Keyword Architecture for Hebrew and English Search Accounts

In Hebrew, exact match leaks in both directions, and that single fact reorganises the account. Close-variant matching was built around English morphology and handles Hebrew badly: no capitalisation to separate a brand from an ordinary noun, two accepted spellings of the same word under ktiv male and ktiv haser, the particles ש, ה, ו, ב, ל, מ and כ fused to the front of the word, and imported brand names that customers transliterate three or four different ways. A Hebrew exact-match keyword misses variants its English equivalent would catch and admits others it would not. Any structure that assumes match type equals control is wrong before it launches. What you receive: A map of the account, not a list of keywords. Which intents get their own campaign because their conversion rate and their margin genuinely differ. Which get folded together because splitting them starves the bidder of signal. Where the budget line sits between them and why. For a food and beverage operator that usually resolves into delivery intent, collection intent, brand, and the category tail behaving as four businesses with four sets of economics. For apparel it is full-price intent against sale-seeking intent, which convert at similar rates and contribute nothing alike. For local services it is emergency against planned, where the same words carry different urgency and a different tolerance for price. Match type as routing, not as control: Match types are assigned to decide where traffic lands once it has already arrived, with negatives doing the routing between campaigns, and the arrangement written down so whoever inherits it can maintain it. Broad match gets used where the value signal is strong enough to steer it and refused where it is not. That is a measurement judgement about your account, not a position I hold in general. The Israeli calendar is not the American one: The working week runs Sunday to Thursday, so ad schedules copied from a US template are wrong from the first day. Demand around the chagim moves in weeks rather than days: the run-up to Rosh Hashana and to Pesach reshapes food and apparel search for a fortnight either side, and August behaves nothing like the August in your benchmark deck. Bilingual accounts want Hebrew and English separated for budget and reporting even when the landing pages are shared, because they behave differently enough that a blended number conceals both. What I refuse to deliver: Single-keyword ad groups. Campaigns duplicated across match types in the old alpha-beta arrangement, which was a workaround for a matching system that no longer exists. A spreadsheet of thousands of terms sorted by search volume, which invites you to judge the work by its row count. Keyword work in languages I do not read — I work in Hebrew and English, and for Arabic, Russian or French markets I will introduce you to someone rather than run your money through a translation tool and hope. Who this is wrong for: Buyers who want the map so a cheaper pair of hands can build it. The structure survives contact with reality because of the argument behind each split, and the argument does not travel in a file. Accounts where the real constraint is the landing page or the offer — you will hear that, and the keyword work will not fix it. How it runs: Two weeks of demand mapping against your own search terms, your sales data and the Hebrew and English query space. Then one long session arguing the structure with whoever owns the budget, because that is where the splits are decided rather than confirmed. Then the written map with a build order and the reasoning attached to each decision. Building it is the next tier, and plenty of clients take the map and build it themselves.