$330 → $0.05
Cost to ask one everyday question 1,000 times — today versus with Tessallite.
Governed analytics, at near-zero cost
The same answer, for a fraction of the cost — on a terabyte.
Tessallite answers the questions your business repeats every day — in Excel, Power BI, the API and the agent — from a tiny maintained summary instead of re-reading the whole terabyte. On a full terabyte of industry-standard data (TPC-DS, 2.88 billion sales records), that turns hundreds of dollars per thousand runs into a few cents, with the identical answer.
A team running 40 dashboards refreshed hourly asks these questions ~350,000 times a year — about $116,000 in BigQuery scan today, about $18 with Tessallite. Same answers, every tool.
Measured results — SF1000 on BigQuery
The same answer, for a fraction of a cent.
Four everyday questions, asked the normal way (raw BigQuery) and with Tessallite, on a ~1 TB workload. Every accelerated answer was checked against the raw source and matched.
66 GB → a few KB
Data scanned to answer it — the same answer from a tiny maintained summary.
4 / 4 identical
Every accelerated answer matched the raw-source answer, within tolerance.
Cost to ask the same question 1,000 times
BigQuery on-demand cost (~$5/TB, 10 MB minimum per query) for three everyday questions — from a simple one to a heavy one — today (raw BigQuery, each time) versus with Tessallite.
Loading chart…
Data scanned to answer one question
Gigabytes scanned for the same three questions — raw BigQuery versus the maintained summary Tessallite reads instead.
Loading chart…
Route and correctness
With acceleration on, every question was answered from a maintained summary; with it off, from the raw source — and all four answers were identical.
Loading chart…
Methodology
How we ran it.
The benchmark measures a curated set of buyer-relevant analytical questions mapped to a semantic model, against direct BigQuery baselines and Tessallite's source and aggregate routes. Want to run it on your own data? Follow the TPC-DS evaluation guide.
Workload and source
The data source is a TPC-DS-derived SF1000 retail analytics dataset loaded into BigQuery. BigQuery remains the source execution engine; Tessallite provides governed modelling, protocol access, route selection, and transparent acceleration.
Run discipline
Published runs will identify cache policy, compute model, BigQuery region, query count, run count, warmup discard policy, Tessallite version, model version, and seeded acceleration state.
Correctness first
Each accelerated answer is compared back to the raw BigQuery result and must match within a stated tolerance before it counts.
Result cache, BI Engine, and BigQuery materialized views are real BigQuery capabilities. The benchmark will describe them accurately and position Tessallite as semantic-model-aware, workload-suggested, version-aware, and multi-protocol acceleration rather than as a replacement for BigQuery.
What we measured
Cost drivers, route behavior, correctness, and operating effort.
Every number on this page is measured on the live terabyte and tied back to the run records linked below.
Runtime metrics
- P50 and P95 elapsed latency by route.
- Bytes processed and cache-hit status.
- Slot-ms where available from BigQuery jobs.
- Variance across repeated uncached runs.
Route and acceleration metrics
- Direct BigQuery SQL baseline.
- Tessallite source route through the gateway.
- Aggregate route hits and selected aggregate IDs.
- Pocket route hits, freshness, and selected pocket IDs.
Human effort and serviceability
- Manual DDL lines and physical objects maintained.
- Time-to-first-accelerated-answer.
- Protocols served from one governed definition.
- Change failure mode when source or model shape moves.
Named sets, aggregates, and pockets
The benchmark story is governed acceleration, not isolated SQL timing.
The same governed question moves through source and aggregate routes while named sets keep business filters reusable across agent, Excel, BI, and API access — one definition, one answer, everywhere.
Source baseline
Curated questions execute against BigQuery through the governed source route, and we confirm the answer first.
Aggregate route
Recurring analytical patterns route to maintained aggregates seeded from the semantic model and reported with route explanations.
Pocket route
Exact executive dashboards, such as CFO weekly revenue, can route to pockets with freshness metadata and lineage visible.
Named sets
Saved business lists, such as high-value customers or strategic categories, are governed objects rather than copied spreadsheet filters.
The numbers in full
Dig into every number.
The full report covers all 17 business questions, the method, and the cost translation. The raw run records and byte matrix are here too, so anyone can check the figures for themselves.
What's behind the numbers
- Raw run records with query ID, run ID, route, elapsed time, bytes processed, cache-hit flag, and BigQuery job reference.
- The correctness check: each accelerated answer matched the raw BigQuery answer within a stated tolerance.
- Aggregate and pocket identifiers, refresh time, route explanation, and freshness metadata.
- Chart-ready summaries for cost, data scanned, and route mix.
- Benchmark environment metadata: date, region, compute model, version, model version, cache mode, query count, and run count.
Evidence manifest pending.
Same answer, a fraction of the cost — proven on a terabyte.
See how governed acceleration cuts your BigQuery bill on the questions your teams ask every day.