You are not buying clicks. You are buying an auction position.

Your maximum bid is the most you are willing to pay. What you actually pay is usually less than that, and it is worked out from the quality of the ads around you. An advertiser bidding less than you can sit above you, and pay less than you, for the whole of a campaign.

Google scores each keyword on how relevant it looks to the query and the landing page behind it, and on how often people have clicked it before. That score multiplies your bid to produce the rank you get. It is the reason two accounts selling the same product can pay very different prices for the same word.

Click-through history feeds the score, and the score feeds position, and position feeds click-through history. An account left alone for a year is usually not losing money because the market moved. It is losing money because that loop has been turning the wrong way.

Broad match will find you buyers you would never have thought to list, and it will also sell you clicks from people who wanted something else with the same name. That is not a fault in the system, it is the setting doing what it was asked. The controls are the match types and a negative keyword list that is genuinely maintained.

It affects the score, so it affects the price, and it obviously decides whether the click turns into anything. Sending paid traffic to a home page is the most expensive habit in this discipline and it is still the most common one.

Smart bidding and Performance Max took the manual bid adjustments away and put the leverage somewhere else: in the targets you set, the conversion data you feed the system, the products you let it promote and the exclusions you apply. The account still needs a person. It needs them looking at different things.

Against the same month last year. The two moved together within four tenths of a percentage point, which is the first thing to look at in a down year: revenue falling roughly in step with spend means the account is still converting money at the rate it used to, and the shrinkage is coming from the market rather than from the management.

The sharpest fall in the set, and close to double the fall in revenue — so the orders that did come through were worth more each than they were a year ago. This is the number that most needs explaining in the report, which is exactly why it should not be the one left out of it.

Fewer conversions from a spend that fell less steeply, so each one cost more. Ten percent is a real deterioration and a survivable one. It is also the honest counterweight to the impression share figure further down, because both are true in the same month.

Essentially flat year on year, which in a year where conversions fell nearly a fifth is the result we would point to. It sits below the 8.81x we quote as our average across accounts, and it is on this page for that reason — an agency average that never has a month underneath it is an average of the months somebody chose to show you.

Impressions were down 8.03% and interactions down 3.65% over the same comparison. So fewer people searched, slightly more of those who did clicked, and fewer of the ones who clicked bought. That pattern reads as softer demand meeting harder competition rather than as something broken in the account.

The clearly good number and the one worth holding on to. In a weaker market, on 10% less money, we were in front of a materially larger share of the searches that existed than the year before. That is the position that pays when demand comes back.

The marketplace that dominates this category, appearing in the same auctions 53.00% of the time and finishing above us in 32.61% of them. Auction insights like these are the difference between "our conversion rate fell" and knowing why. You cannot outbid a marketplace and you can be very deliberate about which auctions you turn up to.

Search term analysis, negative keywords added from it, placement exclusions applied, and the sale campaign built and taken live. The exclusion work is the reason impression share rose while spend fell.

Raise the target return earlier and load the promotional copy earlier, so the campaigns finish learning before demand peaks rather than during it. Automated bidding needs a run of data at a new target before it is any good, and a sale gives it the least time exactly when it matters most. Getting promotion details before they go live on the site, rather than as they go live, is worth more to the result than any bid adjustment made during the week itself.