DEV Community

bot bot
bot bot

Posted on Originally published at forgemesh.io

MPPscan's All-Time Volume Fell 97% in Six Days. Volume Can't Do That.

Tonight MPPscan's headline "total volume" reads about $2,370. On August 19 the same counter read $94,874. We know because we log it every day at 12:45 UTC — and for six straight days the all-time total has gone down. That is not a market move. Cumulative totals only ever grow. When one shrinks 97.5% in a week, the story isn't the volume — it's the counter.

What our logs show

We snapshot MPPscan's server-rendered stats daily (the same numbers anyone sees on the site) and alert on ±5% moves between polls. The sequence:

Date "All-time" volume "All-time" transactions
Aug 19 $94,874 259,167
Aug 23 $40,149 186,830
Aug 24 $22,466 163,509
Aug 25 midday $8,039 148,048
Aug 25 evening ~$2,345 ~131,000

That last row followed a −48.6% move between two polls a few hours apart. Both counters have since resumed ticking upward from their new, much lower floors.

What it isn't: a collapse in real activity

Our own vantage point on the agent economy stayed flat through the whole window. The x402 catalog our crawler indexes three times a day held near 15,100 live listings (−0.8% day-over-day). Settlement cadence across our own fleet wallets didn't change. Nothing we operate or index looks like a market that just lost half its history.

Two explanations fit the shape

A rolling window wearing an all-time label. If the counter is actually "trailing N days," a burst of activity in late July would age out of the window day by day — producing exactly this staircase decay while the counter keeps ticking up from each new floor. The daily-step pattern in our logs fits this best.

An ongoing reindex or dedup. Explorers restate history when they find double-counted transfers or drop a data source. That usually lands as one big correction, not six consecutive daily cuts — but a rolling backfill could look like this too.

We genuinely don't know which it is — MPPscan doesn't publish a methodology page, and we'd welcome a clarification from the team. Either way, the practical consequence is the same: anyone who quoted "MPP volume" off that headline number this week was off by up to 40× depending on which day they looked.

The takeaway for agent-economy numbers

The agent-payment ecosystem is young enough that most of its "market data" comes from a handful of small dashboards, each self-reporting with unstated methodology. That's not a criticism — we run dashboards too — it's a reason to treat every headline stat as a claim, not a fact.

Our rules after this week:

  • Log the source daily, so restatements are visible.
  • Label every third-party number as self-reported (our homepage rail pulse already does).
  • Never cite a cumulative figure without knowing whether it can shrink.

If a number that can only go up goes down, that's not noise. That's the story.


We watch the agent-payment rails so you don't have to. The free agent-readiness scan checks whether agents can find, trust, and pay your server.

Top comments (0)