A k-anonymity-protected buying intent index for the SA automotive market.

1. The data monopoly map

The SA automotive data market is not a competitive free-for-all. It is three monopolies layered on each other:

• NaTIS / eNaTIS — run by the Road Traffic Management Corporation. The sole national vehicle registration database. Not directly accessible to third parties.
• Lightstone Auto — ~2.8 million bank-recorded vehicle transactions. Produces AutoMSA, vehicle valuations, dealer pricing tools. Customers include every major dealer group, every insurer, and most banks.
• TransUnion Auto Information Solutions — Auto Vehicle Verifications + quarterly Vehicle Pricing Index + Mobility Insights Report. Refreshed daily from SAPS, NAAMSA, RMI, BASA.

These three organisations own everything that has already happened: registrations, valuations, sales prices, accident records, finance defaults. They do not own — and have shown no interest in building — the data layer covering what is about to happen:

• Which buyers are downloading brochures right now?
• Which models have rising search velocity in which provinces this week?
• Which financing thresholds are buyers hitting in their pre-approval calculators?
• Which signals indicate a buyer is 2-4 weeks from a decision?

This is the gap. None of the three incumbents publishes any of it. Visio Intent does.

2. The signal feed

Every signal that fires across the Visio Auto suite is k-anonymised, aggregated, and stored in vint_signals_aggregate:

period_start, period_end,
region        text,   -- 9 SA provinces
brand         text,   -- normalised brand name
segment       text,   -- hatch | sedan | bakkie | suv | luxury
signal_type   text,   -- one of 38
count         int,    -- count of signal occurrences
is_demo       boolean -- Honesty Protocol flag

Crucially, no individual record ever appears in this table. The minimum cell size is 5: any aggregation that would produce a cell with fewer than 5 records is hidden from the output, replaced with nulland a “k-anonymity suppressed” indicator. Even an attacker with access to the entire database cannot identify any individual buyer or any individual dealer's exclusive activity.

3. The Index calculation

The headline Visio Intent Index is a single number 0-100 representing aggregated buying intent across SA:

function intentIndex(rows: AggregateRow[]): number {
  const visible = rows.filter(r => r.count >= MIN_CELL); // k-anonymity
  const total = visible.reduce((s, r) => s + r.count, 0);
  return Math.min(100, Math.round(Math.log10(total + 1) * 30));
}

A deliberately simple log-scale normalisation, not a complex composite. We chose simplicity for two reasons:

1. Methodological transparency. Every chart on the public Index page links to /methodology. A reader can reproduce every number from the published constants.
2. Resistance to gaming. Complex composite scores are vulnerable to a single high-velocity signal type inflating the headline. A log-scale on total volume is hard to manipulate without detectable spikes in the underlying distribution.

The Index is published weekly (Monday 06:00 SAST) for the public free version. Subscribers receive daily aggregations (Growth tier) or real-time access (Enterprise tier).

4. Pricing and the marketing flywheel

Above this paid tier sits a free public Index page at /live that anyone can view. This is the marketing flywheel: it gets quoted in BusinessTech, Daily Maverick, Moneyweb every time Toyota's intent score spikes or Chery's overtakes GWM's. Each citation drives subscription enquiries from OEMs and banks who want the underlying drilldowns.

Target: 30 subscribers across the three tiers, blended ARPU R20K/month, generating R7.2M ARR.

5. The OEM use case

A SA OEM marketing team — say, Toyota SA — currently allocates roughly R500M-R1B/year to dealer co-op marketing across their network. The decisions about wherethis money goes are made on lagging indicators: last quarter's sales by region, last year's brand surveys. Toyota does not have a real-time view of which provinces are seeing rising Hilux intent vs falling Hilux intent this week.

Visio Intent's Enterprise tier (R50K/mo) gives Toyota that view:

• Hilux intent up 14% week-over-week in KwaZulu-Natal, down 8% in Gauteng
• Fortuner intent rising in the Eastern Cape, where Toyota has only 3 dealerships
• Corolla Cross intent being captured by Chery Tiggo 4 Pro searches, suggesting cross-shopping

These insights inform a R50K/month subscription paying for itself by reallocating R5-10M of co-op marketing per quarter. The ROI is approximately 20×.

6. The bank use case

WesBank currently buys leads from comparison sites at R200-R500 per qualified application. Lead quality varies wildly because comparison sites optimise for volume, not intent maturity.

Visio Intent's Growth tier (R15K/mo) gives WesBank a feed of which postcodes and brands are seeing the highest application-stage intent (signal: finance_app_started). WesBank can then bid more aggressively on Google Ads for those exact postcodes and brands, reducing cost-per-qualified-lead by an estimated 30-40%. At WesBank's January 2026 volume of 215,050 applications, even a 5% improvement justifies the subscription.

7. The Honesty Protocol in a data product

Selling data is the highest-trust business in any market. A single fabricated number, exposed publicly, kills the entire product. Visio Intent enforces honesty in three ways:

• Demo flag. Until the live signal feed is wired from the visio-workspace signals table into vint_signals_aggregate, every data point is flagged is_demo: true at the database level and surfaced in a banner on every page.
• K-anonymity at the database level, not the application level. The vint_signals_aggregate table itself enforces minimum cell size 5 via a CHECK constraint.
• Methodology page everywhere. Every chart, table, and API response includes a link to /methodology. A subscriber can reproduce any number we publish from the published constants.

8. Comparison to incumbents

We are not competing with Lightstone or TransUnion. They own the historical record. We own the forward-looking signal. The two layers are complementary: an OEM subscribing to both gets a complete picture.

9. Conclusion

Visio Intent is the missing fourth layer of the South African automotive data market. The first three layers (NaTIS, Lightstone, TransUnion) are mature, well-funded, and own the historical record. The fourth layer — real-time, k-anonymity-protected, methodologically transparent buying intent — does not exist anywhere in SA today. We build it as a side-effect of running the Visio Auto suite, and sell it back to the OEMs and banks that need it most.

References: Lightstone Auto product page (2.8M bank transactions); TransUnion Q4 2025 Mobility Insights; TransUnion SA Vehicle Pricing Index Q4 2025; NAAMSA December 2025 sales release; Eighty20 Credit Stress Report (k-anonymity model precedent); POPIA Act 4 of 2013, Section 18 (de-identification standards). Repository: github.com/Iamhamptom/visio-intent, commit bf5d950. Live deploy: visio-intent.vercel.app. Public Index at /live.

Cite as: Hampton, D. (2026). “Visio Intent: A K-Anonymity-Protected Buying Intent Index for the South African Automotive Market”. Visio Research Labs, VRL-AUTO-007.