Opportunities

Opportunities lists the few clients and ad units that earn clearly less for each ad request than they should, and says why. It is for the team that can act on the setup in Google Ad Manager (GAM) or talk to the client. Stealth calculates it once a week from the data it already reads, and never changes anything in GAM.

How it works

Stealth compares how much each client or ad unit earns per 1,000 ad requests with how it earned before, and with how the rest of the network moved in the same weeks. If the whole network fell 30% and a client fell 47%, only the part beyond the market counts as an opportunity.

• Last 4 weeks. The comparison uses the last 4 complete weeks. It needs 16 weeks of history in the network. Until then the page says from which date opportunities appear.
• Compared with what. With at least a year of history, a client is compared with the same 4 weeks a year earlier and with its usual level over the past year. With less history, only the usual level is used. Ad units are compared with their usual level.
• Only real gaps. An item appears only when the drop is clear and holds beyond normal variation from week to week. Small clients and small ad units, and items worth very little per month, are left out.
• Days with incomplete data. Days when Google Ad Manager reports more than Stealth stored are left out of the calculation. Days with an incident are not left out, because a sustained drop is exactly what this page looks for.
• Never mixed. Each network is calculated in its own currency, and amounts are never added across items.

Value

Each item shows what it would add per month if it earned again what the market explains, at the current traffic. It appears as a range, for example R$ 7.8k to 9.7k/month. The lower number is the probable minimum, the upper number is the whole gap. It is a ceiling, not a forecast, and items are never added together.

Confidence

Hover the confidence of an item to read why it has that level.

Causes

When the demand data points to a cause, the item names it:

• A demand channel (such as Header Bidding or Open Bidding) stopped, or fell much more than the same channel across the network.
• The ad server took impressions from Ad Exchange at a much lower eCPM.
• Direct sales ended. This is a matter for the sales team, not for yield.
• Without any of these, the item says whether the lower fill rate, the lower eCPM or both explain the drop.

Network items show a demand channel that moved against the network’s total revenue. They always appear under Under observation, because that can be the market and Stealth has no outside market reference.

What you see

Open Opportunities in the sidebar, next to Incidents. Under the title there are five tabs with counts: Open, Under observation, Makes sense, Applied and Dismissed. With more than one network, a filter chooses the network.

Scoreboard

A line under the tabs says how the applied items turned out. Before any measurement it says when the first result is due. After that it shows, for example, 3 of 5 applied items delivered and how many were inconclusive. Up to 10 measured items it shows counts, after that percentages.

The list

Items come with the largest probable value first. Each row shows:

• The client, ad unit or channel, the network and whether it is a Client, an Ad unit or a Network item.
• The cause in plain words.
• Against expected, such as 47% below expected, with the raw change and the market change underneath.
• The Value range, the Confidence and the verdict, with who recorded it and when.

Click the name to open the item. At the bottom, excluded lists the clients and ad units left out before any test because the comparison would not be fair, each with its reason: it stopped sending most of its traffic or revenue, its traffic changed by more than half, requests grew while revenue held, it lost traffic in every demand channel at once, its traffic moved to other ad units of the same client, or it is the network’s own house inventory.

Item page

The page has a one sentence summary, the value, the confidence and the cause, and then:

• Revenue per 1,000 Requests: a weekly chart of the item against the level expected by the market, with the comparison period and the last 4 weeks shaded.
• Before and Now: weekly requests, impressions, fill rate, eCPM and revenue, and how much of the change fill rate explains against eCPM.
• By Demand Channel: the client’s revenue in each channel, before and now, next to the same channel’s change across the network.
• Ad Units: for a client item, its ad units below expected, largest share of the loss first.
• What to Check: steps based only on what the data shows, and the Google Ad Manager report that rebuilds the numbers (Reports > New report).

The footer shows the date the data goes up to, the days left out and when it was calculated.

Steps

Record a verdict

Any member can record a verdict. Changing a verdict another member recorded needs an owner or administrator. After a verdict, the item shows who recorded it and when, with Change.

• Makes sense moves the item to its own tab.
• Does not make sense hides the item. It comes back if the gap moves more than 30%, a new cause appears or after 8 weeks.
• Applied starts a measurement: Stealth compares about four weeks, starting a few days after the date applied, with the same weeks of similar items that were not changed. The result is Delivered, Not delivered or Inconclusive, and feeds the scoreboard. Network items are never measured.

Verdicts help tune the method and never change the numbers.

Opportunities on the Panel

The Opportunities card on the Panel shows how many items are open and how many have high confidence, the top 3 open items with their cause and value range, and the scoreboard line once a measurement exists. See all opens this page. The card carries no Estimated badge, because the values come from real Google Ad Manager data and are shown as a range.

FAQ