On the moab agency blog on vc.ru, Sergey Dmitrakov published a breakdown of the three-year growth of Shoppers.media—the retail chain Magnit's industry media outlet covering trade, logistics, agribusiness, and marketplaces. At the start of 2024, the site drew around 34,000 search visits per month and was hitting the ceiling of an outdated engine, while ready-made technical recommendations from SEO specialists had sat in a queue for 15 months waiting for a platform rebuild.
What hooked me in this case was the management underside, which the authors aptly called orchestration. Suppose you ordered a deep SEO audit of your site, and then the tug-of-war begins: external consultants ask for dozens of articles targeting broad queries, developers are busy with the core business's servers, and the editor-in-chief protects the publication's voice. A spreadsheet of issues alone will not bring visitors—how, then, do you wire an external SEO team, an IT contractor, and an editorial desk into a single mechanism so technical fixes deliver explosive organic growth?
A Conductor for Three Shops: How Roles Were Split Between moab, Agima, and Editorial
Rebuilding a live publication with multiple contractors without a shared control center is like starting renovations in an operating cafe where you called in electricians, kitchen installers, and a head chef, but forgot to appoint a foreman. Everyone is busy, yet you cannot open the dining room: the outlets were wired nowhere near the stove. At Magnit, the client's product team—Yulia Alemanova and Eduard Utukin—stepped into that conductor role. The client immediately rejected an artificial race for records in the first three months, understanding that an industry media outlet plays the long game for reach and market trust, and clearly divided areas of responsibility:
- Agency moab handled search strategy, technical audits, and evergreen articles—detailed breakdowns on «Честный знак» labeling, certification, and legal giveaway rules that bring readers for years.
- Integrator Agima took over infrastructure migration, backend speedups, and shipping fixes into code.
- The Shoppers.media editorial desk ran the daily news stream and maintained a strict topical filter.
At the intersection of search and editorial policy, a healthy debate broke out right away. The SEOs brought batches of 30 high-demand topics, while the editor-in-chief approved only about 10—strictly about retail, agriculture, and freight, cutting out generic business queries. In my view, it is precisely in these debates that many companies slip up chasing "empty" pageviews, whereas here the filter preserved the core audience: bounce rate holds around 20%, and reading depth grew from 1.2 to 1.58 pages per visit.
Getting in sync also exposed a weak link—a long publishing queue for finished copy. At first, submitted articles waited months for layout: for instance, a step-by-step guide on ad labeling sat in the backlog so long that laws changed and the piece became outdated before publication. That incident forced them to speed up releases, and over a hundred published evergreen articles by the agency brought more than 62,000 search visits and steadily deliver 3,000–7,000 clicks a month regardless of the news cycle.
A 49-Task Backlog and Breaking Into Google Discover: How They Built the Technical Launchpad
The accumulated library of expert articles acted like water behind a closed dam: for the flow to surge at full force, the site's sluice gates themselves needed upgrading. moab packaged the audit findings into a three-tier priority table, writing out a rationale and clear acceptance criteria for every item so developers never had to guess what marketers wanted.
In spring 2025, the Agima team migrated the site into a closed Yandex Cloud perimeter and spun up an exact replica of production—a staging environment for paired acceptance: a developer ships a task to staging, an SEO checks it by hand, and only after fixes does the code go to readers. Along the way, they cleaned up the server: database queries per page dropped from 72 to 13, backend response accelerated from 330 to 221 milliseconds, and out of 241 gigabytes of images they cleared 129 gigabytes of duplicates and drafts, compressing the archive to 14 gigabytes.
Over several months, Agima shipped 44 of the 49 backlog tasks, removing three main roadblocks to search growth:
- Blind spots and server-side rendering (SSR). Previously, the "Blitz News" and "Read Also" blocks were rendered by a browser script without an HTML
<a>link tag. A reader clicked fine, but a search crawler saw emptiness in the source code and did not crawl deeper into the archive. Rendering pages on the server made internal linking transparent to bots. - Duplicate cleanup and a news passport. Configuring
canonicaltags, the sitemap,H2–H6subheading hierarchy, breadcrumbs, andNewsArticlestructured data eliminated clone-page confusion for search engines and gave algorithms a clear article card. - A pass into Google Discover. The SEOs bet early on the recommendation feed on Android smartphones and the Google app, which surfaces fresh articles by user interest without a search query. I checked these requirements against Google Search Central documentation: catchy headlines alone are not enough for the showcase—the code requires the
<meta name="robots" content="max-image-preview:large">directive, clean 16:9 covers at least 1200 pixels wide without logos, and complete author pages with photos and expertise bios.
In June, the upgraded engine connected with the backlog of articles, and traffic jumped nearly fourfold in a single month—from May's 56,000 to a peak of 210,000 visits, while the Discover section in Google Search Console logged 816,000 total clicks across 12.6 million impressions. In July, the figure settled to a natural 75,000 visits: the feed lives on breaking events with a short cycle. Whenever the newsroom quickly published a breakdown of a hot story like the fire at the «Садовод» and «Белая Дача» markets, the algorithm immediately delivered a wave of traffic, whereas any delay in publishing hands the airtime to competitors.
A New Year's Knockout from Their Own Defenses: Why Google Traffic Crashed from 105,000 to 12,000 Visits
No sooner had marketing, editorial, and development hit their stride than a tripwire came from a fourth perimeter—information security. Right before the holidays, on December 30, queries from Google's top ten suddenly vanished from search results, and access to Search Console disappeared.
The symptoms were immediately escalated to Magnit's team, and January diagnostics revealed an ironic picture. During a routine tightening of corporate Anti-DDoS protection (an automated shield against junk traffic attacks), Googlebot search crawlers landed on the blocklist. It is as if a supermarket guard were told not to let rowdy crowds in before the holidays, and out of sheer zeal locked the doors on the store's own cash-in-transit couriers. The system dutifully blocked frequent page crawling as suspicious activity.
To me, this incident is the case's most valuable lesson, because in business security teams and marketers rarely sync watches before the first crash. Google visits collapsed from December's 105,000 to 12,000 in January 2026. The block was lifted immediately, yet organic search is not paid ads where impressions return as soon as you top up the balance: the crawler has to re-index thousands of pages and recalculate weights. By May, overall search traffic had recovered to 63,000 visits per month—nearly double the project's starting point—and continues to grow.
A Resilience Protocol and AI Search Readiness: Protecting Organic Traffic Across Teams
The New Year's outage hardened the system: Magnit's product team pushed a mandatory rule through the IT directorate—any changes on servers, the CDN (content delivery nodes that speed up page loads), and security systems are pre-checked for SEO compatibility. In a large enterprise, getting such a guardrail approved is harder than writing code, yet it is what insures traffic against accidental "friendly fire."
With their flanks covered, the teams audited visibility in answers from ChatGPT, YandexGPT, Gemini, DeepSeek, and AI search blocks: out of 65 queries, the publication was cited in 27, and in narrow topics like coffee vending machines it was recommended by 3 out of 6 AI engines. Now Agima is rolling out an updated robots.txt, Organization schema markup, Q&A blocks, and an answer-first layout with a direct answer in the opening paragraph, while the editorial desk has practical search workshops ahead.
If you want to dig deeper into the economics of these channels, check out our pieces on why live expert texts bring 2.5x more traffic over time than AI-generated streams and how generative search prepares a client for a deal before they even click through to your site. For your own project, two simple rules are worth taking from Magnit's experience: maintain a single prioritized backlog with acceptance criteria on a staging server, and ban any server updates without a five-minute check-in with whoever owns search.
And if tonight your sysadmin or hosting contractor decides "just in case" to tighten bot protection ahead of the weekend—will your marketer find out before traffic stats collapse on Monday?

