> ## Content Index > Fetch the complete content index at: https://www.livain.com/llms.txt > Use this file to discover other available public pages before exploring further. # I raised a client's Google Ads budget 59% while the account could see one lead in sixteen: how to scale on purpose with broken attribution - URL: https://www.livain.com/blog/scaling-google-ads-budget-while-attribution-is-broken/ - Published: 2026-09-24T07:00:00.000Z - Updated: 2026-09-24T06:59:59.000Z - Description: Impression share lost to budget was 8.2% on weekdays and 14.5% on Sunday — the caps were closing exactly where the demand was. Three conditions that make scaling with broken attribution a decision rather than a gamble. - Author: Remco Livain - Tags: Google Ads, Attribution, Performance Marketing, Marketing Analytics Google Ads recorded one conversion last week for a building-materials client. I counted sixteen leads. That is roughly 94% of the outcome invisible to the system that is supposed to be optimising toward it. On Monday I raised the daily budget by 59% anyway. I want to write down why that was a defensible decision rather than a reckless one, because "we scaled while the measurement was broken" is a sentence that normally appears in a post-mortem. ### The cap was biting on exactly the days that matter The finding that moved me had nothing to do with conversions. Saturday and Sunday took 32% of the week's spend across 28.6% of the days — which on its own is unremarkable, a business whose customers are private homeowners doing their research at the weekend. Mobile was 72% of spend, consistent with the same story. The number that mattered was [impression share lost to budget](https://support.google.com/google-ads/answer/7103314?hl=en&ref=livain.com), split by day of week rather than averaged across it. Monday to Friday: 8.2%. Saturday: 11.6%. Sunday: 14.5%. Sunday was also the single most expensive day in the entire fortnight. The weekly average hid that completely. Averaged, the account looked mildly budget-constrained and mostly fine. Split by weekday, the daily caps were slamming shut precisely when the demand showed up, and the account was buying a smaller share of its best auctions than of its worst ones. That distinction — lost to budget versus lost to rank — is the one that decides whether more money does anything at all. [Adalysis put it plainly](https://adalysis.com/blog/ppc-kpi-monitoring-how-to-diagnose-and-work-with-impression-share-loss-due-to-budget/?ref=livain.com): raise bids when budget is the constraint and you just burn through the cap faster; raise budget when rank is the constraint and you buy more impressions from a losing position. Here, every campaign was running on a maximise-clicks strategy with no bid ceiling, which makes budget a direct lever on both — more budget permits higher bids, which reaches into the rank-lost share as well. One of the starved campaigns flipped from budget-constrained to eligible within minutes of the change. The cap was real, not theoretical. So there is a genuine, measurable, conversion-independent reason to spend more. Clicks were up 14% week on week for 7% more money, and the cost per click had come down. The efficiency was already improving before I touched anything. ### And here is what I am giving up With attribution at roughly 6% coverage, I will not be able to say which part of the extra spend worked. Not "it will be hard to say." Will not be able to. If leads go from sixteen to twenty, I will know the account produced four more and I will not know which campaign, which query or which day produced them. The two fixes that would close that gap have been sitting with the client's web developer since late August — persisting the attribution cookie as order metadata, and replacing a hardcoded zero conversion value with a real enquiry value and a transaction ID. Neither is complicated. Neither is done. The honest version of the decision is therefore: I am buying volume with a strong structural signal and no causal read, for a defined period, and I have written down in advance what I will not learn. > Bezos frames this in *Invent and Wander* as one-way and two-way doors: most decisions are reversible, and the mistake is treating them as though they were not. "If it's wrong, it's wrong. You'll change it." A daily budget is about as two-way as a door gets. That framing is what makes the call reasonable rather than reckless. If the cost per lead — which I count by hand, outside the platform, from the client's own inbox — stays under a threshold I set before I made the change, the scale-up holds. If it crosses it, the budgets go back. I am not betting the quarter on a hunch; I am opening a door I can walk back through in ninety seconds, and I have said out loud what would send me back through it. ### The rule I am taking from this Scaling with broken attribution is allowed, on three conditions. You need a non-conversion reason that is structural rather than hopeful. "Lost impression share to budget concentrated on the highest-demand days" is structural. "It feels like it's working" is not. Everything downstream depends on that distinction. You need a hand-counted outcome metric the platform cannot see. In this account that is the client's own inbox, sixteen leads a week, counted by me. It is unglamorous and it is the only trustworthy number in the entire arrangement. If I did not have it I would have no business increasing anything. And you need the abort condition written down before the spend goes up, with a date attached. I check next Monday, comparing like for like. I have learned the hard way that [a number moving is not the same as a number meaning something](https://www.livain.com/blog/test-whether-a-dip-is-a-signal-before-hunting-for-a-bug/), and that a threshold invented after the results arrive is not a threshold at all. What I am not doing is waiting for the tracking to be fixed first. The season for this product ends in autumn; the developer tickets are three weeks old; the auctions are happening this weekend whether or not the conversion snippet ever gets its transaction ID. Perfect measurement that arrives in November is worth less than imperfect measurement plus a hand count in September. But the gap does not get to be invisible either. I have written before about what happens when a measurement hole goes unnamed — [nineteen days of spend with the tag gone from the page](https://www.livain.com/blog/gtm-snippet-removed-from-page-nineteen-days-no-measurement/), or a [ratio quietly broken by a filter](https://www.livain.com/blog/channel-filter-broke-roas-attributing-ad-spend/). In both cases the damage came from the reporting continuing to look normal. Here, the report says 1 conversion against 16 leads, in the note, above the budget table, where the client reads it. Scale when you can see the constraint. Say plainly what you won't be able to prove. Put a date on going back. ### Sources & further reading **External** [Get impression share data — Google Ads Help](https://support.google.com/google-ads/answer/7103314?hl=en&ref=livain.com) [Diagnosing impression share loss due to budget — Adalysis](https://adalysis.com/blog/ppc-kpi-monitoring-how-to-diagnose-and-work-with-impression-share-loss-due-to-budget/?ref=livain.com) [About conversion windows — Google Ads Help](https://support.google.com/google-ads/answer/3123169?hl=en&ref=livain.com) **Related posts** [Nineteen days of ad spend with no measurement](https://www.livain.com/blog/gtm-snippet-removed-from-page-nineteen-days-no-measurement/) [The channel filter on my dashboard silently broke ROAS](https://www.livain.com/blog/channel-filter-broke-roas-attributing-ad-spend/) [How to test whether a dip is a signal](https://www.livain.com/blog/test-whether-a-dip-is-a-signal-before-hunting-for-a-bug/) [Three ways in, none of them a funnel](https://www.livain.com/blog/three-ways-in-none-of-them-a-funnel/)