Google reported 5.8×. The store collected 4.7×.
PPC audit for Hydroponika, an online store selling hydroponic growing equipment. We set the ad account side by side with the orders the store actually took.
- Client
hydroponika.ro - What we audited
- Google Ads · Google Analytics 4 · orders from MerchantPro
- Period analysed
- 7.5 months · every order in the store, read one by one
- Service
- PPC audit →
What did the PPC audit find at Hydroponika?
That Google Ads reported a return of 5.8× while the store's orders confirmed 4.7×; that Google Analytics had recorded no orders since February 2026 without any report showing it; and that almost 60% of revenue came from a single paid channel.
Hydroponika is an online store with almost 60% of its revenue coming from one paid channel. The budget was set on the figures the ad platform reported, which nothing was checking any more.
The audit read every order in the store and its source, reconciled three data sources, and delivered an executive presentation plus a work list for the team that runs the campaigns.
months
with no order recorded in Analytics. No report showed it.
1 in 3
orders brought in by ads was worth under 100 lei (29.5%) — and advertising consumed about 70% of their value.
→
one customer in six brought in almost 40% of revenue. Email accounted for only 3% of their orders.
How we run a PPC audit
- 01
Three sources, not one
The Google Ads account, Analytics, and the real orders from the store's admin (MerchantPro). The differences between them are usually the story.
- 02
Read-only access
We changed nothing in the account. We used no personal data of end customers.
- 03
Order-by-order reconciliation
Every order read straight from the admin and attributed by source, then compared with what the platform reported.
- 04
An executive deliverable
A presentation with big numbers, in the language of the person paying, plus a 7-sheet working file: Summary, Monthly, Sources, Campaigns, Customers and basket, Tracking, Recommendations.
Five findings, in the order they matter for the budget
01 · Measurement
months
A single witness was deciding the monthly ad budget.
Analytics had received no purchase event since February 2026. The only source of truth left was the Google Ads tag — which nothing was checking any more.
What worked
The ad tag measured correctly: the store confirmed 82% of its value.
What we recommended
Three witnesses — complete Analytics, correct consent, server-side data — and a monthly reconciliation against the store's orders.
- Orders seen by Analytics
- Real orders in the store (index)
Chart data
| Month | Orders seen by Analytics | Real orders in the store (index) |
|---|---|---|
| Mar | 0 | 64 |
| Apr | 0 | 92 |
| May | 0 | 91 |
| Jun | 0 | 134 |
| Jul | 0 | 128 |
| Aug | 0 | 106 |
| Sep | 0 | 83 |
02 · Real return
For every 100 lei put into ads, the store collected 470. Not 580.
We read every order and its source. The gap was not a bug; it was the way the platform attributes sales. But the budget was being set on its figure — and by summer the store was confirming only 60% of what the platform reported.
What worked
The ads brought in revenue above their cost every month. In March, the store confirmed even more than the platform had reported.
What we recommended
A budget tied to the season and to the confirmed return: cost per order had risen 67% at the same monthly budget.
- February98%
- March109%
- April89%
- May73%
- June78%
- July60%
- August87%
- September84%
Chart data
| Month | Confirmed by the store, % of reported |
|---|---|
| February | 98% |
| March | 109% |
| April | 89% |
| May | 73% |
| June | 78% |
| July | 60% |
| August | 87% |
| September | 84% |
03 · Small orders
On small orders, advertising consumed almost 70% of the order value.
The volume came from small orders; the money came from large ones. Almost one order in three brought in by ads (29.5%) was under 100 lei — and advertising consumed roughly 70% of their value, before margin and shipping.
What worked
Orders over 500 lei, 8.5% of the total, brought in 42% of revenue — also through ads.
What we recommended
Ads on value, not volume: cheap products out of the bidding, a free-shipping threshold, product bundles.
Estimate from the audit
About a third of the annual ad budget could be moved towards profitable orders. An estimated figure, not a measured result.
- % of orders
- % of revenue
Chart data
| Order value (lei) | % of orders | % of revenue |
|---|---|---|
| <100 | 29.5 | 7.9 |
| 100–199 | 37 | 21.4 |
| 200–299 | 14.9 | 14.1 |
| 300–499 | 10.2 | 15.1 |
| 500–999 | 5.2 | 13.5 |
| 1,000+ | 3.3 | 28.1 |
04 · Returning customers
of existing customers' orders went through a paid ad again.
The product brought the customer back on its own. Only, in almost half of the cases, the route went through a pay-per-click ad again. Email, which doesn't cost per click, accounted for only 3% of these orders.
What worked
About one customer in six had bought at least twice — the base of loyal customers already existed.
What we recommended
Automated replenishment flows by email and SMS; existing customers taken out of paid ads.
Customers
- One order only — 84%
- Returning customers — 16%
Revenue
- From one-order customers — 61%
- From returning customers — 39%
Chart data
| Measure | One order only | Returning customers |
|---|---|---|
| Customers | 84% | 16% |
| Revenue | 61% | 39% |
- Search ads45%
- Direct / assisted46%
- Social ads6%
- Email3%
Chart data
| Channel | % of orders |
|---|---|
| Search ads | 45% |
| Direct / assisted | 46% |
| Social ads | 6% |
| 3% |
05 · Dependence
of revenue came from unpaid search. 58% — from a single paid channel.
Almost every order was bought. Any rise in click price or account problem hit sales directly.
What worked
The brand was searched for directly (16% of revenue), and the first orders from AI assistants had appeared with no effort at all.
What we recommended
Expertise content and pages that search engines and AI assistants can cite.
- Search ads58%
- Direct16%
- Assisted orders13%
- Social ads8%
- Unpaid search2%
- Email2%
- AI assistants0.1%
- Other sources~1%
Chart data
| Source | % of revenue |
|---|---|
| Search ads | 58% |
| Direct | 16% |
| Assisted orders | 13% |
| Social ads | 8% |
| Unpaid search | 2% |
| 2% | |
| AI assistants | 0.1% |
| Other sources | ~1% |
What the client received
Executive presentation, 22 screens
Big numbers, structured problem → opportunity → solution, with prices on each recommended module and a forecast simulator.
A 7-sheet analysis file
Summary, Monthly, Sources, Campaigns, Customers and basket, Tracking, Recommendations.
The work list for the campaign team
Tracking, a seasonal budget structure, excluding cheap products, retention flows.
Terms used in this study
- Reported ROAS
- The return the ad platform reports: the revenue it attributes to ads, divided by their cost. It depends on the platform's own attribution rules.
- Store-confirmed ROAS
- The same ratio, but with revenue taken from the real orders in the store's admin, attributed by source.
- Attribution
- The rule by which a sale is credited to a channel. Different platforms can attribute the same order in different ways.
- Reconciliation
- Comparing, order by order, the data from several sources — here Google Ads, Analytics and the store — to see where they differ and why.
- Returning customer
- A customer who has ordered at least twice. At Hydroponika, about one customer in six.
- PPC audit
- An analysis of a paid-advertising account and the data around it, with conclusions and a work list. It does not include managing the campaigns.
What readers of this study ask
Why did Google Ads report 5.8× when the store confirmed 4.7×?
The gap was not a bug; it was the way the ad platform attributes sales. The audit compared, order by order, what the platform reported with what the store collected: 100 lei put into ads brought 470 lei into the store, not 580. In summer, the store confirmed only 60% of what the platform reported.
What does it mean that Analytics had stopped recording orders?
From February 2026, Google Analytics received no purchase event for 8 months. The only source left was the Google Ads tag, which nothing was checking; it measured correctly, with the store confirming 82% of its value. The recommendation: three measurement sources and a monthly reconciliation against the store's orders.
What did the client receive after the audit?
A 22-screen executive presentation, a 7-sheet analysis file (Summary, Monthly, Sources, Campaigns, Customers and basket, Tracking, Recommendations) and the work list for the team that runs the campaigns.
What access did Websem have to the ad account?
Read-only. We changed nothing in the account and used no personal data of end customers. Orders were read from the store's admin and attributed by source.
How does Hydroponika show up in AI assistants?
Only 0.1% of revenue came from AI assistants, but the first orders had appeared with no effort. The audit recommended expertise content and pages that search engines and AI assistants can cite.
Did Websem improve the campaign results?
No. The study describes an audit, not campaign results: implementing the recommendations is up to the client and the team that runs the campaigns. The one forecast-type figure is marked as an estimate.
Why don't you publish the budget, revenue or number of orders?
For the client's confidentiality. The study uses percentages, ratios and indexes, which support the conclusions without disclosing commercial data a competitor could use as a benchmark.
If you recognise the situation in this study
- Marketing audit The service the PPC audit belongs to: account, tracking and data, set against real sales.
- Tracking & data GTM, GA4, server-side and consent — the three witnesses from finding 01.
- AEO audit See what ChatGPT, Gemini and Perplexity answer when asked about your brand — the starting point of finding 05.
- Another PPC audit study: Sablare Laser 28% of the budget, spent with the door closed — a local-services account, audited before the new site launched.
Want to know what your numbers say?
Percentages and ratios are real and published with the client's consent. Absolute sums and volumes (budget, revenue, number of orders) have been omitted or expressed as percentages or indexes, for confidentiality. Estimates are marked as such.