Frequently Asked Questions

A: Five visible steps: find the public deposit behind the paper, load and sanity-check it, re-express the paper's core analyses independently, verify item-by-item against the published statistics, then report what reproduced and what didn't. The whole flow is on the How We Work page.

A: Small differences have three honest sources: the original analysis ran in different software (e.g. Mplus), standardisation details differ (e.g. coefficient scaling), and public deposits occasionally drift between versions. Across the audits so far, the direction and significance of every headline finding reproduced; effect sizes should be read as ranges, not single best numbers. Genuine differences are flagged in the report, never smoothed over.

A: That is informative, not a failure — reproducibility reports are a legitimate publication category, and a careful account of where and why results differ is a useful research output in itself. It is also exactly what the trial guarantee covers: if the trial deliverable is not useful to you, you do not pay.

A: Yes. Journals judge whether the statistics are correct, fully reported, and reproducible — not which software produced them; no reputable journal requires SPSS or any particular package. Standard analyses (t-tests, ANOVA, regression) return identical numbers across tools because they are the same formulas. The trend actually favours open-source: Science, Nature, PNAS and the leading psychology journals increasingly require code and data to be shared, which scripted pipelines satisfy by design. JASP's reporting guidelines were themselves published in Psychonomic Bulletin & Review (2021), and jamovi-based analyses appear in BMJ-group journals (2025). The software and version are stated in the methods, exactly as those papers do.

A: Possibly, with months of training — and an RA who can build and debug these pipelines reliably costs HK$15–30K per month. You pay a fixed fee once and keep everything: the pipeline, the scripts, and the documentation, runnable by your team after the engagement ends. You buy the outcome and the durability, not my time.

A: Yes — nothing is hidden. On a call I can walk you through any stage of the pipeline, and every demo report documents the key decisions in plain language. If you want the full implementation — every script, documented so your team can run it — the trial hands it all over, run on your own published data, in a matter of days. You judge the whole machine, not just the output.

A: Three levels, each with a job. The demos are public-data showcases — the machinery demonstrated once, so you can see it work before anything touches your data. The trial runs that same machinery on your already-published data, checked item-by-item against your own paper's numbers — a matter of days, a fixed HK$3,000–8,000, free if not useful. A full project is production work on your current or new data: robustness checks, iterations until the results are publication-defensible, full documentation and handover, milestone-based. The price difference is honest: the trial runs on data with an answer key — your paper; new data has no answer key, and that is where the weeks go.

A: All processing runs on your institution's environment or a dedicated server you control — never on my personal devices, and never used to train models. That baseline covers unpublished data, student records, and ethics-scoped material, and it aligns with university data rules. Privacy is built in, not an extra.

A: Yes — two options. If you have an already-published dataset (OSF or journal supplementary files), the trial runs on that with zero confidentiality concerns. If not, the trial can run on your current project's data under the same privacy baseline; restricted data is handled with extra care (anonymisation, segregation), quoted transparently if it adds work.

A: One trial deliverable on your topic at a fixed HK$3,000–8,000 fee. If the result is not useful to you, you don't pay. You judge the quality before deciding anything else.

A: The data, the analysis, the scripts, and the documentation — everything your team needs to re-run the pipeline themselves. Invoiced as a fixed fee per deliverable, in a form suitable for grant claims.

A: Fixed-fee quotes per deliverable, tiered by scope: HK$6,000–15,000 for a small scoped piece, HK$20,000–40,000 for a full pipeline or analysis package, and HK$40,000–80,000 for grant-scale work (milestone-based, so you see progress before the full amount commits). All fit the kind of line item a GRF or FDS grant carries.

A: Replies within 24 hours on business days, same-day for anything urgent — written into every agreement. Delivery timelines are set in the scope of work; the trial itself is typically a matter of days, not weeks.

A: Every question about the delivered result is answered as part of the engagement — understanding the result is part of receiving it, and adjustments within the agreed scope are included. If you want to go further — new analyses, different model variants, another cut of the data — those are small, separately scoped pieces (typically a few thousand dollars, grant-claimable like the rest). No open-ended hourly meter: the scope and price of any follow-up is agreed before it starts.

A: Yes. After one or two deliverables, most academics add a research retainer — an "AI-accelerated RA": you choose the monthly hours (typically 10–19) at a fixed rate, so the fee is predictable (typically HK$10,000–19,000/month). It covers upkeep of the systems we've built plus routine analysis and summaries on them. The month's scope is agreed in advance, so there's never an open-ended meter. You can run it month-to-month, or lock the rate for 3 or 6 months if you'd rather fix the cost line for a whole grant period. If you'd rather keep everything in-house after handover, the documentation — and a lighter support plan if wanted — covers that instead.

A: No — the trial and every deliverable are one-off purchases. There is no subscription, nothing auto-renews into a retainer, and you keep everything (the data, the scripts, the documentation) whether or not you ever buy anything else. Ongoing support is a separate, optional decision you make only after you've seen the work — month-to-month, cancel anytime with notice. And if you do have ongoing needs, the retainer replaces separate per-piece quotes at a predictable monthly fee: it's the cheaper option for recurring work, not an extra layer of cost. If you don't need it, you simply never hear about it again.

A: No. If you'd rather not talk to anyone yet, you can get an indicative quote from the assistant on the Find Your Case page (it asks a few questions and emails a summary to me — I review it before anything is quoted), or start the trial directly by email: send your published paper or data link and you'll get a comparison table back (HK$3,000–8,000, free if not useful). A 15–30 minute call is optional, not required — it's there if you want to watch a demo run or discuss a larger project.

A: Yes — one monthly plan covers all your projects. Each project is set up as a named workstream with its own hour budget, and the monthly report itemises hours per project, so each grant can claim its own share of the fee from its own line.

A: Unused hours carry forward to the next month, then expire — so nothing you pay for is ever lost, and hours can't pile up into a debt. The month-end report shows carry-in, hours used, and carry-out, so the position is always visible (and grant-claimable).

A: An hour is an hour of my active working time on your requests — scoping, running the pipeline, verifying the results, writing the summary. It's not how long a person would take by hand (that comparison is about the value, not the clock), and it's not AI runtime either. Every task is estimated before it starts, with the breakdown shown to you, so you always know the cost before I begin.

A: Yes — grant applications, ethics/IRB paperwork, reference letters, meeting minutes, and inbox triage. A two-week free pilot on one inbox and one meeting stream, then an AI admin RA on retainer (HK$5–8K/mo — versus HK$15–30K for a real RA), with human confirmation on every action. Funded from dean's discretionary, department, or TDG budgets — not grant lines.

A: Yes — send one lecture deck and I'll rebuild it: gap analysis against the current field, a rewrite that keeps your voice, and redesigned slides with speaker notes throughout. One-deck trial, free if not useful; per-deck HK$2–5K or a per-course package, funded from TDG or department teaching budgets.

A: The demos are the track record. Nineteen replication audits of published studies (PLOS ONE and others), each with a per-paper comparison table you can check yourself in 15–30 minutes. Background: PhD in psycholinguistics (UNSW), BA in linguistics (CUHK), former EdUHK researcher — including co-developing a language-learning app with the Dean at EdUHK. I don't claim history I can't show.

A: I work fully remotely on HK hours — a video kickoff in the first week, replies within 24 hours (same-day for urgent). That reliability is why pricing is fixed-fee rather than hourly. If an on-site visit is ever genuinely useful, it is quoted as a separate pre-agreed day rate — never assumed, never a surprise.

A: No — and there's nothing to install or subscribe to either. I operate the pipelines; you receive the results and a report you can judge directly. If you later want your team to run the pipeline themselves, the documentation makes that straightforward — or I can simply keep running it for you under a support arrangement. Either way, the technical work stays on my side.

A: Both, at different levels. Every deliverable comes with a technical report — what the numbers show, significance, effect sizes, and the caveats. That is my job, and it is written for you to judge. The scientific interpretation — what the findings mean for your theory and your next paper — is yours; that's your expertise. If you'd like drafting help for a results or discussion section, that is scoped as a separate piece.

A: Yes — and that's a common frustration: those tools are designed to start fresh every conversation. I set up a private AI assistant that remembers every conversation, correction, and decision — and acts on them. It gathers data from the web, runs analyses, drafts documents, builds spreadsheets, reads PDFs, and writes code; because it remembers, it does all of it in the context of your work, not from scratch. You choose the models it uses, and for sensitive material it can run entirely on your own server — nothing trains on your data, nothing leaves your control. The setup includes onboarding: a guided walkthrough and a plain-language guide. It begins with a two-week free pilot on one recurring task, after a live walkthrough of the setup I run my own work on.