Case study: Stormotion — copy (EN)

IT outsourcing · mobile and web development · US market

Stormotion: we cut a channel we were being paid for, to save the site

Lead: The site had lost roughly 80% of its traffic. Over two years of work we brought back 2.75× from the bottom. The most important decision in this case is something we stopped doing, not something we added.

• 2.75× traffic from the low point across 24 months
• We ended a link building channel we were being paid to run
• Inbound inquiries up 160% in six months, two thirds in the core verticals

Small print under the bar:

Ahrefs data, October 2023 to September 2025. Values were read off the Ahrefs chart; traffic there is an estimate from its own index rather than a counter reading.

What happened

Stormotion’s traffic peaked around the turn of 2022–2023 at roughly 6,000 visits a month. Then it fell, kept falling until autumn 2023, and settled at about 1,200 visits.

The site had lost around 80% of its organic traffic.

When this work started

In October 2023 a follow-up appears in the project with a blunt title: “Work resumed.” Before that the engagement had been occasional — mostly under 10 hours a month. From that point it became systematic, averaging 38 hours a month for two years.

That stretch is this case.

Work out why the site lost its traffic and get it back — without guesswork, and without “let’s just publish more articles.”

When a site loses 80% of its traffic the temptation is obvious: grab at everything at once. We went through our WEDGE framework instead. It answers where a limited resource should go, and in a crisis that question is sharper than usual.

W — Where the client can win

Stormotion’s competence map is dated January 2023 — reworked exactly when it became clear the decline wasn’t random.

Which verticals rose is visible in how the client still answers inquiries today: each one gets a matching case deck — fitness, IoT, mental health, healthcare, EV.

Those four directions now account for two thirds of all contacts.

E — Evaluate whether we can get in

The tracker carries a task with a telling name: “Analysis of 10 target markets.”

Ten directions went through the second screen — competition, available traffic, whether the site in its current state could reach them. With a site under a ranking loss, that last question decides everything: some directions are out of reach not because of competitors but because the site lacks weight.

6.1. First we checked whether anything was actually broken

Worth showing, because it happened earlier — in August 2021, during a previous dip.

We ran an analysis then too. The conclusion was the opposite: nothing needed fixing. Updates had an effect, but the site was recovering on its own; the main driver was seasonality, an annual decline from May to September three years running; plus one traffic-heavy page had lost clicks to a Google answer box while its positions barely moved.

The decision that time: carry on as normal.

An agency that answers every dip with an urgent remediation plan bills more. But in 2021 there was nothing to treat, and we said so.

D — Decode who signs the contract

Step D is visible directly in the task names: “Brief for 1 blog text (persona)”, “Marketing review of 1 article (Persona)” — monthly.

The persona doesn’t sit in a document. It’s a separate review stage on every text.

G — Guide them to a decision

The content plan was deliberately paused. The project record states it plainly: there is no content plan, because we’re doing reoptimization right now.

A site in trouble doesn’t need new articles. It needs the existing ones to start working.

E — Expand the coverage, and cut it

Here step E worked in reverse, and that’s the most valuable part of the case.

A backlink audit (May 2023) — the first systematic step after the decline. The order matters: establish the cause, then act.

Duplicates (January 2024): found, but we didn’t rush to rewrite everything. We rewrote two articles and watched what happened. Test the hypothesis small.

⭐ And we cut a channel we were being paid for. In February 2024, in a Project Overview document, we argued the case for stopping Web 2.0 link building.

It was a live, billed workstream. We ended it, because in this site’s situation we believed it was doing more harm than good.

Expanding coverage isn’t always about adding. Sometimes the most useful thing to do with a limited resource is remove what’s eating it without paying back.

Documents follow hypothesis → test → conclusion

You can see it in the internal correspondence: the lead insists on that structure and on removing phrasing like “if this is indeed the case.”

An analytical document either proves a statement or doesn’t contain it. Assumptions dressed as facts don’t pass our internal review.

The project’s technical conditions

• Two separate admin panels for publishing — one for the blog, one for service pages;
• three staging environments, with a note in the project record that “it’s important to confirm exactly where implementations should be checked.”

A detail that costs a week if nobody writes it down at the start.

The client gave us access to their leads

This is the only project in our practice where the client shared their inbound lead table from Make.com. We could see not just positions but what came out of them.

From the bottom to the end of the work: 2.75× across two years.

Caption:

Per Ahrefs. Values read off the chart; traffic in Ahrefs is an estimate from its own index.

Inquiries. The client kept their own inbound table. Across the final nine months of our work the flow grew: Q3 2025 produced 160% more contacts than Q1.

The useful part isn’t the count — it’s the subject matter. The client replies to each inquiry with a matching case deck, which gives a clean read on what people come for:

Four core verticals account for two thirds of all inquiries. The channel wasn’t bringing in random visitors — it brought the people the content was built for.

Market: 88% of traffic from the US, then Poland, India, Germany.

Recovery runs slower than the fall

The site lost 80% of its traffic in about eight months. Getting 2.75× back from the bottom took twenty-four.

That ratio is normal, and it’s better known in advance. A fall happens in one update. Recovery is years of reoptimization, cleanup, and waiting for the updates after that.

It didn’t fully return to the peak

By September 2025 traffic was around 3,300 against a peak of ~6,000. We don’t claim we got it all back, because we didn’t.

Traffic fell again after the work ended

As of September 2026 the site shows 204 keywords and 859 visits. We stopped working on it in September 2025.

We include this for two reasons. The chart is public. And it’s the best argument available that on projects like this, the result holds exactly as long as the work does.

Directions came from the screen, not from instinct. Ten target markets went through a reachability assessment; four core verticals now produce two thirds of the inquiries. In 2021 the same process concluded “leave it alone.”

Hypotheses get tested small. Two rewritten articles instead of a mass cleanup, then scale on the result.

We stopped what was hurting. Web 2.0 link building was cut from the scope despite being billable.

Not every dip is a problem. In 2021 the analysis showed seasonality plus clicks lost to a Google answer box while positions held. The correct move was to do nothing.

Test a hypothesis on two pages, not two hundred. Cheaper, faster, and if the theory is wrong you haven’t broken the site.

A bad channel should be stopped even when it’s in the contract. Web 2.0 was cut here on purpose.

A site in trouble doesn’t need new content. It needs the existing content to work. The content plan was deliberately paused.

Recovery takes about three times as long as the decline. Eight months down, twenty-four up — and even then, not all the way.

Work concluded in September 2025. Later changes to the site were made without our involvement.

Lost traffic and don’t know why?

It starts with an audit, not with a plan of works. Sometimes the audit says nothing needs doing — we’ve written that one too.

Питання