Position
Transparency in display advertising
Display advertising has spent a decade being told to be more transparent, and most of the response has been a page like this one. So here is ours, written as specifically as we can manage, including the parts where the honest answer is that we have not built it.
Transparency is not one thing. It gets used to mean at least six different problems, and a vendor can be genuinely good at one while being useless at the others. Below is each of them, what usually goes wrong, where we stand, and what the product actually does about it today.
01
Cost and fees
The oldest transparency problem in the trade is not knowing how much of a media budget reached media. Undisclosed markup, rebates and platform fees can sit between a brand and its inventory without ever appearing on a report.
A brand should be able to see what it paid, what the media cost, and what sat in between. That is an uncomfortable number to publish, which is exactly why it belongs in the product rather than in a quarterly conversation.
Emptora models wholesale cost and client revenue as separate records, so gross profit and margin are calculated rather than asserted. Cost lines are typed, which means an agency fee and a platform fee do not disappear into one bucket called media.
In the product today
- Cost lines carry wholesale cost, client revenue and a cost type covering media, creator, usage, production, platform fee and agency fee.
- Margin, markup and gross profit are computed from those records and shown to entitled roles only.
- Entitlement is enforced in the database with row level security and again in the query layer. Internal economics are never sent to a client browser and hidden with styling.
02
Supply path
Programmatic buying can put five intermediaries between a budget and an impression. When reporting stops at the channel level, nobody can tell whether the money reached quality inventory or drained into a long tail nobody chose.
Channel level reporting is where most dashboards stop, and it is the level at which almost no useful question can be answered. Knowing you spent on display tells you nothing. Knowing which publisher and which placement took the money tells you what to change.
Emptora stores delivery at the grain it was bought at, so the supply path is available rather than reconstructed.
In the product today
- Programmatic delivery is stored daily by channel, platform, publisher, placement, line item and audience.
- The programmatic overview breaks spend and outcome down by publisher and by placement.
- Every row records the provider it came from, so two sources reporting the same campaign stay distinguishable.
03
Measurement and data quality
Platforms mark their own homework, partners report their own results, and competitive tools model what they cannot observe. Blend the three into a single total and you get a number that looks authoritative and cannot be defended.
This is the problem Emptora was built around. Every quantitative record in the system is classified by how it was obtained, and that classification travels with the number all the way to the screen.
When figures with different provenance are added together, the total reports the weakest source in it and is marked as mixed. A blended total that quietly inherits the confidence of its best input is a lie told by arithmetic.
In the product today
- Every record is classified as actual, reported, estimated or derived, and shows a badge wherever it appears.
- Totals that combine classifications report the weakest one present and flag themselves as mixed.
- A metric that cannot be calculated returns an em dash rather than a zero, so a missing denominator never renders as a real result.
- Competitive spend is shown as a range at the level it was observed, and is never attributed down to an individual ad.
04
Creative
Creative reporting tends to rank whatever spent the most, then call it a top performer. Competitive ad libraries have the opposite failure, ranking whatever has run the longest as though longevity were proof of results.
Longevity is evidence that somebody keeps paying. It is not evidence that the ad works, and a product that blurs the two is training its users to draw the wrong conclusion.
Emptora keeps the two populations apart. Creative you run is ranked on measured outcomes. Creative observed in market is ranked on presence, and the card says plainly that there is no outcome data behind it.
In the product today
- The ad library ranks your programmatic and paid social creative on spend, impressions, CTR, CPA, ROAS or a derived efficiency score.
- Observed competitor creative sits in a separate ranking based on how long each ad has been running.
- Comparing your creative against observed creative shows the unmeasured columns as not measured rather than as zero.
- Each ad carries its share of channel spend and its share of the whole paid mix, so a good rate on a small line item is read in proportion.
05
Brand safety and fraud
Invalid traffic, misrepresented placements and unsafe adjacencies are handled by verification vendors sitting outside the reporting stack, which means their findings often never reach the report a client actually reads.
Emptora does not run verification, and it would be dishonest to imply otherwise. What it can do is refuse to launder a source. Where verification data is imported, it stays attributed to the vendor that produced it.
This is a gap we would rather name than paper over. Pre-bid and post-bid verification belongs with the vendors who specialise in it, and the integration work to bring those findings into Emptora reporting has not been built.
In the product today
- Every delivery record stores the provider that supplied it, so a verified figure and a platform figure never merge silently.
- Imports are validated row by row, and an invalid row is rejected and reported rather than dropped.
- Any import can be rolled back exactly, because every row it created is tagged with its import job.
06
Accountability
Transparency that depends on asking your agency for a report is not transparency. It is a favour, and it stops the moment the relationship gets difficult.
The point of putting this in software is that a client should not have to request it. If the number is in the system, the people entitled to see it can see it whenever they want, in the same view the agency uses.
That is also why the permission model is explicit rather than implied. Five roles, a written matrix, and isolation enforced at the database rather than by hiding buttons.
In the product today
- Five roles with a documented permission matrix covering what each one can read and change.
- Tenant isolation enforced with row level security, with tests that assert one client cannot reach another's data.
- An audit trail of imports, including who ran them, what changed and what was rolled back.
What we do not claim
A transparency page that only lists strengths is an advertisement. These are the limits we would tell you about on a call, so they may as well be on the page.
- Emptora does not run ad verification, viewability measurement or fraud detection. Those belong to specialist vendors and are not built here.
- It is not a demand side platform and does not buy media. It reports on buying that happened elsewhere.
- Competitive figures are estimates. They are shown as ranges with a confidence value and should be read as directional.
- Where an integration is not built, the documentation says so rather than describing it in the future tense.
Further reading
Bannerflow published a useful survey of the same territory, covering the supply chain, fee structures, verification vendors and the in-housing argument in more depth than we go into here.
The ultimate guide to transparency in display advertising, Bannerflow
Hold us to it
If any of the above does not match what you see in the product, that is a bug and we want to hear about it.