{"id":4970,"date":"2025-10-26T10:41:02","date_gmt":"2025-10-26T09:41:02","guid":{"rendered":"https:\/\/www.htt.it\/?p=4970"},"modified":"2026-02-08T16:56:28","modified_gmt":"2026-02-08T15:56:28","slug":"andromeda-metas-new-update","status":"publish","type":"post","link":"https:\/\/www.htt.it\/en\/andromeda-metas-new-update\/","title":{"rendered":"Andromeda: Meta’s new update"},"content":{"rendered":"\n\n\n
\n
\n
\n
\n
\n

Hello Andromeda, we\u2019re really curious to get to know you.<\/h2>\n<\/header>\n

No, this isn\u2019t a new colleague who just joined HT&T, nor Perseus\u2019s mythical wife, nor the spiral galaxy closest to Earth.
\nAndromeda is a key piece in Meta\u2019s complete rebuild of its advertising infrastructure, the most significant AI-based update since the iOS 14 crisis.
\n\"Andromeda,We are therefore facing a watershed moment for anyone investing in advertising campaigns on Meta (Facebook and Instagram).<\/p>\n

Origins and functions<\/h2>\n

The core function of this new AI system is to decide in real time which ads to show and to which users. Its gradual rollout, which began in late 2024 and was completed over the last few months of 2025, had a clear goal: to make campaign management even more automated, intelligent, and predictable.<\/p>\n

The need for this change was driven by an internal operational issue at Meta: the exponential growth of ad content due to the introduction of tools like Advantage+ Creative and GenAI.<\/p>\n

The result was a steep increase in the number of ads in circulation\u2014from millions to hundreds of millions\u2014creating a load that Meta\u2019s old infrastructure could no longer sustain.
\nAndromeda can therefore operate in the most critical stage of ad delivery (Retrieval<\/em>) using deep neural networks.
\nThis new AI engine can process between 100 and 1,000 times more ads than before, learning complex patterns and personalizing delivery at a hyper-granular level.<\/p>\n

In practice, the system can show the right content to the right person.<\/p>\n

\"campaigns<\/p>\n

\n

A practical example:<\/h2>\n

Let\u2019s take Marco, who is considering buying a V-neck cashmere cardigan.
\nWhile browsing Instagram, he interacts with an ad from Brand A. This first click is the signal that activates the algorithm and triggers:<\/p>\n

    \n
  • Brand A\u2019s action:<\/strong> Brand A\u2019s algorithm, optimized for Andromeda, won\u2019t just show him the same ad repeatedly. It will orchestrate a varied path: perhaps a Story highlighting the craftsmanship of the cashmere, then a carousel with different styling combinations.<\/li>\n
  • Meta\u2019s ecosystem action:<\/strong> Meta has now labeled Marco as an in-market<\/em> user for \u201cluxury cashmere apparel.\u201d As a result, Brands B and C\u2014set up with similar targeting\u2014will also compete for his attention.<\/li>\n<\/ul>\n

    The outcome is that Marco\u2019s feed turns into a truly competitive space. He won\u2019t just see ads from Brand A, but a strategic mix of content from multiple companies.
    \nIn this scenario, the system hasn\u2019t just identified Marco as the ideal buyer persona; it has created a discovery experience where several brands compete for his preference. The winner won\u2019t be the brand that shows the most impressions, but the one that\u2014thanks to a richer, more relevant creative ecosystem\u2014tells the best story and convinces him along his purchase journey.<\/p>\n

    The 3 Pillars of Change: Targeting, Creativity, and Structure<\/h2>\n

    So, what are the main consequences for advertisers?<\/p>\n

    The end of manual targeting<\/h3>\n

    Before Andromeda, best practices for concrete results relied on precise, manual audience segmentation (interests, specific demographics), which has now been largely surpassed by Meta\u2019s AI capabilities.
    \nThe winning strategy has become broad targeting, allowing the algorithm to find the user segments most likely to convert. Over-restricting the audience worsens performance because it limits the system\u2019s learning capacity.<\/p>\n

    \"How<\/p>\n

    <\/h3>\n

    Creativity is the new targeting<\/h3>\n

    If the past rewarded the best targeting, now the advantage goes to those who produce the greatest creative diversity: the best practice is to diversify creatives as much as possible by creating multiple ads within the same campaign. The algorithm is no longer looking for a single \u201cperfect ad\u201d; the goal is to match each creative variation (videos, carousels, copy addressing different pain points) to its specific micro-segment. The more creatives you have, the more usable signals the AI gets.<\/p>\n

    Simplified campaigns<\/h3>\n

    Complex and fragmented structures, with numerous ad sets to manually control variables, are now counterproductive. The new best practice calls for consolidated structures: a single campaign per objective with Campaign Budget Optimization (CBO) enabled, and dozens of different creatives inside it. This simplicity allows Andromeda to learn and optimize the budget more effectively.
    \n<\/section>\n

    \n

    The crucial role of marketing consultants<\/h2>\n

    Andromeda therefore presents itself as a significant\u2014and potentially very profitable\u2014automation, with estimated ROAS uplifts of +22% when using optimized creatives.<\/p>\n

    However, the price to pay is the loss of the granular control we had in the past when everything was managed manually: the system manages budget and creative distribution, and its decisions are often hard to interpret.<\/p>\n

    A \u201cblack box\u201d is emerging\u2014one that requires human expertise to be governed effectively. Meta\u2019s system won\u2019t take advantage of advertisers\u2019 budgets only if a team of experienced marketing consultants, like us at HT&T, supervises and guides the AI.<\/p>\n

    Here\u2019s why:<\/p>\n

      \n
    • Strategic setup:<\/strong> AI optimizes, but it doesn\u2019t set strategy. Experts must define the true objective (leads, sales, etc.), identify the ideal customer, and build a robust funnel to convert the generated traffic. If the objective is wrong or the landing page isn\u2019t optimized, Andromeda will waste budget optimizing for weak signals.<\/li>\n
    • Budget management:<\/strong> The system may suggest increasing budget too quickly or focusing on low-yield tests to exhaust spend. Experts know how to read real data (CPL, cost per purchase, ROAS\u2026) and can set the right amounts for the objectives without blindly trusting automation.<\/li>\n
    • Creative architecture:<\/strong> We\u2019re not just marketing consultants\u2014we\u2019re also \u201ccampaign architects\u201d and strategic content producers. We must ensure Andromeda gets the volume and variety of creatives (15\u201320 assets with constant refresh) it needs to reach its full potential.<\/li>\n<\/ul>\n

      In conclusion: well done, Andromeda!<\/h2>\n

      But let\u2019s never forget the importance of carefully planned strategies, targeted creatives, and strong data analysis skills.<\/p>\n

      Without expert guidance, artificial intelligence can scatter budget across a sea of ineffective attempts.<\/p>\n

      Only with the right expertise can this system become a genuine competitive advantage.<\/p>\n<\/section>\n

      \n

      Frequently asked questions about Andromeda (Meta)<\/h2>\n
      \nWhat is Meta\u2019s Andromeda?<\/summary>\n
      \n

      It is the new retrieval<\/em> engine for ads: it filters millions of ads in real time and selects the most relevant candidates before ranking\/auction, improving relevance and performance.<\/p>\n<\/div>\n<\/details>\n

      \nWhen was it released and how widespread is it?<\/summary>\n
      \n

      Meta described Andromeda in engineering channels in late 2024; in 2025 the update was widely rolled out and now affects ad delivery on Facebook and Instagram.<\/p>\n<\/div>\n<\/details>\n

      \nWhat changes for targeting and manual segmentations?<\/summary>\n
      \n

      The weight of manual targeting decreases while automation (e.g., Advantage+) increases. Broad audiences with strategic exclusions work better than rigid micro-segmentations.<\/p>\n<\/div>\n<\/details>\n

      \nHow does it impact campaign structure?<\/summary>\n
      \n

      Less fragmentation, more consolidation: fewer campaigns\/ad sets, pooled budgets, many creatives under test. The algorithm optimizes distribution across formats and placements.<\/p>\n<\/div>\n<\/details>\n

      \nWhat is the role of creatives?<\/summary>\n
      \n

      Creative diversity becomes central: more variants (formats, angles, hooks) help Andromeda find the right match. In practice, \u201ccreativity is the new targeting.\u201d<\/p>\n<\/div>\n<\/details>\n

      \nDo I need to change anything in tracking?<\/summary>\n
      \n

      Yes: prioritize clean, complete signals (CAPI\/server-side, deduplication, event quality). Well-configured objectives and conversions help the algorithm optimize correctly.<\/p>\n<\/div>\n<\/details>\n

      \nWhich metrics should I track to assess impact?<\/summary>\n
      \n

      Beyond CPA\/ROAS, look at creative distribution, reach\/frequency, learning phase stability, downstream traffic quality (onsite engagement, AOV, conversion rate), and delivery stability.<\/p>\n<\/div>\n<\/details>\n

      \nImmediate best practices?<\/summary>\n
      \n

      Consolidate budgets, use broad audiences with exclusions, increase creative variants, keep conversion\/CAPI setup clean, and run cyclical tests without frequently interrupting learning.<\/p>\n<\/div>\n<\/details>\n<\/section>\n<\/article>\n