How to Forecast Results From a Creator Marketing Budget: Inside Almedia's Model

Most companies treat creator campaigns as a leap of faith and reconcile the numbers afterwards. Almedia — the company behind Freecash, with more than 70 million users and over $300 million paid out to them — decides whether a campaign will pay back before the money is spent. Their model isolates a cohort, predicts how it would behave without a campaign, launches one large video, and attributes the difference. Reviewed 14 days after launch, it calls roughly 80–90% of creator outcomes correctly.

Tatyana, CBDO at Zorka.Agency, sat down with Martin Hartmann, Director of Ventures at Almedia, and Marianna Grollo, Influencer Marketing Manager, for a 65-minute conversation about how that works in practice. What follows is the substance of it.

Why a Performance Company Went Back to Influencer Marketing

Almedia was founded around six years ago, and for its first stretch influencer marketing was not one channel among several — it was the only one.

Back then we were unable to run profitable campaigns with Google, because our product was not optimized enough.

— Martin Hartmann, Director of Ventures at Almedia

The company grew to several million dollars in revenue on creators alone. Then the channel went quiet for roughly eighteen months to two years while the product was optimised and the team scaled Google and TikTok. About a year ago they came back to it deliberately, having concluded that paid user acquisition tops out at a level they had not yet reached but could already see.

The condition for coming back was that the channel had to submit to the same standard as everything else:

Everything we do needs to be measured. We are not the fan of throwing money somewhere and then hoping it will give us a return. It's not in our DNA. Our DNA is performance.

— Martin

What Almedia Actually Sells, and Why Trust Is a Metric

The business only makes sense once you know what the product promises its advertisers.

We do not sell installs. We sell engagement.

— Martin

Almedia partners with mobile game publishers who pay a fixed amount per install and need to break even on it within six or twelve months. A studio paying $10 CPI does not want ten dollars' worth of installs; it wants users who will still be playing and spending later. So Almedia's job is not volume, it is match quality.

That reframes trust from a brand value into a performance input. On the consumer side, a play-to-earn product has to convince people it is legitimate and that the money can genuinely be withdrawn. And that is where the internal numbers get interesting:

We see in our internal metrics that this trust with influencer marketing is way higher than it is with Google.

— Martin

Specifically, creator-acquired cohorts behave like referral cohorts — people invited by a friend — rather than like paid search cohorts. The channel sits closer to peer recommendation than to advertising, and the retention profile follows.

Marianna put the mechanism in the audience's terms:

Usually the influencer audience chooses to be there. I'm actively opting in to hearing what you have to say to me. When it comes to Google and Meta, the ad fatigue is so real that you're just sitting there waiting the 15 seconds for this to be over.

— Marianna Grollo, Influencer Marketing Manager at Almedia

The commercial consequence is blunt: a dedicated user who trusts the product is worth roughly ten times a random install that converted on a promise of payment.

The Forecasting Model, Step by Step

This is the part most teams want and few have. Almedia's data team built it over the past year, and while the exact weightings stay private, the shape of it does not.

1. Isolate and predict. Take a single geo or cohort and predict how it would behave over the coming weeks with no campaign running. That prediction is the baseline.

2. Launch one large video. It has to be big enough to move the numbers above noise. A small integration cannot be measured this way.

3. Measure the whole lift, not just the clicks. This is the step that separates the model from ordinary click attribution:

If we have a big YouTuber dropping, we see that our organics go up, our conversion rates on Google go up. Maybe 70% of the value from the video is literally clicking on the video, scanning the QR code — and the other 30% is incremental value we see in our organic signups and in our paid campaigns.

— Martin

Attribution that counts only the click credits the creator with roughly two-thirds of what they actually delivered. On a channel judged against paid social, that gap decides whether it survives the budget review.

4. Read the result at day 14. Fourteen days after launch there are seven mature day-7 data points. Martin notes the window is not sacred — day-3 metrics could produce a read in ten days — and that the model will look different in a year.

5. Split the diagnosis in two. When a campaign is reviewed, performance decomposes into two independently measurable things:

• Quality of each signup — driven mostly by who the creator's audience is.
• Conversion rate per signup — driven by how the creator presented the product: their understanding of it, the craft of the integration, where it sits in the video.

Separating them is what makes a result reusable. A campaign that produced few signups of excellent quality calls for a different next step than one that converted heavily into users who never returned.

The honest ceiling, in Martin's words:

We cannot 100% accurately predict which influencer is going to perform. But once we work with them, we can very accurately measure whether something worked or not.

Where the Model Stops and Judgment Starts

Both guests were careful not to oversell it.

You can never truly crack the code. There are so many variables all the time that you need to consider and factor in that you will never have a perfect guide book.

— Marianna

Marianna described relying partly on instinct in the first thirty seconds of watching a creator — and Martin's defence of that is the most useful thing said about it:

The gut feeling doesn't come from nowhere. There is a reason why our team is getting a gut feeling for some creators that they will work and for others that they will not.

Instinct here is compressed pattern recognition from campaigns already measured. It is what the model is trained on, arriving faster than the model can.

One failure mode they call out explicitly: a video that underdelivers on views cannot be judged on conversion at all. Expect 200,000 views, get 80,000, and the campaign never had the chance to perform. Diagnosing why a campaign missed is what decides whether the creator gets another attempt — and the two guests differ on how freely to grant one, which is a more honest answer than a policy would have been.

Deal Structures That Put Both Sides on the Same Side

Almedia's answer to unpredictability is not tighter forecasting. It is contracts that stay fair whichever way the forecast lands.

CPM and CPA with a guaranteed floor. A creator is promised a minimum — Martin cites figures of $10,000 and $20,000 — and takes a share of everything above forecast.

It brings us and the influencer on the same side, which is the main goal we have with all of our partnerships.

— Martin

Two effects follow, and the second is the one most brands miss:

• Almedia pays the upside but not the downside, so a video that underperforms costs the floor rather than the forecast.
• Creators know their own channels far better than any advertiser does, and they can feel which video will overperform. Paid on upside, they have a reason to place the brand there rather than in whatever slot is next in the calendar.

The same logic extends to agencies, which Almedia treats as "satellite companies" — real ownership of their campaigns, reimbursed more when performance is better. And it mirrors the consumer product itself, which shares revenue with its users. As Marianna put it, if the company already shares money with users, sharing performance with creators is simply consistent.

Free mentions become paid partnerships. Because Freecash ranks among the more profitable side hustles, creators in the finance and side-hustle niches were already mentioning it unprompted. Almedia's approach was to find those people and formalise it — pay monthly, or per signup, whichever suits the creator.

Briefing: Talking Points, Not Scripts

On creative control both guests landed in the same place from different directions.

Almedia supplies talking points it has tested and knows move conversion or quality. It does not supply scripts.

We should not give too strict guidelines to the creators, because usually they know themselves the best. Authenticity is one of the key things for influencer marketing, and they also know their audience the best.

— Martin

If I was giving them "you need to be super excited, super enthusiastic" on a creator that is usually more toned down, it would be so obvious that it's scripted that people would actually be angry about it.

— Marianna

The example they gave of an integration that worked is worth copying. Rather than a 30-second ad block early in the video, a YouTube creator threaded the mention through a long video — noting mid-stream that he had just spent thirty minutes on something and could have earned around $10 in that time on Freecash, then returning to it naturally minutes later. Marianna's summary of why native placement matters now and did not in 2020:

Maybe it was working in 2020 when this channel was kicking off. But now that people are so used to this, you really need to have the creator on the center of the attention and build from there.

Where the Channel Is Heading at Almedia

The target is $20–30 million in creator ad spend at run rate, and the budget framing is unusual enough to quote directly:

We don't have a budget. We are at a size where whether we spend 10, 20 or 30 million on influencer marketing — we could, as long as it's profitable. But it needs to be profitable.

— Martin

That sentence is the entire argument for the forecasting model. Without a way to know profitability in advance, no one can responsibly authorise the difference between twenty and forty million a year. The model is not a reporting nicety; it is what makes the budget question answerable at all.

The constraint they name is not money but scale mechanics. Paid social scales by raising a number. Creator marketing scales by doing more work — managing 100 campaigns is nothing like managing 10, whereas a media buyer's job is much the same at $1 million or $10 million a month. Their answers are long-term creator relationships that need less hand-holding each cycle, and agencies operating with genuine autonomy.

What Transfers to Other Advertisers

Freecash is an unusual product, but four things here are not specific to it:

1. Measure incrementality, not clicks. If roughly a third of a creator's value lands in your organic and paid numbers, click-only attribution will systematically underfund the channel.
2. Split quality from conversion rate. One is the audience, the other is the creative. Blended into a single ROAS figure, neither tells you what to do next.
3. Structure deals so both sides win from the same outcome. A floor plus upside beats a flat fee for the advertiser and gives the creator a reason to choose their best slot.
4. Give talking points, not scripts. The tested points protect conversion; the creator's own voice protects the trust that made the channel work.

The conversation ends with a promise to revisit these numbers in a year. We intend to hold them to it.

---

Marianna also spoke to Influencer Marketing Hub about the same shift, in an interview by our Influencer Marketing Director: How Influencer Marketing Is Shifting From Reach to Trust. Almedia's Freecash campaigns are covered in more detail in our Freecash case study. If you're planning creator campaigns that need to answer to a CFO, our influencer marketing and performance marketing teams work this way by default — see the case studies for what that looks like in practice.