The Empty Chair Is in the Auction

Imagine an ordinary Monday pipeline meeting. The empty chair was supposed to last six weeks.

It belongs to the Plains territory at a company that sells air-quality retrofits for warehouses and light-industrial buildings. The previous rep left in May. Recruiting has taken longer than expected, so the leads are being shared by two people who already have territories of their own. The arrangement has now lasted fourteen weeks, which is how temporary things begin applying for permanent status.

The VP of Sales has brought a slide showing that Plains leads qualify less often and close at roughly half the rate of the rest of the country. The CFO has brought the obvious question: Why hire another rep for the territory that produces the weakest pipeline?

Across the table, the paid-search lead looks at the vacant chair.

She knows something the slide has made easy to forget. The company did not discover a weak market and then lose its rep. It lost its rep and then began measuring a market through the loss.

That distinction is about to matter in the auction.

The Search account has a deliberately boring conversion architecture. A form becomes sales-eligible when the building is in the service area, clears a square-footage threshold, names an eligible project, and gives a plausible start window. The rule is mechanical. It fires before Sales touches the lead, and it is the primary event used by Target CPA.

Qualified lead and closed sale are imported later from the CRM as secondary actions. The team wants those outcomes for diagnosis. It does not want the quickest caller, the most permissive qualifier, or the rep with unusual discount authority to decide what kind of demand Marketing buys next.

For years, “secondary” served as a small treaty between the two departments. Sales could report what happened after the form. Search could optimize to the form. Both could argue about quality without quietly giving one side control of the other’s machinery.

Then Google announced journey-aware bidding.

The public claim is short. In beta, Search campaigns using Target CPA can learn from biddable and non-biddable conversion goals across the lead-to-sales journey. Google’s accompanying product page calls it a backend upgrade and says the way to give it visibility is to import the relevant conversion actions.

This sounds like a welcome escape from form-fill optimization. A sales-eligible form is better than a raw lead, but it remains a proxy. If later outcomes help the bidder recognize which eligible forms become customers, perhaps the campaign can keep the volume of an upper-funnel target while gaining some of the judgment of a lower-funnel one.

That is the optimistic reading, and it may be right.

But the empty chair supplies another reading.

Nothing about the form rule has changed in the Plains. The same buildings still clear the same threshold. The same projects are still serviceable. What changed is what happens after the form: a lead waits until an overloaded rep has room, an unfamiliar local number gets a second attempt instead of a fourth, a discovery call is shortened because another territory has a proposal due. By the time the CRM says not qualified or says nothing at all, buyer demand and company capacity have been folded into one outcome.

Now allow that outcome to teach.

The bidder does not need to announce, “I have developed a view on regional staffing.” It only needs to find patterns that improve its predictions. Geography may be one. Query language, industry, device, hour, audience membership, and countless combinations may carry some of the same information. If leads resembling the well-covered territories travel farther through the CRM, a learning system may become more willing to win those auctions and less willing to win the Plains versions.

The company can still hit its sales-eligible-form CPA. That is the disorienting part.

Target CPA has not been changed to Closed Sale. The visible goal can remain exactly where the paid-search team put it. Yet the policy behind the bids may begin answering a different question:

Among all the forms that satisfy this rule, which ones arrive where our company is currently best at turning them into revenue?

That can alter the distribution of demand without ever asking a late-stage event to count toward the target.

On the next slide, the Plains problem appears to improve. Spend moves toward places where the company closes more. Aggregate form CPA holds. Sales rate rises. Everyone has a pleasant quarter.

Meanwhile the under-covered territory receives fewer chances to contradict its reputation.

This is where journey learning can become more than a quality improvement. It can become a policy for rationing market attention according to the company’s current abilities and preferences. The policy is especially hard to see because it does not need to appear as a setting. It emerges from the examples the company supplies.

The Plains rep leaves. Response slows. Later outcomes weaken. The system may learn to favor demand that reaches staffed desks. Plains pipeline shrinks. Finance sees less evidence for the replacement hire. The chair stays empty.

The company has not merely adapted spend to capacity. It has allowed capacity to help decide which demand will be observed, and then used the observed demand to justify the capacity.

The first loop is efficient. The second is self-confirming.

No conversion-tracking audit can settle which loop the company wants. The CRM records can be timely, deduplicated, correctly categorized, and entirely faithful. The problem is not dirty data. The problem is that a perfectly true closed sale says what happened when a buyer met the company as it existed that day. It does not say what the buyer would have become in the company the leadership says it is trying to build.

Back in the meeting, the VP of Sales makes the strongest case for enabling the beta.

“We have people ready to sell in four regions,” he says. “Why wouldn’t we send more opportunity there?”

He is not confused. If the company’s actual strategy is to maximize near-term sales through its current footprint, the empty chair is relevant information. A bidder that directs more demand toward staffed, responsive territories may be doing exactly what the business needs. The gain does not become illegitimate because some of it comes from operational fit rather than a purer essence called lead quality.

The paid-search lead’s objection is narrower.

“Is the chair part of our strategy,” she asks, “or is it a problem we plan to fix?”

The answer changes what the beta is authorized to learn.

If the Plains territory is being wound down, using later outcomes to favor the remaining operation is coherent. The company has chosen where it wants to be good. Letting bids reflect that choice is honest capacity allocation.

If the vacancy is temporary, the same behavior can work against the stated plan. The company needs continued contact with Plains demand while it recruits, both to serve customers and to learn whether the replacement role deserves investment. Allowing today’s understaffing to reduce tomorrow’s opportunities would make the rebuilding case weaker precisely because rebuilding is needed.

The distinction is not automation versus control. It is present efficiency versus the right to discover a future different from the present.

At this point the product’s opacity stops being an abstract complaint and becomes an authority problem.

The ordinary Google Ads help page still describes secondary actions as observation only, reported in All conversions and not used for bidding, apart from its documented custom-goal exception. The journey-aware announcement speaks instead of learning from non-biddable goals. The public pages do not reconcile those concepts closely enough to show this team exactly which of its secondary CRM actions would participate.

Nor do they give the paid-search lead a documented way to authorize the sentence she actually wants: learn from stable qualification patterns, ignore closed-sale differences during the vacancy, and do not carry the vacancy forward after the new rep starts. Google does not publicly describe per-stage influence, a contribution view, or what a removal causes the model to forget.

Those unknowns matter here for one reason. Without a narrower control supplied by the beta itself, the company is not consenting to “better lead quality” in the abstract. It is consenting to an unresolved mixture that may include its territory coverage, response times, rep judgment, pricing practice, and buyer characteristics. The whole mixture may improve revenue. The company still has to want the whole mixture imitated.

The CFO closes the laptop halfway, which is the universal finance gesture for please arrive at a decision before this becomes a philosophy seminar.

The paid-search lead does.

They will wait.

Not forever, and not because the announcement fails to explain a model architecture that no advertiser was going to inspect anyway. They will wait because the Plains outcome is currently unstable in the exact direction that could ration away the evidence needed for a decision the company has not made.

They will revisit the beta after the territory is staffed, the call burden returns to something intentional, and Sales has used one qualification standard long enough for “qualified” to describe a repeatable business choice. If their account-specific beta terms can show a real boundary around which journey stages teach, that could justify an earlier test. The public announcement alone cannot.

The opposite decision would have been defensible if the VP had said, plainly, that the company no longer intended to build in the Plains. Then the later outcomes would express a durable operating choice rather than a temporary failure. Turning that choice into auction preference would be a use of the beta, not an accident of it.

The meeting ends without anyone changing the primary conversion. That was never the live decision.

As people stand, the shared phone beside the empty desk begins to ring. One of the overloaded reps looks at it, looks at the proposal open on his screen, and reaches across.

The market has not gone quiet. The office has.

An acquisition objective can keep the same name, the same target, and the same conversion count while becoming a policy about which parts of the market deserve the company’s attention. Teaching without counting makes that possible.

Before the empty chair gets a vote in the auction, decide whether it represents the company you have chosen—or merely the company you have not fixed yet.