Should dimension queries go into Search or Shopping?

Both, but for different reasons. Search lets you write the exact size into the headline, which is the strongest relevance signal you have for a shopper who has measured. Shopping and Performance Max can only match a size if it exists in the product data, so treat the feed as the prerequisite and Search as the place you control the message.

How many ad groups is too many for a furniture account?

The limit is not a number, it is landing pages. Every additional ad group needs a headline that changes and a page that proves the promise. If you cannot land a 3 seater sofa query on a page filtered to 3 seater sofas, the extra ad group adds reporting fragmentation without adding relevance.

Can I use a free delivery callout if it only applies to some cities?

Only if the campaign is geographically scoped to those cities, or the callout itself names the limit. An asset that promises something the checkout then withdraws costs you the click and creates a mismatch between the ad and the landing page, which is a policy exposure as well as a conversion problem.

Why does my ad group ROAS swing so much between months?

Because one order moves it. In the worked example above, removing a single 45,000 TL order from a month with 20 orders takes reported ROAS from 6 to 5.25 with no change to the campaign. At two or three orders per ad group per month, month to month movement is mostly sample size, which is why targets should be reviewed on a longer window.

Do I need offline conversion import if most orders complete online?

If a meaningful share of high value orders close by phone or in a showroom, yes. A bid strategy optimises towards the conversions it can see, so unimported offline sales quietly teach it that your most valuable traffic is worthless. Import them with the order value attached rather than as a flat count.

Google Ads for Furniture Retailers

Google Ads for furniture retailers turns on two questions a shopper asks before they ever click: will it fit, and when will it arrive. Search demand here is unusually literal about both: people type dimensions, seat counts and room names straight into the box, and the account that answers those exact words in the headline takes the click from one answering with a generic category page. This article covers three decisions specific to the category: mapping room and dimension queries to ad groups, putting the delivery and assembly promise into your assets, and choosing a bid strategy when orders are few and each one is large.

None of it is a creative problem. It is structural, and it hides well: an account wasting half its budget on searches it cannot serve still reports a respectable ROAS on the orders it gets.

What makes Google Ads for furniture retailers different

Two facts about the category drive every decision below.

The query is a constraint, not just a product name. Someone searching for a 160x200 bed has already measured the room. The search is a filter, and the ad has to show the filter is satisfied.
Orders are rare and expensive. A furniture campaign can spend for days between conversions, then land one order worth more than a fast moving store takes in a week. Every automated system in the account has to cope with that shape.

Those two facts pull in opposite directions: the first pushes you to split so each query family gets its own headline, the second punishes splitting because thin data makes automated bidding unreliable. Resolving that tension is the job.

Room queries and dimension queries are different intents

Furniture search separates into families that deserve different treatment.

Room level. Living room furniture, bedroom set, home office desk. Browsing intent, no constraint stated yet.
Dimension level. A 160x200 bed, a 3 seater sofa, a 180 cm dining table. The shopper has measured, and this is the highest intent traffic in the account.
Configuration and fit context. Corner sofa with storage, extendable dining table, sofa for a small living room. A functional or spatial requirement rather than a stated size.

The rule for splitting is narrow: give a family its own ad group when the headline has to change and a landing page proves the claim. A dedicated ad group for 160x200 beds is worth building only if you can land that click on a page already filtered to 160x200. Send it to the generic beds category and the split is cosmetic: you paid for a promise the page does not keep.

Handle the notation before you handle the keyword

Sizes arrive written every way a keyboard allows: 160x200, 160 x 200, 160*200, and the same size in words. Match types treat these inconsistently, so build the ad group around one canonical form, then read the search terms report for two weeks and add the variants that actually turned up. This is also where Search earns its place against Shopping: in a Shopping or Performance Max unit the title comes from the feed, so the size must exist in the product data to match at all. Getting size and variant fields right in the feed is its own project, covered in product feed automation.

Room level queries rarely deserve tightly matched ad groups; they are browse traffic, better served by Shopping inventory and a category page. Keep negatives short and category specific: assembly instructions, how to build, second hand, repair, upholstery service and rental.

Delivery and assembly belong in the assets, not only on the site

The objections that kill a furniture sale come in order: does it fit, what does it cost, when does it arrive, who carries it upstairs, who assembles it, and what happens if I hate it. The ad copy handles two; the rest belong in what Google now calls assets rather than extensions.

Sitelinks with descriptions pointed at real pages: delivery and assembly, size guide, returns for bulky items, showrooms. A shopper who clicks the delivery sitelink is qualifying themselves, which beats a bounce.
Callouts for short absolutes you can back everywhere the campaign runs: assembly included, free delivery, 14 day returns. If a promise applies only to certain cities, scope the campaign geographically or rewrite it per region.
Structured snippets under a header such as Types or Styles, which widens the ad and signals the breadth of the range.
Price assets to pre qualify: in a wide price spread, a starting price loses you the clicks you did not want.
Location assets if you have a showroom. At this basket size, seeing it in person is part of the offer.

Two disciplines matter. Assets are eligible to serve, not guaranteed, so never put a promise in an asset the ad would be incoherent without. And whatever the asset says must appear on the landing page in the same words: a delivery claim the checkout contradicts is a wasted click and a policy risk.

Bidding when orders are rare and baskets are large

Work the arithmetic with your own figures. Take an illustrative account: a campaign spends 60,000 TL in a month at an average cost per click of 12 TL, which buys 60,000 divided by 12 = 5,000 clicks. If your own analytics shows those clicks converting at 0.4%, the campaign produced 5,000 x 0.4% = 20 orders. At an average order value of 18,000 TL that is 20 x 18,000 = 360,000 TL of revenue, and 360,000 divided by 60,000 = a reported ROAS of 6.

Now take one order away. Remove a single 45,000 TL sofa set and revenue falls to 315,000 TL, so ROAS reads 315,000 divided by 60,000 = 5.25. The reported figure moved 12.5% because one household changed its mind, while nothing in the campaign changed at all. That volatility is the defining condition of the category, and it has three consequences.

Ad group structure is not reporting structure. Spread those 20 orders across eight ad groups and the average ad group carries 20 divided by 8 = 2.5 orders a month. Judging an ad group on its ROAS at that volume is reading noise. Ad groups control relevance and headlines; the campaign is the unit where a bid strategy accumulates data. So split rooms and sizes into ad groups inside one campaign rather than separate campaigns wherever budget and geography allow.

Reach the target in stages. Starting straight on Target ROAS anchors the system to a number produced by too few sales. Run Maximize conversion value without a target until you have a trailing period you trust, then derive the target from it rather than from an aspiration; the target ROAS calculator turns your own margin into that figure. After that, move the target in small steps, because a large change sends the campaign back into learning. Google's help documentation states minimum conversion recommendations for value based strategies and revises them, so check the current figure in the official entry rather than trusting a blog post. The trade offs between strategies are covered in our guide to PPC bid management strategies.

Give the model every real sale. Furniture buyers finish by phone, over messaging or in the showroom far more often than most categories, and those sales are invisible to a bid strategy unless you import them. Offline conversion import with the order value attached is worth more here than any bid tweak, because it multiplies the examples the system learns from. Check the click through conversion window too: a research cycle running for weeks against a short window deletes your own training data.

A sensible build order

List your real dimension and configuration query families from the search terms report, not a keyword tool.
Confirm a filtered landing page exists for each family before creating the ad group. No page, no split.
Write delivery, assembly and returns into sitelinks and callouts, then check the same wording on the destination page.
Consolidate campaigns until each carries enough conversions to justify a bid strategy, and recover the granularity in ad groups.
Import phone and showroom orders with values, then review targets monthly rather than weekly. A weekly review at this order frequency is a reaction to one customer.

If you are weighing external automation on top of this, treat it like a bid strategy: decide what it controls before you connect it. ZenoxAds manages Meta and Google Ads campaigns, and the automatic scaling page sets out that scope. Two or three orders per ad group per month rewards structure long before it rewards tooling.