August in review, and where our data work stands

What we are confident about

January and February were the exception, not the baseline. Looking back two full years, site traffic ran between 18,700 and 31,700 visits a month for fifteen straight months. Then December hit 34,500, January 47,893 and February 57,527. August’s 30,960 sits above almost every month of 2024 and 2025. Against any normal baseline there is no decline. What ended was a two-month spike.

Year on year the business is well ahead. $66.9M against $44.7M, 2,276 orders against 1,169, and 123 online transactions in August against 79 last August. Order value is holding: the August average was $28,173, above July, and order counts fell evenly across every size band rather than through the loss of a few large buyers.

Metal prices cooled over the same stretch. Silver ran to $90.39/oz in January, fell to $58.36 by July and recovered to $65.45 in August. Bullion demand climbs with a rising price and cools when it stalls. How much of the change that explains we cannot prove. Eight months is a short series, and the relationship rests almost entirely on January and February. It is consistent with how this market behaves, which is not the same as measured, and the search data below complicates it further.

Public demand in August was flat year on year, and you sold more into it. Google’s index of how often Americans search “monster box” reads 18 for August 2026 and 18 for August 2025, the same number. Over those same two months your online transactions went from 80 to 123. On identical demand you converted half again as many buyers. This is the one benchmark in this brief that comes from outside every system we manage.

Order count is up on last August; the revenue comparison is composition. WooCommerce Analytics shows 144 orders this August against 121 a year ago, up 19%, on net sales of $4.45M against $15.62M. That is because August 2025 averaged $129,096 per order and this August averaged $28,173. A handful of unusually large orders made last August; more customers bought this August.

Demand cooled off the website too. Orders entered by your team over the phone and by wire, which never touch the site or any tracking, fell 59% across the same months. Whatever is happening in the measurement, some of this is real market movement.

Where the advertising is allowed to run, the returns clear your bar. Over the last three months every Shopping and Performance Max campaign is above the 100:1 this business needs. Catch All is running around 400:1 and brand search 240:1. The two strongest are capped at $10 and $40 a day. Non-brand search is the drag, and its “conversions” are inquiries rather than sales.

The “currency” product blocks are real and recurring. Merchant Center keeps reclassifying bullion items into a policy category that removes them from Shopping. We can clear them, and they come back. That is a standing maintenance job rather than an explanation for the year.

What concerns us, and is not yet settled

Fewer of the people arriving are buying. Measured on real orders rather than on tracking, the site converted 1.14% of visits in January and 0.32% in August, a 72% fall. That number comes from order records rather than from a pixel, so it is not an artifact of how sales are being counted. Taking out the orders your team enters by phone and wire changes almost nothing: those were already excluded, and putting them back gives a 68% fall instead of 72%.

Public demand hit a twenty-two year high in April, and April was your worst month. Search interest in “monster box” ran 29 in January and 64 in April, the highest reading in the whole record apart from a news spike in early 2019. Over those same months your sales fell from $19.1M to $3.1M. Interest kept climbing through May and June before dropping back to normal in July and August. So the market explains the summer, but it does not explain the first half of the year: demand was rising while sales were falling. That is the part we would want to keep digging on.

Product-level value has largely stopped flowing back. Only 7% of August’s Shopping campaign value is attached to a specific product, down from 20% in July. Across the account, the value being reported per sale has drifted far below what your orders actually average, which means the buying engine is optimizing against numbers that do not describe your catalog.

Our working concern is that a tracking change disconnected that feedback. Item-level cart data stopped reaching Google on January 10, which sits alongside the pixel and tag work done in the same week. If the high-value products are not reporting their value back, the engine is spending against values it cannot see properly. We have not proved this, and we are not going to claim it until we have.

What we cannot yet reconcile:

• Four tracking systems are running on the site at once, and we do not yet know whether they interfere with each other.
• Reported sales values do not match the order records. For August, GA4 reports $2.34M of product revenue against $1.76M of real website orders, about a third too high. Earlier in the year the gap is far larger, and one corrupt April order alone put a number into the data bigger than the whole year of real sales.
• The Search Console setup needs reviewing. The property was switched to a domain property when Answerbase came on, and there are now two properties reporting different histories. Until we know which one reflects the live site, we would not draw any conclusion from organic search.