Only the World That Happened Gets a Packing Slip

Take a hypothetical homewares account in the short stretch between graduation and move-in. Its warehouse holds 2,400 starter cookware sets already paid for. They are useful stock now and expensive archaeology once the first-time renters have moved. The business has another $180,000 in media spend it can risk without threatening the quarter. What it cannot store for later is the remaining buying window.

A nine-day graduation promotion, with a discount, ends with 1,682 orders. The previous comparable stretch produced 1,032. Target CPA had been loosened from $70 to $92, and when the late conversions finish arriving, actual CPA settles at $72—inside the retailer’s contribution limit. Of the 650 additional orders, 182 came through a new family of searches around “first apartment cookware.”

Those 650 orders do not remain in Google Ads. They become skillet boxes, label rolls, an extra collection at the loading dock, and a warehouse supervisor trying to remember where the emergency tape gun went. This is success with weight. You can stack it on a pallet.

What the promotion does not leave behind is the other week.

There is no pallet from the promotion at the old target. No packing slip from the looser target at full price. No duplicate auction in which competitors made the same bids while graduation demand politely held still. The missing counterfactual, in a familiar failure of cross-functional support, ships absolutely nothing.

The visible week can sustain two serious accounts of itself. Google says Target CPA uses historical information and auction-time signals to set bids, and that an overly low target can suppress traffic. Giving the bidder more room may have reached a real group of young renters the account had been declining one auction at a time. A coherent query family and an acceptable mature CPA are not hallucinations. There are 182 real orders attached to those searches.

The promotion was also a price change during a demand event. Google’s own seasonality guidance exists because short sales can produce large temporary changes in conversion rate. Search interest and auction competition can move at the same time. Perhaps the bidder found a durable customer pool. Perhaps graduation, a discount, and unusually kind auctions assembled a convincing impression of one.

Maturity makes the CPA honest. It does not make the history singular.

That distinction would be easier to preserve if evidence arrived with equal physical force. It does not.

Google Ads Change history can show the user attached to a settings change and place that change beside performance. This is useful. It is also an astonishingly efficient little biography. A person loosened the target; volume rose; a new query family appeared. Subject, verb, result.

Temporary demand has no email address in the User column. Auction softness does not join the video call. The promotion belongs partly to merchandising, the price belongs partly to finance, the orders belong partly to the site, and the new searches belong partly to people moving into apartments. Their combined explanation has accuracy going for it, but terrible stage presence.

So the account’s language begins to improve itself.

The first recap says: “After the target change, during the graduation promotion, volume increased at a stable mature CPA.”

The budget request says: “The target change reached first-apartment shoppers at a stable CPA.”

The planning slide says: “Looser targets unlocked a first-apartment customer pool.”

Nobody has falsified a number. One modest word—after—has simply lost its job. In its place is unlocked, a verb with an actor, a mechanism, and excellent prospects for advancement.

This is how a commercially right action begins to manufacture an authorized narrator of the past.

The operator who made the change is not grabbing credit from a smoke-filled room. They have the most complete story anyone can tell without stopping for three minutes: I changed this, then customers came. They are asked to present the result, then to apply the lesson, then to review other campaigns for the same opportunity. Their judgment deserves more trust than it did before; the result was acceptable within the bounded risk they took. But the organization quietly upgrades a warranted claim about the decision into an unwarranted claim about the cause.

The newly trusted operator chooses where to loosen targets next. Each success produces more orders associated with the same theory. Each failure can be assigned to inventory, query coverage, ads, or time—perhaps correctly. Meanwhile, the alternative explanation becomes harder to voice without sounding like an objection to revenue already in hand.

No one has to be vain or dishonest. Action has a structural advantage over restraint: it leaves receipts. If the company had kept the old target and missed a durable pool, the absent customers would not arrive as an angry delegation. They would remain perfectly quiet, each carrying the product they bought from somebody else.

This is also why giving “the causal readout” to a supposedly neutral role does not solve much. The paid-search operator owns the lever and sees the auction reports. Merchandising knows what the promotion changed but not which searches almost served. Analytics can define a comparison but cannot conjure the unrun week. Finance can decide how much loss is survivable but not which customers would have arrived anyway. Authority is distributed because the event itself was distributed. The story concentrates because organizations need a sentence they can reuse.

That $180,000 has to be placed while the boxes are still useful and first-time renters are still searching. A campaign experiment could create a cleaner contrast by splitting traffic or budget, but Google notes that an unresolved experiment may need four to six weeks to gather enough data. The warehouse does not get those weeks back. A small test can preserve optionality and still be too small to capture the short buying window or show whether first-apartment demand is durable. Waiting for better evidence can be a disciplined way to donate customers to a competitor.

In this account, I would release the budget.

Not because unlocked has won. The warehouse defined the wager before the argument began: paid-for stock, a loss the quarter can carry, and customers whose need expires. Even if the promotion did most of the causal work, the downside remains survivable; if the new pool is durable, the missed buying window cannot be repurchased. Experimentation is not morally superior to accepting a bounded risk, and uncertainty is not an instruction to stand still.

But releasing the money does not end the puzzle. It intensifies it. Every additional order will make unlocked feel more true, even though more orders under the new setting cannot recreate the old setting. The spend may produce margin, customers, and useful observations. It will also produce political evidence for the person whose theory chose the action.

The next honest observation is already hiding inside the 182 orders: they averaged about 20 a day. After the discount disappears and a full conversion cycle matures, do “first apartment cookware” searches still deliver anything like 20 daily orders at a CPA inside the contribution limit? If they do, the claim that graduation and price assembled a temporary pool has to work harder. If they disappear with the promotion, unlocked loses its right to travel across the account. Neither result is a performance review. It is pressure on what the next decision may claim.

Back at the packing table, the promotion has left 1,682 labels. Each has an order number, a postcode, a shipping class, and a cost.

None has a field for who was right.