Iris sees what happens after the click.
Ad platforms count clicks and cost. Iris adds what visitors do on your own site, and turns both into one decision per campaign.
Every visit gets a score
A model trained on your site scores each visit from 0 to 100, by how the visitor behaves. Visitors who act like past buyers score high.
High intent and add-to-cart overlap by only 20 to 30%. The score finds buyers the cart never shows.
One client's site. Values rounded.
- Score 0–9: 0.1% of visits end in an order
- Score 10–19: 0.3% of visits end in an order
- Score 20–29: 0.6% of visits end in an order
- Score 30–39: 1.0% of visits end in an order
- Score 40–49: 1.6% of visits end in an order
- Score 50–59: 2.2% of visits end in an order
- Score 60–69: 3.1% of visits end in an order
- Score 70–79: 4.8% of visits end in an order
- Score 80–89: 8.5% of visits end in an order
- Score 90–100: 24% of visits end in an order
Each visit belongs to the campaign that paid for it
Iris reads the campaign id in the landing link, from Google or Meta. So each visit, high intent visit and order from your ads is tied to its campaign.
One way of counting, for every channel
Google and Meta each count orders their own way, and both can claim the same order. Iris counts each order once, on your own site.
One number decides: eCPO or eROAS
Iris judges every campaign on one number. Read it as a cost per order or as a return on ad spend, whichever your business already measures.
So a new or small campaign can be judged weeks before it has enough orders. You find the winners and the losers sooner.
- Expected cost per ordereCPO
- What a campaign pays for one order.
- Expected return on ad spendeROAS
- What a campaign brings back for each krona it costs.
- A campaign with many orders is judged on those orders.
- A campaign with few orders is judged partly on its high intent visits, valued at the rate your account turns them into orders.
- Each campaign is compared with similar campaigns: same platform, same channel, same kind.
Written rules turn the numbers into a decision
The rules are written down, so every decision can be traced. Our customer success team sets them together with you.
- StopIts visitors are ready to buy far less often than the account's, over the whole period and the last 3 days.
- LowerIts intent is weak and the trend is falling, or it pays more per order than similar campaigns.
- RaiseStrong intent, a fair price, and the budget runs out before the day ends.
- Leave as isNothing moved enough, or the fix is the bid, not the budget.
Brakes
These levels are an example. Yours are set for your account.
- The rules only cut a campaign that spent at least 5 000 kr in the period.
- A campaign with over 10% of the account's spend is lowered, never stopped outright.
- A price that rests on fewer than 5 expected orders decides nothing.
- No raise or cut is bigger than half the budget.
How sure Iris is
- High 21 days of data or more.
- Medium 7 to 20 days.
- Early signal Under 7 days or 100 visits. Iris watches, and moves no money yet.
It tells you how much data each decision rests on.
An AI council makes the final decision
The rules give every campaign a first decision. Then four AI models read the whole account and your instructions, and make the final one.
- Growth
- Looks for campaigns that deserve more money.
- Guardian
- Looks for money that is wasted.
- Auditor
- Questions every decision the numbers do not support.
- The fourth model
- Weighs the three answers and picks one per campaign.
Not tied to one AI model
The council works with GPT from OpenAI and Claude from Anthropic. Each role can use a different model, and we can switch when a better one arrives.
Checked by code
Code checks every number the council writes against your measured data. A decision that fails is thrown out, and the rules' decision is used instead.
Iris follows up on its own advice
After Iris advises a budget change, it checks what happened next, whether anyone acted or not. We also test the rules on past data, on weeks they had not seen.
Why this gives you an edge.
It can tell you to spend less
Iris is not paid by the platforms, and its data does not come from them. So it can say a campaign is not worth its money.
Your data, not theirs
Your high intent visitors are measured on your own site. Competitors bidding on the same searches cannot buy that.
Faster answers on small campaigns
High intent visits give a campaign a price long before it has many orders. Test a new channel without waiting months.
No black box
Every decision names the number that decided it and what it was compared with. You can check the math yourself.
See Iris on your own numbers.
Book a meeting. We show you Iris on a demo account. Then we measure your potential and prove the value in your own accounts, before you commit.
Already using Iris? Sign in