Conversions Fell Off a Cliff

In a single week, conversions dropped 35% and the team scrambled to cut ad spend. The cause wasn't a broken campaign. It was a setting nobody remembered changing.

By WebSignalytics Inc.  ·  6 min read

The scenario

Tindle & Roe is a fictional online supplier of specialty baking ingredients — the kind of small operation that runs paid search alongside a steady trickle of organic traffic. Their growth lead, a sharp marketer named Elena, watches the weekly numbers closely because the ad budget is real money she has to justify.

One Monday the dashboard showed something jarring. Reported conversions for the previous week had fallen 35% against the week before. Paid search, which usually carried the bulk of the credit, looked like it had simply stopped working. Branded campaigns that reliably produced sales were suddenly showing a fraction of their normal contribution. From where Elena sat, the picture was unambiguous: the channel had collapsed.

The confident wrong conclusion

Elena reached the conclusion almost everyone reaches when a familiar channel craters. The ads have stopped converting. Maybe the market had shifted, maybe a competitor had outbid them, maybe the creative had gone stale. Whatever the cause, a 35% drop in a week was not noise — it was a signal, and the responsible move was to stop pouring money into something that had clearly broken.

So she paused the lowest-performing campaigns and drafted a note recommending the team pull a chunk of next month's budget until they understood what had gone wrong. The logic was sound given the number she was looking at. The problem was the number itself.

The overlooked metric

Before the budget cut went through, someone asked a quieter question: were total conversions actually down, or just the conversions assigned to paid search? When Elena pulled the all-channels total, it was flat. Week over week, the business had closed essentially the same number of sales. Nothing had fallen off a cliff. What had changed was who got the credit.

The cause was buried in the settings. The account's attribution model had been switched the previous week from last-click to data-driven attribution. Under last-click, the final touchpoint before a sale takes 100% of the credit — which is why branded paid search, often the last thing someone clicks before buying, had always looked like a powerhouse. Data-driven attribution does something different: it distributes credit for each key event across all the touchpoints on the path, weighted by their actual contribution.

So the credit that last-click had concentrated on paid search was now being spread out — to organic search, to email, to the earlier visits that introduced the customer in the first place. The channel hadn't lost a single sale. It had lost its monopoly on the credit for them.

The corrected interpretation

Read correctly, nothing in the data was alarming. Total conversions were stable. A change to the attribution model had reassigned credit across the path, so last-click-heavy channels appeared to collapse while assisted touchpoints — the ones that had always been doing quiet early work — finally showed up in the numbers. The collapse was an accounting artifact, not a business event.

The honest comparison is like-for-like under a single model. When Elena looked at paid search the way you have to after a model switch — same model, this week versus a comparable week — performance was flat. The drop existed only because she was comparing a last-click week against a data-driven week, which is a bit like comparing a revenue figure in dollars to one in euros and concluding the company shrank.

Her budget cut would have starved a channel that was working fine — and, worse, would have hit the very paid-search touchpoints that data-driven attribution had just revealed were assisting conversions all along.

What to do next

If conversions appear to drop suddenly in one channel, check the plumbing before you touch the budget.

• Check whether total conversions actually changed. If the all-channels total is flat, you have a credit-allocation shift, not a performance problem.
• Confirm which attribution model is active and whether it changed recently. A switch from last-click to data-driven attribution will move credit between channels without changing a single sale.
• Compare like-for-like. Never read one model's week against another model's week — hold the model constant when you judge a channel's trend.
• Treat a sudden single-channel drop as a question, not a verdict. Reallocating budget on a 35% swing that turns out to be a counting change can do real damage.
• Write down when you change a setting. Most "the channel collapsed" panics trace back to an undocumented model or key event change a week earlier.

Tindle & Roe's ads were never the problem. The team just needed to know that the rules for counting had changed before they reacted to a number those rules produced. For more on why this happens, see why your conversions dropped after switching attribution models.

Tindle & Roe and Elena are illustrative — a composite created to demonstrate a real and common pattern.