Target ROAS for a new Shopping account: set it from margin or from history?

Setting a tROAS on a new Shopping account looks simple, but where the number comes from matters more than the number itself. Most people work backwards from margin. A 30% gross margin means you need roughly 333% ROAS just to break even, and then they wonder why delivery falls off a cliff. Setting it from last month's actuals is gentler, but it locks in whatever the account was already doing, waste included. I'd start 20 to 30% below break-even and only step up once the campaign has a steady run of conversions, so the algorithm has room to find volume. The catch is that a brand-new account has no actuals at all, so how do you pick a starting point there without just guessing?

Agreed on the margin math, but on a brand-new account I'd skip tROAS at first and run Maximize Conversion Value for two or three weeks. Once you've got around 30 conversions in that window, set the target at your trailing actual ROAS minus 15 to 20%, so the starting point comes from your own data instead of a guess.

Mostly agree, but the 30 conversion rule really depends on how many orders you get per week. At 15 a week you're there in two or three weeks and the approach works fine. At 3 a week you'd sit in Maximize Conversion Value for two months, and by then seasonality or a price change may have made that data stale. In a low-volume store I'd take the margin-based number, cut it by the same 20%, and accept that the first few weeks will be noisier.

Agreed, and at low volume the conversion lag matters too. If people tend to buy 10 days after they click, a three-week window at 3 orders a week only holds a handful of finished conversions, so the trailing ROAS you're reading is partly guesswork as well.

Yeah, the lag is the real issue, and the fix is to read the window by conversion date instead of click date. In Google Ads the conversions column is tied to the click day by default, so a recent week looks worse than it really is because purchases from the last few days haven't landed yet. Pull trailing ROAS on a conversion-date basis, or just wait one full lag cycle before you trust it, and you'll start from a much cleaner number.

I'm not convinced the Conversions column is click-date by default, so check your segment and column settings in your own account before you change any reports. And even if you switch to a conversion-date view, the recent days don't get fixed. The incomplete period just moves to the right edge of the chart. The last lag cycle is still filling in on either basis, so trailing ROAS will look a bit low on the newest days no matter what. Waiting out the lag is what actually cleans up the number. At 3 orders a week that wait matters for the 30-conversion threshold too, because the conversions you're counting need time to finish before they tell you anything about ROAS.

Fair, I overstated it. Switching to conversion date only pushes the incomplete days to the edge of the chart, so the lag is still there. How much it matters depends on your typical click-to-purchase gap. If most orders close within a day or two, the newest days barely distort trailing ROAS. If it's closer to 10 days, you just have to wait it out.

Find the lag where about 90% of your conversions have landed, and don't trust trailing ROAS for any day newer than that. The conversion lag bucket segment in Google Ads shows you this. If most orders close within two days, the last week is readable. If 90% only closes around day 10, a three-week window is really about two weeks of finished data, and only that finished data should count toward your 30 conversions.

I prefer this over a fixed "wait a week" because the right wait depends on the store. A fixed cutoff is too long for a fast-closing account and too short for a slow one. The account's own lag distribution tells you how much of the recent data is still pending, so both the ROAS reading and the conversion threshold stay honest.

The 90% cutoff holds up at high volume, but at 3 orders a week the lag distribution is too thin to trust. A three-week window might hold 10 to 15 purchases, and the 90th percentile of that is mostly noise. Two late orders can push your "safe" cutoff from day 2 to day 9.

For a low-volume store, pull the lag bucket over 90 days instead. That gives you enough orders to actually read the curve, and you can use that single cutoff for your ROAS readings. Re-check it every quarter, since promos and seasonality change how long people take to buy.

If 90 days still doesn't give you a clean curve, the trailing ROAS isn't reliable enough to set a target from, and the margin-based starting point is the safer call.

Agreed on the 90 day pull, but be careful what you read off it. At 3 orders a week, one big basket can be a large chunk of the window's revenue, so the trailing ROAS is mostly just that one order. Calculate it as total conversion value divided by total spend over the whole window, not as an average of daily ratios. Then see how much of the revenue comes from the top two or three orders. If one order is a third of it, I'd widen the window or lean on the margin-based number instead.

How spread out are your order values, or do a few baskets dominate?