How to Measure Programmatic Advertising ROI: A Practical Guide

Programmatic campaigns produce a lot of numbers, and very few of them are ROI. Clicks, conversions and even ROAS are steps towards the answer. ROI is the question your finance team actually asks: did this campaign make us more money than it cost?

The formula is the easy part. The hard part is agreeing what counts as cost, what counts as a conversion, and how much of the return the ads actually caused.

What is the difference between ROI and ROAS?

ROAS tells you how much revenue was linked to each rupee of ad spend. ROI tells you how much profit the campaign added after all its costs. They answer different questions, and mixing them up is one of the most common reasons a good-looking campaign loses money.

Work out break-even ROAS before launch. If your gross margin is 40%, you need a ROAS of 2.5 just to stand still. A campaign reporting a ROAS of 2 looks healthy on a dashboard and is still losing money.

What costs should go into the calculation?

Everything the campaign needed to run. Using media spend alone flatters the result.

• Media cost, the inventory itself.
• Platform and data fees, if you run your own seat.
• Service or management fees, if a partner runs the campaign.
• Verification and ad serving fees, such as IAS, DoubleVerify or Campaign Manager 360.
• Creative production, including language versions and resizes.
• GST, where you can't claim the input credit.

One trap in DV360 reports: the metric called "Revenue (Advertiser Currency)" is the amount spent buying impressions and serving ads through DV360, not sales revenue. It belongs on the cost side of the sum.

How do you track the return side?

You need each valuable action recorded, ideally with a value attached. In DV360 that usually means Floodlight, the conversion tracking system shared across Google Marketing Platform. Sales activities can pass order value, and counter activities can record leads or sign-ups. Our guide on tracking conversions in DV360 covers the setup step by step.

If your team works mainly in Google Analytics, you can also connect DV360 with GA4 and use GA4 key events in custom bidding.

For lead generation, the tag only gets you halfway. A form submission isn't revenue. Follow leads into your CRM and work out what a lead is worth:

Value per lead = qualified rate x conversion to sale x average gross profit per sale

Here's an illustrative example for an NBFC running a personal loan campaign. Say 30% of leads are qualified, 15% of qualified leads disburse, and the average gross profit per loan is ₹12,000. Value per lead = 0.30 x 0.15 x ₹12,000 = ₹540. If you're paying ₹700 per lead, the campaign loses money however cheap the CPM looked. Our page on NBFC and lending campaigns shows how we tag each step of an application flow for exactly this reason.

Where the sale closes offline, as in real estate, cars or B2B, upload those outcomes to a Floodlight activity as offline conversions, so they sit next to your online data instead of in a separate spreadsheet.

Which attribution rules should you agree before launch?

Decide these before the campaign starts, write them down and don't change them halfway through.

1. Post-click and post-view. A post-click conversion follows a click on your ad. A post-view conversion follows an ad view with no click. DV360 lets each line item count all of both, only post-click, or post-click plus a percentage of post-view that you choose.
2. Lookback windows. These set how many days after a view or click a conversion can still be credited. Match them to how long your customers take to decide. A food delivery order and a two-wheeler purchase need very different windows. If you also use Campaign Manager 360, the shorter of the two windows applies.
3. One source of truth. Floodlight, GA4 and your CRM will never agree exactly. Choose which one decides ROI and use the others as cross-checks. Our guide to CM360 campaign attribution explains how attribution models shift credit between channels.

Report post-click and post-view results in separate columns. Post-view conversions are real signals, but some of those people would have bought anyway. That's where the next step comes in.

How do you prove the sales were extra?

Attribution tells you which ads a buyer saw. It doesn't tell you whether the ads caused the sale. For that you need a comparison group that didn't see the ads.

• Holdout tests. Keep a random slice of your audience out of the campaign and compare conversion rates. In DV360, Google's Conversion Lift studies do this for YouTube and Google Display Ads line items. Only a Google sales representative can set one up, and the campaign needs enough budget and conversions to give a readable result.
• Geo tests. Run the campaign in some cities and hold it back in similar ones. You might advertise in Pune and Ahmedabad and hold back Jaipur and Lucknow, then compare the change in sales in each group over the same weeks.
• Marketing mix modelling (MMM). For brands with a couple of years of weekly sales and spend data, a model such as Google's open-source Meridian estimates each channel's contribution, including TV and offline effects.

The share of attributed sales that turns out to be incremental varies a lot by brand, category and channel. Measure yours rather than borrowing someone else's number.

A worked example: attributed ROI vs incremental ROI

Here's an illustrative month for a D2C skincare brand running programmatic display, video and retargeting. All figures are made up for the example.

On attributed numbers, ROAS is 3.75 on media cost, or 3.0 on total cost. Gross profit is ₹30 lakh x 55% = ₹16.5 lakh, so ROI = (16.5 - 10) / 10 = 65%.

Now suppose a geo holdout shows that 70% of the attributed revenue was truly incremental. Incremental revenue is ₹21 lakh, incremental gross profit is ₹11.55 lakh, and ROI = (11.55 - 10) / 10 = 15.5%.

Both numbers describe the same campaign. The second is the one to plan next quarter's budget on. It's still positive, and the incremental ROAS of 2.1 on total cost clears the break-even ROAS of about 1.8, so the campaign earned its keep. But the margin for error is far thinner than the dashboard suggested: a 10% drop in conversion rate would take ROI down to about 4%.

How do you measure ROI on branding campaigns?

Awareness campaigns rarely show a clean ROI inside the campaign window, and forcing last-click numbers onto them will always make them look bad. Measure them in stages instead.

• Did the right people see it? Look at unique reach and frequency against your target audience, on ads that were actually viewable. Our guide to improving ad viewability covers the second part.
• Did it change minds? A brand lift study gives you the lift in recall or consideration, and a cost per lifted user you can compare across campaigns.
• Did it add new people? Incremental reach shows what digital added on top of TV or other channels.
• Did it move sales over time? MMM or long-running geo tests connect brand spend to sales over quarters, not weeks.

Which mistakes make ROI look better than it is?

• Leaving fees, verification and creative out of the cost.
• Counting every post-view conversion as fully earned.
• Comparing channels measured with different lookback windows.
• Reading DV360's "Revenue" metric as sales.
• Judging a campaign that ran through a sale period against a normal month.
• Ignoring lag. Real estate, cars and B2B often convert weeks after the first ad, so measure across the full buying window.

Next step

If we run your campaigns through our DV360 Managed Service or a Campaign-Based Engagement, our DV360 Reports Section gives you the cost and delivery side: spend, impressions, unique reach, frequency, CPM and video completion. Pair it with your Floodlight or CRM data and the formulas above. To set targets before you spend, ask for a free media plan and work back from your break-even ROAS.