If you sell sofas, beds or dining sets online, the default 7/14/30-day retargeting windows in Meta and Google are working against you. A ₹60,000 sofa gets browsed, measured against a wall, argued about with a spouse, cross-shopped against a local carpenter's quote, then bought six or seven weeks later. Retargeting for furniture ecommerce has to survive that gap without burning the same three creatives into someone's feed 50 times. The fix is structural: three audience windows (0–3 days, 4–21 days, 22–60 days), different creative in each, and a frequency budget calculated from pool size rather than whatever number the media plan says.
Most furniture accounts we inherit have the opposite setup. One catalogue ad set, all site visitors, 30-day window, spend cranked up because ROAS looks good on last-click. Frequency sits at 14. The 45-day browsers, the ones actually deciding, are excluded entirely.
The arithmetic that caps your retargeting budget
Retargeting spend has a mathematical ceiling and almost nobody calculates it. Work through a real case.
Say the store gets 18,000 unique product viewers over a rolling 60 days. Meta CPMs for a furniture audience in India, warm traffic, land somewhere around ₹200. The plan says ₹2,00,000 a month on retargeting.
₹2,00,000 at a ₹200 CPM buys 10,00,000 impressions. Spread across 18,000 people that is 55 impressions per person per month, about 13 a week. Nobody needs to see your dining table 13 times a week. Past roughly four or five, the marginal impression stops persuading and starts irritating, and your negative feedback rate climbs, which pushes CPMs up, which buys you fewer impressions for the same money. The plan defeats itself.
Now run it the other way. Split the pool:
- Hot, 0–3 days, high-intent events only: 1,200 people at 24 impressions/month = 28,800 impressions = ₹5,760
- Mid, 4–21 days: 5,000 people at 12/month = 60,000 impressions = ₹12,000
- Long, 22–60 days: 11,800 people at 6/month = 70,800 impressions = ₹14,160
Total: about ₹32,000 a month to reach every warm visitor at a frequency they can tolerate. That is the honest number. The other ₹1,68,000 belongs in prospecting, or in a fourth window, or nowhere. We have had this conversation with founders more than once and it never lands well the first time. It is still true.
Three windows, three different arguments
The reason default windows fail is that a person three days into a sofa decision and a person forty days in have nothing in common. Same product, entirely different objection.
0–3 days. They remember the product. Show it. Dynamic catalogue ads earn their keep here and only here. Price, fabric, one line about delivery time. No brand film.
4–21 days. This is the measurement and comparison phase. The objection is not price, it is fit and risk. Ads that work: dimension overlays in a real room, fabric swatch delivery, the assembly process, warranty length, what happens if it does not fit through the stairwell. We have seen a plain 20-second video of two people carrying a sofa up a narrow flat's staircase outperform three polished lifestyle cuts, because it answers the question the customer is actually stuck on.
22–60 days. Trust and cost-spreading. Customer photos in homes that look like the buyer's home, not a studio. EMI on the exact SKU with the monthly figure spelled out. Store visit invitation if you have a showroom. Anything that de-risks a ₹45,000 transfer to a brand they have never touched.
Run these as separate ad sets with hard exclusions between them, otherwise the 0–3 day creative keeps chasing someone into week six and you have rebuilt the problem.
Frequency caps that survive contact with the auction
Meta will not give you a true frequency cap in standard auction buying. Reach and frequency buying does, but the minimum reach requirements make it impractical for a 12,000-person warm pool. So you control frequency with the three levers you actually have: audience size, daily budget, and mutual exclusion between ad sets.
Practical caps we work to on furniture accounts: 6–8 impressions per week on the 0–3 day pool, 3–4 on the mid window, 2 on the 22–60 day window. Check the delivered number weekly in Ads Manager with a 7-day window, not lifetime, because lifetime frequency on a long-running ad set is a meaningless average.
Google Display and Demand Gen do let you set a real per-user cap. Use it. Two per day, ten per week on Display for the long window is generous. YouTube is the one place a longer, slower asset justifies itself in the 22–60 day bucket, because a two-minute factory or materials video is watched by people who are close to buying and want a reason to trust you.
Stop building audiences on page views
A product page view is a weak signal for furniture. People browse sofas the way they browse property listings. The signals worth building audiences on are the ones that cost the visitor effort:
- Variant or swatch selection
- Opening the dimensions or materials tab
- Pincode or delivery-estimate check
- Using more than two filters in a collection, or repeating a site search
- Add to cart, obviously, and swatch sample requests if you offer them
- Return visit within 7 days to the same product
These need custom pixel events. It is an hour of theme work and it changes the quality of every audience downstream. Filter and search behaviour is the most under-used of the lot: someone who filters to