboard, not exported. Values are accurate to roughly ±5% and the shape is
reliable, but no argument here rests on their precise level — and the one argument that would
(the magnitude of the spike) is exactly the one I am telling you not to trust. Export the
underlying CSV before anyone puts these figures in a board deck.
Section 2
Four explanations, tested one at a time
Everyone in this thread has a hypothesis. Rather than argue about which is most
plausible, each one makes a prediction that the data can check. Three fail.
H1
The FYF campaign created a brand halo
Ruled out
The prediction: if a top-of-funnel brand campaign built awareness, more spend
should build more awareness, and the newly-aware should show up as branded searches.
What happened: neither. tof_awr_eng_fyf spent nothing
before February, $15,588 in February, $38,737 in
March, then $357 in April. Awareness went
318k → 325k → 380k across those same three months. Spend rose 2.5× and
awareness moved 2%. Spend fell 99% and awareness hit its peak.
Dose and response, side by side
If the left panel caused the right panel, the two shapes would rhyme. They are
close to inverted.
FYF campaign spend · $ per month
Customers in awareness · thousands
The honest complication: across all twelve months,
awareness/engagement spend is the strongest correlate of awareness in the whole dataset
(r = +0.58). That is precisely why correlation isn't enough. Both series happened to rise
in February and March. The month-by-month dose-response is what separates a cause from a
coincidence, and it fails.
H2
Our own Amazon media put us in front of more people
Ruled out for awareness · partial for sales
The prediction: awareness tracks our Amazon ad reach. More impressions, more
people who saw us.
What happened: the ratio between the two swings by 14×
across the window, which is another way of saying there is no stable relationship at all. Two
observations kill it outright:
August 2025 had more Amazon impressions than February 2026
(3.88M vs 3.55M) and one-eighth the awareness
(40k vs 318k).
April 2026 is the awareness peak at 380k — on the
second-lowest impression month in the window (2.54M) and a
37% spend cut. The peak of awareness sits at a trough of our own activity.
Awareness per 1,000 Amazon ad impressions
If our media drove the metric this line would be roughly flat. It moves 14-fold.
Where our media does matter is sales, and only partly — see
Section 3.
H3
The metric itself changed, and no behaviour did
Substantially true
The prediction: if this is a definitional change or a backfill, the series will
behave in ways the stated definition forbids.
What happened: it does. Amazon defines Customers in Awareness as people who
searched our brand or viewed our product in the last 12 months. Under that definition, the
roughly 240,000 people who entered the pool in February are still inside
the 12-month window in July. The count should have stayed high. Instead it went
318k → 325k → 380k → 270k → 118k → 82k — all the way back to its
pre-spike level in five months.
A trailing-12-month unique count cannot do that. Either the window is not really 12 months,
or Amazon restated the series, or the axis means something other than what the tooltip says. Any
of those has the same consequence: this number cannot carry a strategic argument.
Note the two Actions event markers Amazon places on the chart at 31 Dec and 31 Jan —
the second sits directly on the discontinuity. Worth asking your Amazon rep what they were.
H4
A competitor activated the category and we drafted off it
Leading explanation
The prediction: category demand rises, our branded demand doesn't, and we pick
up customers who were shopping the category rather than looking for us.
What happened: every observable matches.
RYZE launched nationwide in 1,900+ Target stores on 8 January 2026 — its first
national retail launch — and Forbes ran a feature on the brand and the category on 9 January.
That is a mass-audience introduction to mushroom coffee, one month before our spike.
Our branded search stayed flat on Amazon (+5%) and Google (+10%). These
shoppers were not looking for MUD\WTR.
Our Google conquesting impressions rose 27% January to March — we can only
appear against competitor brand terms if people are searching them.
Fewer, better clicks. Amazon paid clicks fell 13% while new customers rose
48%. That is higher-intent traffic arriving, not more traffic bought.
It faded when the news cycle did.
New evidence, added after the first version of this report. The Google
Ads matched-search-term report (Section 5) gives a test the other hypotheses cannot pass.
Comparison queries — "mudwater vs ryze", "mud wtr vs ryze" and their variants — rose
20% from January to February, 15,601 to 18,706 impressions, hitting their highest
share of all our search impressions until May. You cannot bid a comparison query into
existence; someone has to type it. Meanwhile our own brand queries rose 7% and generic
category queries 6%.
So February's shift was not broad category curiosity and it was not new interest in us. It
was head-to-head cross-shopping against one competitor — precisely what a
rival's national retail launch produces, and the one pattern no other hypothesis predicts.
Confidence: medium-high, up from medium. Still short of proof: +20% is
modest, our conquesting bids partly confound the RYZE line (Section 5), and I have no
Amazon-side search-term data to show the same thing happened on Amazon — which is where the
sales actually moved. Amazon Brand Analytics → Search Query Performance closes that gap.
Section 3
What actually happened in February
Strip out the awareness metric and there is still a real event to explain. Amazon new
customers stepped up sharply on 10 February and stayed up. Amazon went from 21% to
35% of all new customers in a single month.
Amazon new customers per day · February 2026
A step, not a ramp. 211/day before the 10th, 395/day after.
Amazon ad spend per day · February 2026
We stepped spend up on the same day, +65%.
The number that decides it
We raised Amazon spend 65% and got 88% more new customers, which looks like a media win. But
across the full month paid clicks fell 13% (33,625 → 29,280) while new customers rose
48%. You cannot buy a 48% customer increase with 13% fewer clicks. Most of February's
lift walked in on its own; our spend increase rode along with it, at a 12% better cost per customer.
The two facts sit together like this: our media got more efficient because the traffic got
warmer, not because we got better at buying it. Cost per Amazon new customer improved from
$5.79 to $5.12. That is the signature of demand you didn't create — the same ads convert better when
the person already wants the category.
Amazon: what we bought versus what we got — indexed, Jan 2026 = 100
Indexed because impressions run in millions and customers in thousands; on a shared
raw axis the two that matter would flatten to nothing. Watch February: clicks go down, customers go
up. That divergence is the whole argument.
Data gap you should know about
The Daily Stand's Amazon rows stop after 1 March 2026 — the feed broke and only
one day of March was ever recorded. The sheet still shows a March Amazon figure of $13,553 and 424
orders; both are artefacts of that single day, not the month. Northbeam has the real March spend
($57,515), so spend is recoverable, but we have no Amazon
new-customer counts from March 2026 onward. That is why this section stops at February.
Fixing that feed is the single highest-value thing on the list.
Section 4
Everything, twelve months, one panel each
Same window, same x-axis, each metric on its own scale. February and March shaded.
Month over month
▲ / ▼ mark moves of 8% or more. Direction only —
no judgement implied. Fewer Amazon clicks in February is the strongest evidence in this report,
not a problem.
What correlates with awareness — and why that's a trap
Ranked correlation across all twelve months. Note that nothing correlates strongly, the top
correlate fails its dose-response test, and site visits — a real measure of demand — correlates
negatively. This table is here to show that the correlational approach cannot answer the
question, not to answer it.
Section 5
Your question, answered: almost nobody types the backslash
You asked whether more people search mud, mud water,
mudwtr and misspellings than MUD\WTR, and what our brand-term list actually
contains. I pulled the Google Ads matched-search-term report — real query strings, full depth,
10,000 rows a month, Nov 2025 to Jun 2026. You were right, and the split is almost
exactly two to one.
Phonetic "water" spellings share of brand impressions
61.3%
"mud water" + "mudwater"
The canonical "mud wtr" share of brand impressions
31.3%
1.96× less than the phonetic spellings combined
Distinct brand queries Nov 2025 – Jun 2026
1,133
1.24M impressions, 345k clicks
Queries containing a literal backslash
~0
Nobody types it. It is a logo, not a search term.
How people spell us Tier A
Brand search splits into near-perfect thirds. Any brand-term list missing
"mud water" and "mudwater" misses about two-thirds of brand demand.
What this changes. It makes the central finding stronger, not weaker. My
"branded search was flat" claim was originally built on Amazon's branded-search metric plus our
Google branded campaign impressions — both of which depend on someone else's term list.
This pull doesn't: it counts actual typed queries across all 1,133 spellings. And it agrees.
Our brand queries went 169,832 in January to 181,296 in February — up 6.8%.
Three independent measurements of branded demand (Amazon +3%, our Google campaign +4%, raw
query strings +7%) now say the same thing. Nobody was newly looking for us.
One incidental finding worth passing to whoever owns the site: "mudwater login" drew
15,007 impressions over the window. People use Google as our account-login door.
Query volume by type, monthly Tier A
Our brand queries are flat all year. RYZE-term queries — which we appear
against because we bid on them — climb from below ours to nearly 4× ours.
Brand · competitor · category
Comparison queries — own panel, 20× smaller
The confounder, stated plainly
These are our ad impressions, so they depend on our bidding as well as on query
volume. We raised conquesting spend 64% in February and 144% by March, so the RYZE line is
partly us buying more, not purely more people searching. The exception is our own brand
terms: we hold essentially full impression share on those continuously, so that line
is a genuine volume read — and it's the flat one.
Section 6
Every real measure agrees. One number doesn't.
This is the whole report in one chart. Twelve measures of what changed between
January and February 2026, from four independent systems. Eleven of them land between −13% and
+89%. One lands at +318%.
Change from January to February 2026, every measure available
Amber = the Amazon awareness metric. Teal = it went up. Pink = it went down.
Hover any row for its reliability tier.
An estimate can be wrong. Twelve estimates from Northbeam, the Daily Stand, Google Ads and
Amazon do not all drift the same way by accident. The real February event was a lift somewhere
in the range of +20% to +50% — visible in Amazon new customers (+48%),
new-to-brand orders (+89%) and cross-shopping queries (+20%). The awareness metric is
roughly 6–15× larger than anything it is supposed to be measuring. That is what
"overstates by an order of magnitude" means, and this chart is the receipt.
The ordering is the interesting part
Look at which real measures moved most. Comparison queries (+20%) and RYZE-term queries
(+15%) rose faster than generic category queries (+6%) or our own brand queries (+7%). The
February shift wasn't broad curiosity about mushroom coffee and it wasn't new interest in us —
it was people actively comparing us against a specific competitor. That is the
signature H4 predicts, and no other hypothesis predicts it.
Section 7
What to do with this
Settle it
1
Pull Amazon Brand Analytics → Search Query Performance for Dec–Apr. If category
and competitor queries rose while ours didn't, H4 goes from leading to proven. This is the single
query that ends the debate, and it also answers the brand-terms question.
2
Ask your Amazon rep what the two "Actions" markers on 31 Dec and 31 Jan were,
and whether Customers in Awareness was restated or redefined in early 2026. One email.
3
Get Sponsored Products new-to-brand orders. We only have Sponsored Brands, which
is why the paid-versus-organic split in February is bounded rather than settled.
4
Add "mud water" and "mudwater" to every brand-term definition we own —
Amazon's branded-search reporting, our Google brand campaign, any dashboard that segments
branded vs non-branded. Two-thirds of brand demand uses a spelling the canonical list misses,
so every branded metric we report is understated by roughly 3×. This is a reporting fix, not a
bidding one; we already capture the clicks.
Fix
5
Repair the Daily Stand Amazon feed. Dead since 1 March. We are five months blind
on Amazon new customers — the channel that was 35% of all new customers in February. This is the most
expensive gap in the report.
6
Stop reporting Customers in Awareness. Until Amazon explains the discontinuity it
is worse than no number, because it invites exactly the argument this thread has been having.
Decide
7
Treat the category as a channel. If competitors' retail launches send shoppers to
Amazon to compare, then Amazon share-of-shelf and conquesting are how we monetise other people's
marketing. February shows the return on that is good: 48% more customers at 12% lower cost. We then
cut Amazon spend 65% by July.
8
Separate the halo debate from the decline. DTC new customers went 18,300 in
February to 10,355 in July. Site visits fell 45% from January. Whatever happened in Q1, that is the
number that needs a plan, and arguing about attribution for a spike is a comfortable way to avoid it.
The reframe worth having
The thread is asking "did FYF cause the halo?" Both possible answers are consolations. If yes, we
spent $54k and got a spike we then failed to sustain. If no, a competitor's Target launch moved more
volume for us than our own brand campaign did. The more useful question is the one neither side is
asking: we now know what warm category traffic is worth to us on Amazon. Why are we not
buying more of it?
Section 8
Sources, reliability, and what I could not verify
Reliability tiers
Tier ANorthbeam (Clicks Only, cash
accounting) — Amazon ad impressions, spend, clicks, Sponsored Brands new-to-brand orders, Google
campaign spend and CPM, site visits. Pulled month by month, 12 separate exports.
Daily Stand — DTC spend, DTC new customers, blended CAC, Amazon new customers
through February. Google Ads matched-search-term report via Supermetrics —
real query strings, 10,000 rows per month, Nov 2025 – Jun 2026, account 5666511290.
Tier CAmazon Brand Analytics —
Customers in Awareness and Branded Search Customers, digitised from your screenshot. Shape reliable,
levels ±5%, and the series is internally inconsistent with its own stated definition. Every
conclusion in this report is built to survive without it.
Derived figures
Google campaign impressions are computed as spend ÷ CPM × 1,000;
Northbeam does not expose campaign impressions directly on this export.
February daily figures use the 28 calendar days. The sheet's month-to-date cell sums 29 values
and therefore overstates February by one day ($51,642 / 9,839 versus $49,456 / 9,415). The 28-day
spend reconciles to Northbeam's independent $49,485 within 0.1%.
Amazon new customers come from the Daily Stand (Shopify-side). Northbeam's new-to-brand orders
are Sponsored Brands only and are roughly 10–13% of that total — the two are not interchangeable and
are never mixed here.
Confounders I could not remove
Conquesting impressions are contaminated by our own bids. We raised conquesting
spend 64% in February and 144% by March. Rising impressions are therefore partly purchased. Impression
share would fix this; we don't have it on this export.
"Mushroom Coffee" non-brand impressions track our budget almost exactly and carry
no independent information about category demand. I have deliberately not used them as evidence.
Paid versus organic Amazon growth is bounded, not resolved. The 13%-fewer-clicks
argument holds regardless, but the precise split needs Sponsored Products new-to-brand.
No Amazon new-customer data after 1 March 2026. March and April, the two months
where awareness peaked, have no Amazon customer counts at all.
Corrected after publication
Google Trends is not blocked by 1Password — it is not on the Supermetrics
licence. The first version of this report said the key was unreachable. Wrong: the key
was in a local env file all along, and it authenticates fine. The actual error is
LICENSE_DATA_SOURCE_NOT_AVAILABLE — Data source Google Trends (GT) is not
available in your licence. No amount of unlocking fixes that; it needs a plan change, and
it is not worth one. The Google Ads search-term report in Section 5 is strictly better data
anyway — real query strings and absolute volumes instead of a normalised 0–100 index.
Authenticated Supermetrics sources on this licence, verified 20 Aug 2026: Amazon Ads,
Google Ads, Facebook Ads, Google Analytics, Shopify, Recharge. Not Google Trends, not
Instagram.
Still blocked
Meta campaign detail via API — the Meta MCP server is returning a field error on
every insights call. FYF spend in this report comes from Northbeam's campaign-level export instead,
which independently reproduced the figures.
