Cheap Impressions, Unclear Brand Impact
A national pet-supply brand spends $120,000 per month on open-market programmatic display through a demand-side platform. Average CPM is $1.80, measured viewability is 41%, and an outside audit found that 62% of impressions served on sites flagged as made-for-advertising. Last-click reporting shows display contributing 0.4% of online conversions, and the growth team has consequently deprioritized the channel. A brand lift study ran for six weeks and showed aided awareness up 1.2 points, but the result was not statistically significant at the chosen sample size. The chief marketing officer wants a recommendation: double the programmatic budget because CPMs are the cheapest in the media mix, or shut the channel down because it does not convert. The media lead suspects both options are wrong but needs a defensible position.
1. Audit the economics. What do the CPM, viewability, and made-for-advertising figures actually mean together?
Read the worked answer
The headline CPM is misleading, and the first job is to restate cost in terms of what was actually delivered.
- Viewable CPM: $1.80 divided by 41% viewability is roughly $4.39 per viewable impression. Already the cheapness is less impressive.
- Quality-adjusted cost: if 62% of impressions sit on made-for-advertising sites, then only about 38% of impressions are on plausible inventory before viewability is even applied. Combining the two, the effective cost per viewable impression on legitimate inventory is many multiples of $1.80.
- Frequency and reach: neither is reported, so the brand cannot say whether it bought broad reach or repeatedly hit a small pool.
This pattern, very low CPM plus low viewability plus high made-for-advertising share, is the signature of a supply-path problem rather than a channel problem. The judgment is that the channel has not been tested fairly; what has been tested is a low-quality inventory strategy. Shutting down would discard a legitimate upper-funnel tool, and doubling down would buy more of the same waste.
2. What changes to inventory, supply path, and measurement would you make before recommending a budget decision?
Read the worked answer
Restructure first, then decide, with a clear timeline.
Inventory and supply path:
- Add pre-bid filters for viewability thresholds, brand safety, and made-for-advertising exclusion, and cut the number of supply-side partners to shorten the path.
- Shift a meaningful share into private marketplace deals and allowlists with publishers whose audiences match pet owners, rather than open exchange auctions.
- Set frequency caps and report unique reach, not just impressions.
Measurement:
- Stop judging display on last-click. Build a holdout-based read: geo-matched control markets with no display, or a platform-level ghost bid or intent-to-treat design.
- Redesign the brand lift study with adequate sample size per cell, a pre-registered primary metric, and a duration long enough to detect a realistic effect. The prior study was underpowered, so its null result is uninformative rather than negative.
- Add a marketing mix model or at minimum a consistent weekly geo-level regression once enough weeks of clean data exist.
Run the restructured buy for one quarter at current budget, then compare cost per viewable impression on quality inventory, reach, and lift against baseline before scaling or cutting.
3. How would you frame the recommendation to the CMO, who is choosing between doubling and shutting down?
Read the worked answer
Give a third option with an explicit decision gate, and make the reasoning legible in business terms.
Frame it as: the current buy is not a fair test of programmatic, so neither doubling nor shutting down is supported by the evidence. The $1.80 CPM bought a large number of impressions that were mostly unseen or on low-quality sites, which is why last-click conversion is near zero and the brand lift study was inconclusive.
Proposal: hold budget flat for one quarter while moving at least half of spend into private deals and allowlisted publishers, applying viewability and made-for-advertising filters and capping frequency. In parallel, fund one properly powered brand lift study and a geo holdout.
State the gate in advance: if quality-adjusted reach improves and lift becomes significant at a cost per incremental aware consumer that beats the alternative channels, scale. If lift remains flat with adequate power, cut.
Present both outcomes as acceptable and pre-committed, which is what makes the recommendation credible to a CMO who is being asked to choose between two extremes.