I Spent 7 Months Hiring One Geospatial Engineer in Singapore Before I Understood the Role: the 7-Step Method I Use Now

The requisition said geospatial engineer. It stayed open for seven months, attracted about thirty applications, and produced two candidates I would have hired, both of whom took other offers before we finished our process. The role was eventually filled by someone whose previous job title contained neither the word geospatial nor the letters GIS. That is the whole lesson, and everything below is me trying to make it reproducible rather than lucky.

Step 1: Split the 2 Jobs Hiding Under One Title

Spatial work contains two roles that share a vocabulary and almost no skills. Decide which one you are funding before you write a word.

• The spatial analyst. Answers questions with geography. Where should the next outlet go, which catchment does this address fall in, how does a coverage gap change under a scenario. Output is an answer, a map or a model, and the audience is a decision maker.
• The spatial data engineer. Builds the systems that make geography usable at volume. Ingests feeds arriving in mismatched coordinate reference systems, validates geometry, builds tiles and indexes, and keeps a service serving polygons under load. Output is a pipeline or an interface, and the audience is other software.

My original advert described the second and interviewed for the first, which is why two good candidates walked. Analysts were being asked database performance questions they had never needed, and the one engineer in the pipeline was asked to critique a choropleth. Everybody left the loop feeling the process was unserious, and they were right.

Then run the simple test: if your problem is a question, you need an analyst, and the work may well be finite. If spatial data arrives continuously, or accuracy has to be defended to a client or a regulator, or there is a map in front of users, you need the engineer and you need them permanently.

Step 2: Write the Brief Around Your Coordinate System and Your Accuracy

Most spatial job adverts are a list of software names. That tells a candidate nothing about whether the work is interesting or whether they are capable of it. Three facts do almost all the qualifying:

1. Which coordinate reference systems your data actually arrives in. If you do not know, that is itself a finding and probably the first thing the hire will fix. Data arriving in a local projection, a global one and whatever a vendor exported is the normal and difficult case.
2. The accuracy your product needs, in metres. Not “high accuracy”. A product that needs five metres and a product that needs half a metre are different engineering problems with different costs, and candidates can tell immediately which one you have.
3. Who is harmed when the accuracy is missed. A wrong catchment boundary on a dashboard is an annoyance. A wrong berth position or a wrong service boundary in a regulated filing is not. This single sentence sets the seniority of the role more reliably than years of experience.

When I rewrote the advert around those three facts, the application volume fell by roughly half and the proportion worth interviewing roughly tripled. The applications I lost were people who matched a tool list. Our notes on API design practice in Singapore cover the related habit of specifying an interface by its technology rather than by its contract.

Step 3: Source From the 3 Adjacent Domains, Not the Job Title

This is the step that ended my seven months. In Singapore the pool holding the job title is small, and everyone in it is being contacted by everyone. The pool holding the skills is far larger, and almost nobody is contacting it, because the skills sit inside domains rather than inside a software job family.

Three places to look first:

1. Surveying and land administration. People here live inside coordinate systems, datums, tolerances and error propagation as a daily discipline. The gap is usually software engineering practice, which is teachable, rather than spatial judgement, which is not.
2. Maritime, port and logistics operations. Routing, vessel and vehicle positions, geofencing, time and position joins at volume. This is where I found the hire. Our notes on hiring port and logistics engineers in Singapore cover the overlap in more detail, and building a logistics platform here covers the spatial components that tend to be underestimated.
3. Urban planning, utilities and environmental analysis. Strong on multi-layer analysis, data provenance and the politics of a disputed boundary. Often weaker on production systems, which makes them a better fit for the analyst half of step 1.

Search on the skill, not the title. In practice that means searching for coordinate systems, reprojection, spatial indexing and specific data formats rather than for a job title, and reading project descriptions rather than headlines. It is slower per profile and it multiplies the field.

Step 4: Screen With a Projection and Accuracy Exercise

This replaced a tooling interview and it is the highest value hour in the process. Hand the candidate two small datasets, deliberately in different coordinate reference systems, and one question that cannot be answered correctly without reprojecting one of them. Something concrete: how many of these points fall inside this polygon, and how far is the nearest one to the boundary.

Ask for three things back:

1. The answer. A candidate who combines the layers without reprojecting will produce a confident, wrong number. This is the most common real-world spatial failure and it is the fastest discriminator you will find.
2. The error margin. Not optional. A practitioner volunteers uncertainty; someone who has followed tutorials gives a single figure to three decimal places.
3. What they would refuse to conclude from the data. This is the senior signal. Strong spatial people will tell you the dataset is too coarse to support the question you asked, and they will be right, because I built the datasets that way.

Cap it at ninety minutes and pay for the time. Of the candidates I have run this on, the split is stark: most produce the unreprojected answer, a few reproject correctly and give a bare number, and a small group do all three. That last group has been worth interviewing every single time.

Step 5: Test the Production Half of the Role

If you concluded in step 1 that you need the engineer rather than the analyst, there is a second gap to check, and it is invisible in an analysis interview. Strong spatial analysts have frequently never operated a service. The work they have done ran in a notebook on a laptop against a dataset that fitted in memory, and it was correct.

Production spatial work is a different discipline: tiling strategy so a map renders at every zoom level, spatial indexing so a containment query against complex geometry returns in milliseconds rather than seconds, caching so a popular area does not recompute, and simplification so a polygon with tens of thousands of vertices does not arrive at a browser intact.

Ask three direct questions. What is the largest geometry volume you have served to a live interface. What did you do when a query got slow. What have you simplified, and what did you lose by doing it. The answers separate people who have shipped spatial software from people who have produced correct spatial analysis, and both are legitimate careers.

Step 6: Price Against Both Bands the Title Straddles

Spatial engineering sits in two salary markets at once in Singapore, and the choice has to be made explicitly. Benchmarked as a domain specialist, the band follows the sector the person came from, and in surveying, planning and utilities that band is frequently below software engineering. Benchmarked as a data or backend engineer who happens to work in geometry, the band follows your engineering band.

Both are defensible. What is not defensible is advertising against the domain band, interviewing against the engineering bar, and then concluding the market is short of people. I did that for a quarter.

So benchmark both from your own recent offers, decide which band you are competing in and write down why, then set the walk-away number before the first call and record which band it came from. Candidates who come from a domain background will often not know they can ask for the engineering band, and quietly paying them the lower one is both unfair and a retention problem that shows up in month nine. Our notes on improving offer acceptance rates in Singapore cover how band confusion shows up as late-stage declines.

The other local variable is the employment route, which determines who can accept your offer and how long the offer realistically stays open. If your shortlist includes candidates who would need a pass, the sequencing of the search changes, and our notes on preparing a developer employment pass application cover what to settle before the offer rather than after it.

Step 7: Define Accuracy Acceptance Before the Offer

Spatial outputs are uniquely arguable. Nobody disputes whether a payment reconciled. Everybody has an opinion about whether a boundary is in the right place, and in the absence of an agreed standard the loudest opinion wins. Your new hire will spend their first quarter defending outputs instead of shipping them.

Settle three things in writing before the offer goes out.

1. What counts as correct. An accuracy threshold in metres, per dataset, with the authoritative source each one is measured against. “Looks right on the map” is how this goes wrong.
2. How it will be measured. A repeatable check someone other than the hire can run. If only the new hire can verify their own work, you have no quality signal at all.
3. Who signs it off. One named person. Spatial disputes expand to fill every available stakeholder, and a single owner is the only thing that stops it.

Share all three with the candidate before they sign. In my experience good spatial people visibly relax when they see an accuracy standard, because it is the thing they have been missing in previous jobs, and the ones who argue with your threshold usually have a better one to propose.

The 3 Mistakes That Cost Me Most

Advertising a title instead of a skill. Seven months. The people I eventually wanted were reachable the entire time under other job titles, and I never searched for them because I was searching for words.

Interviewing the analyst half for an engineering job. Two good candidates left the loop thinking we did not know what we wanted, which was accurate. Decide in step 1 and staff the panel to match.

Leaving accuracy undefined until after the hire. Our first spatial hire spent most of a quarter in meetings about whether a boundary was right, because no threshold existed and no single person could sign anything off. That is not a performance problem, it is a brief problem, and it is free to prevent.

FAQ: Hiring a Geospatial or GIS Engineer in Singapore