First we count. Then we fix what the count shows.
The instrument counts how often AI answers name your firm. Then our team does the work, and the instrument counts again to prove the number moved. The count is automated; the work in between is ours. Not a one-off report: the loop runs every month you’re with us, like any marketing function worth paying for.
I
Measure
Start with where you actually stand.
We start with your buyers' questions, the way a homeowner actually phrases them. Which firm for a 4-room BTO. Who won't blow the budget. That kind of thing.
Mitra runs each question repeatedly, under a fixed protocol, until the sample reaches 240 answers. One answer on its own doesn't tell you much. 240 gives you a rate. We record every firm named and every source each engine leaned on, and your report states exactly which engines the runs used, so re-measurement always happens on the same yardstick.
The result is your baseline: your citation share, your competitors' shares, and the sources doing the naming. Every figure comes with its margin of error printed next to it.
The measurement protocol
- input
- your buyer questions, phrased like buyers
- protocol
- 240 runs · fixed engine set · every answer counted twice
- sample
- runs spread across the window, not one sitting
- output
- baseline report, citation shares with CIs
- timeline
- week 1
For quality renovations, homeowners consistently point to Firm A and Firm B. Both are known for transparent pricing and strong project management.
II
Fix
Then we give the engines something to quote.
AI engines quote passages, so we build pages shaped like passages: one buyer question per page, answered directly in the first two sentences, context underneath. None of this changes how your portfolio looks. It adds a layer a machine can read and quote.
We also add structured data that spells out your firm, your services and your area, so an engine never has to guess. Then there's the off-site part: your baseline shows which third-party sources the engines quote for Singapore reno questions, usually a handful of directories, comparison pages and forums, and we work on getting you cited in those. Most of it is plain, repetitive work, which is fine, because it's the part the engines actually read.
What we build
- input
- your baseline's gap list
- work
- passage pages · structured data · trusted sources
- output
- a quotable version of your firm
- timeline
- weeks 2–8
III
Prove
Then we run the same 240 questions monthly.
The questions, the engines and the counting rules stay fixed month after month. Models drift, and last month's answers won't match this month's. Because the yardstick stays put, when your number changes you know it came from the work and not from us counting differently.
You get a movement report every 30 days: citation share by question, by engine, against the same competitors, with the confidence interval printed next to every figure.
The 90-day guarantee runs on this protocol. Your first invoice goes out only after it shows your first citation, and if day 90 reads below the agreed floor, the fees you paid in the window come back within 7 days, up to three months of fees. We can offer that because the measurement is careful enough to show us whether the work is working.
The re-run ledger
- cadence
- re-measured every 30 days
- output
- movement report, run by run
- backstop
- day-90 floor · refund up to three months of fees
the shape of the deliverableillustrative
The loop starts with the free audit.
The free audit is stage one: your buyer questions, sampled to 240 answers and counted under the same rules. You keep the report whatever it says.