Supply Path Optimization for Mobile: Your Path Is Probably Not the Problem

Supply path optimization gets sold as the fix for programmatic waste. The published evidence does not support that, and the most useful thing a mobile buyer can do is find out which problem they actually have before spending a quarter on the wrong one.

What a supply path is

Between your budget and a publisher sits a sequence of intermediaries. A demand-side platform buys from a supply-side platform or exchange, which may in turn be reselling inventory it sourced from another intermediary instead of from the publisher directly. On the mobile side of that chain, a mediation layer inside the publisher's app decides which demand source gets each impression.

Two properties of the chain matter, and they are different problems.

Depth. Each hop takes a cut, so the longer the path, the less of your money is left to buy attention. Your effective CPM is higher than the rate you agreed to, by the sum of what every hop took.

Duplication. The same impression can reach you through several routes at once, at different prices. Mediation resolves which route wins, and it is not obliged to pick the one that was cheapest for you. So you can end up paying a more expensive route's price for inventory you had a cheaper claim on, and standard reporting will not show you that it happened.

We raised the first of these in how to choose a mobile DSP, where the opening question is what the fee structure looks like per hop.

What the benchmark data actually says

This is the part that should change how you plan, and it cuts against the premise of most writing on this topic.

The ANA's Q1 2026 Programmatic Transparency Benchmark found higher-performing advertisers converting 54.0 percent of programmatic spend into qualified impressions, against 32.1 percent for lower-performing advertisers. ANA described the resulting 21.9-point gap as "the largest recorded by the Benchmark to date." Of the weaker cohort, the release says it "loses more than two-thirds of every dollar before it reaches the end consumer."

Read what is being counted before drawing the obvious conclusion. ANA's TrueAdSpend Index "measures the share of programmatic investment delivering fraud-free, measurable, viewable, and MFA-free impressions," and the Benchmark's qualified impressions appear to be counted against those same criteria, though the release does not spell the equivalence out. Either way it is a measure of inventory quality, not of how much money the intermediaries kept. A dollar can become a qualified impression after a long chain has taken its margin, and a dollar can reach a publisher intact and produce an impression nobody could see.

And when ANA looked at what drives that gap, fees were the smaller component by a wide margin. Transaction costs differed between the two cohorts by just 2.4 percentage points. Media productivity losses differed by 19.4 points. The release states the conclusion directly: "quality, not cost, is the primary differentiator in programmatic performance."

So the honest position for a DSP writing about supply paths is this. The Benchmark measures what separates two cohorts rather than diagnosing any one advertiser, but on its evidence the likelier candidate for underperformance is what you are buying, not how many hands it passed through on the way. Path depth is real and worth knowing. It is not where most of the money goes.

Three caveats on the data itself. These are benchmarks across the advertisers who participate, not universal rates. ANA does not break out in-app inventory separately, so treat the figures as the shape of the problem rather than as your numbers. And the Benchmark reports quarterly: the figures above are from Q1 2026, a later quarter has since been published, and the gap it measures moves. Check the current release before quoting any of it.

The one quality signal that moved the wrong way

ANA reports made-for-advertising exposure running between 0.4 and 0.6 percent throughout 2025, then rising to 1.1 percent in Q1 2026.

That is small in absolute terms, and one quarter is not a trend. It is worth watching because MFA is the quality failure most directly under a buyer's control.

An MFA property is built to serve ads rather than to be used: high ad density, auto-refreshing placements, heavily sourced traffic, generic templated content. It is not fraud. Real people are sometimes there and the impressions are genuinely served, so filtering for invalid traffic will not remove it.

It persists for a structural reason: MFA inventory is cheap. An algorithm optimizing toward a low cost per impression will keep finding it, and the metrics that would catch it are not the ones such a goal rewards. Set the goal at the top of the funnel and you have built a machine for buying it.

That is also the fix, and it is the one the benchmark points at. A model optimizing toward a properly measured downstream event stops rewarding cheap impressions that lead nowhere. In practice that means naming a post-install event your product actually earns on, instrumenting it before the campaign starts, and accepting a longer read while that event resolves.

One scope note for a mobile reader: the industry definition of MFA was written around websites, and ANA does not break out in-app exposure separately.

Why path work is still worth doing

Three reasons that survive the data above.

Unauthorized selling is a different problem from inefficiency. A seller offering inventory it has no right to sell is not taking a margin from you. It is selling you something that may not be what it claims. Publishers declare which companies may sell their inventory, and those declarations are public, so this is the one part of path analysis a buyer can audit without a vendor's cooperation. Worth doing whatever the fee analysis says.

Duplication has a direct price effect. It is also something you can act on, once you can see which routes reach the same inventory and at what price.

Fee transparency is a proxy for everything else. A platform that will not state its take rate per hop, or pass complete supply chain data, is usually not forthcoming about the other things either. The answer matters less than the willingness to give one.

What to ask your DSP

Take rate, stated as a percentage, per hop. Not a blended number, and not an assurance that the platform is competitive. Ask whether it varies by channel, format, or supply source, because if it does, the platform has a financial preference about where your money goes and the interface will not show it to you.

Complete, unedited supply chain data. Ask whether the platform passes it on in full, and whether it will show you the chains behind your own spend instead of a summary of them. A platform that filters or truncates what it passes on is making a choice about what you get to see.

Log-level data. Impression-level records showing what was bought, from which supply source, and at what price. Ask at what account size it becomes available, in what format, at what frequency, and at what cost. Some platforms will not provide it. That is not automatically disqualifying, but it should change what you are willing to pay, because you are buying a service you cannot verify.

The direct and resold split, by market. This matters most in smaller markets, where resold paths often carry inventory no direct integration reaches.

What path work costs you

Cutting a genuine duplicate costs no reachable inventory. If two routes carry the same impression, removing one leaves it reachable through the other. It can still cost you money if you cut the cheaper route and keep the expensive one, which is why the exercise starts with knowing the prices, not the count.

Cutting unique supply costs reach outright. Paths are rarely purely duplicative. A reseller you drop may be the only route to a long-tail publisher, a regional network, or inventory one platform holds exclusively. That constraint binds fastest in smaller markets and on thin supply.

Keeping a path list current takes ongoing work. Integrations change and resellers appear, so a list curated in January describes January.

Measure the result on conversions, not on CPM. A change that lowers your CPM and your conversion count has made things worse, and a dashboard comparison will not separate a cheaper buy from a worse one, because the inventory mix moved underneath the comparison. That is what a holdout test is for: suppress the channel in some markets, run it in others, and compare totals rather than trusting a before-and-after.

Where we sit in this

Edge226 is a DSP, which makes us one of the hops. Taking supply path advice from an intermediary is reasonable only if the intermediary will answer the same questions it tells you to ask, so put the four questions above to us in writing.

The more useful conversation is usually the one before that. If you want to work out whether your problem is the path or what you are buying, get in touch.

FAQ

What is supply path optimization?

Supply path optimization is the practice of reducing the number of intermediaries between an advertiser's budget and the publisher serving the ad, and removing duplicate routes to the same inventory. The aim is a larger share of spend reaching working media and fewer cases where a duplicate route wins an impression you could have bought more cheaply.

Is supply path optimization where programmatic waste comes from?

Mostly not, on the published evidence. The ANA's Q1 2026 benchmark found transaction costs differed between higher- and lower-performing advertisers by 2.4 percentage points while media productivity losses differed by 19.4 points, and concluded that "quality, not cost, is the primary differentiator in programmatic performance." Path work is still worth doing, for narrower reasons than it is usually sold on.

What are qualified impressions?

Qualified impressions are what ANA's TrueAdSpend Index counts: it measures the share of programmatic investment delivering fraud-free, measurable, viewable, and MFA-free impressions. It measures inventory quality, not how much of your budget intermediaries retained. The two are frequently confused.

What is MFA inventory?

Made-for-advertising inventory is content built for serving ads rather than for reading, typically with high ad density, auto-refreshing placements, heavily sourced traffic, and generic templated content. It is not fraud, since the impressions are genuinely served, which is why invalid-traffic filtering does not remove it and why an algorithm optimizing toward cheap impressions keeps finding it.

Does supply path optimization reduce reach?

Only when it cuts unique supply. Removing a genuinely duplicate route leaves the impression reachable through the path you kept. Reach is lost when a dropped reseller was the only way into a long-tail publisher or a regional market, which is why the first job is telling duplicates apart from sole routes.

‍