How to Measure a Site Acquisition Pipeline
Market depth, evidence and the backlog behind delivery
This research note explains how to distinguish a list of possible sites from a measurable pipeline, separate counted facts from assumptions, test a sourcing hypothesis against ground truth, and keep owner contact outside automated scoring.
A useful backlog is a dated supply of opportunities that have survived the checks that matter. The question is not whether a map can show many parcels. It is whether the evidence can tell us which ones deserve the next minute of human attention.
The question
A development business may know how many settlements it wants. That is a destination, not a supply map. To plan responsibly, it also needs to know how many opportunities exist, how long they take to move, where they fall out and which parts of the process are still guesses.
The first useful number is not “how many sites are on the list?” It is “how many opportunities survive the next check?”
- Market depth
- How many opportunities exist in the segment and geography being studied, under a stated definition.
- Pipeline depth
- How many qualified opportunities are live at each dated stage, relative to the delivery requirement.
- Evidence depth
- How much of each claim is counted, derived from counted inputs or still assumed.
The loop
Site acquisition sits inside a larger development loop. Delivery, finance, planning, feasibility and market conditions all affect the result. Acquisition is not the only constraint, but it is the part of the loop where a deliberate search process can change the supply of future work.
From site to settlement
- Sitea candidate becomes a commitment
- Pipelinedated supply
- Deliveryapproval, civil works, build
- Settlementcapital and market evidence
- Next searchthe loop learns
Read as: A site enters a pipeline, moves through delivery, settles, releases information and capital, then changes what the next acquisition must achieve.
The loop is why timing matters. A shortfall in site sourcing may not appear as a shortfall in settlements for years. That lag makes an unmeasured backlog look healthy until the delivery programme reaches the point where it needs the missing sites.
Three pipeline questions
- Is there enough supply?
- The market may be too thin for the target. More searching cannot create opportunities that do not exist in the chosen segment.
- Does the process convert?
- A list may be large while the qualified subset is small. The remedy is better checking, sequencing or follow-up, not necessarily more records.
- Does it arrive in time?
- A channel can produce real opportunities too slowly to feed the delivery programme. Cycle time and time in stage are part of the result.
Why separate themThe three failures have different fixes. A study should diagnose them separately before recommending more data, more outreach or a change of geography.
The funnel
From record to decision
- Sourcea published record
- Identifya parcel and context
- Screenthe mandate fits
- Qualifythe evidence survives checking
- Decidea human chooses the next action
- Measurethe outcome returns to the method
Read as: Each step removes a different kind of uncertainty; the next step should not be treated as complete until its evidence is recorded.
The general principle is portable: every transition needs an observable event, a date and a reason for the change. Without those three things, a funnel is a drawing rather than a measurement system.
required live supply
= completion rate per year × average years in process
─────────────────────────────────────
conversion from identified to completed
Rate and duration must use the same time unit.
Conversion is measured from identified to completed.
The equation is a relationship, not a forecast by itself.No invented thresholdPass, review and fail bands belong to a defined pilot with a named market, sample and baseline. Until those exist, the equation describes the relationship but does not supply a forecast.
Count before you model
A metric is not a fact merely because it appears in a spreadsheet. The full research inventory contains public source families and candidate market metrics, but each metric is graded against the quality of the source that supports it.
Three states of evidence
- Measuredcounted or directly observed
- Derivedreproducible from measured inputs
- Assumeduseful placeholder, not a result
- Unknownkeep the blank visible
Read as: Measured values come directly from a named source; derived values are reproducible transformations of measured inputs; assumed values are placeholders awaiting evidence.
- Measured
- The source directly counts or records the thing being discussed, with a known reference period and coverage.
- Derived
- The value is calculated from named measured inputs and can be reproduced from them.
- Assumed
- The value is a planning placeholder because the source is incomplete, indirect, paid, unavailable or not yet checked.
- Unknown
- The honest value is not available yet. An em dash is better than a plausible default.
RuleA formula built from an assumed input remains assumed. More decimals do not make an assumption more empirical.
Build a point-in-time record
A point-in-time test can be quietly wrong if it uses a fact that was learned later than the opportunity it is supposed to evaluate. A dated copy of the source, the date the fact was true and the date it was learned are therefore part of the evidence, not administrative decoration.
- Source date
- When the publisher’s record or observation existed.
- As-of date
- The historical point at which the research asks what could have been known.
- Learned date
- When the research system acquired or processed the record.
- Licence
- What the source permits: internal analysis, public display, owner contact, derivative work or none of these.
- Provenance
- The chain from published material to canonical fact to model output.
The same rule applies to geography, planning records, valuation, rents and any other input. A public source is not automatically a public-use source, and a current value is not automatically the value that was knowable at an earlier date.
Use public data carefully
Public research can establish the shape of a market and test whether a record is available. It cannot, by itself, reveal a private person’s intentions. The public-safe method therefore separates parcel evidence from claims about an owner, and treats any contact decision as a separate ethical and legal step.
- Parcel evidence
- Planning status, geometry, approvals, dwelling context and other facts about the physical or regulatory setting.
- Owner inference
- A hypothesis about willingness, motivation or timing. It is not a fact simply because several public records appear nearby.
- Contact
- A human action governed by consent, privacy, licensing, duty and reputation. It is not an automatic consequence of a score.
- Human release
- A person reviews the evidence, rejects weak or inappropriate candidates and records why.
- Do not publish private names, addresses, financial distress, health or family circumstances as examples of a method.
- Do not infer a sensitive attribute when a parcel-side signal is sufficient for the research question.
- Do not use a source for direct contact unless its licence and the applicable privacy rules permit it.
- Do not present an inferred motivation to an owner as though it were known.
- Do not automate an approach before a human has checked the evidence and the contact protocol.
Test the method
A small research loop
- Questionwhat could change a decision
- Boundarywhat was knowable and permitted
- Runsame rules for every record
- Compareagainst ground truth or control
- Decidecontinue, revise or stop
Read as: Define a question, freeze the information boundary, run the method, compare it with known outcomes and decide what changes.
- Market test
- Estimate the reachable universe and its uncertainty before building a large acquisition system.
- Screen test
- Run a defined rule across the full available universe rather than selecting only attractive examples.
- Ground-truth test
- Compare the screen with parcels whose relevant outcome is actually known.
- Process test
- Measure whether evidence reaches a human decision in time and whether the result is recorded for learning.
A failed method is still useful if the failure is attributable. It may have poor precision, narrow coverage, too much delay, an unacceptable cost, or an ethical boundary that makes the proposed use inappropriate. Those are different findings and should not be collapsed into one score.
A public evidence trail
The privacy boundary follows official Australian guidance on personal-information handling and direct marketing. Any future performance result needs its own dated evidence record.
Source boundaryThese links support the privacy boundary applied to collection, profiling and contact. No channel-performance claim is made without a dated pilot.
Glossary
- Candidate
- A record that a rule has flagged but that has not survived the relevant checks.
- Qualified opportunity
- A candidate whose evidence, fit and next action have been checked by a human.
- Pipeline
- Qualified opportunities currently live, each with a dated stage and next action.
- Coverage
- Live qualified supply compared with the supply the delivery plan requires.
- Precision
- The share of flagged records that survive qualification.
- Recall
- The share of known relevant outcomes that the method finds.
- Cycle time
- Time from identification to the relevant completed outcome, measured by stage where possible.
- Ground truth
- A set of records whose relevant outcome is known well enough to test a method.
- Point-in-time
- The information boundary used to ask what could have been known at a historical moment.
- Provenance
- The source, date, licence and transformation history attached to a fact.
- As-of date
- The date against which a value or state is interpreted.
- Assumption
- A labelled placeholder used until a named source or experiment can replace it.
What this note does not claim
- It does not claim that a public data method can reveal a private person’s intentions.
- It does not claim that a market contains enough opportunities for a particular commercial target.
- It does not replace planning, title, feasibility, legal, privacy or investment advice.
- It does not report channel performance until a dated baseline and appropriate ground-truth set exist.
A public research article should make the method easier to inspect without making private people easier to target.
Next research step
The next defensible step is a bounded pilot: choose one clearly defined geography and product, freeze the information boundary, document the source permissions, define the outcome before looking at results, and test one transparent screen against known outcomes. The result should say what improved, what failed, what remains unknown and whether further work is justified.
- Proceed
- The method clears the pre-defined evidence, quality and ethics checks, and the remaining uncertainty is acceptable for the next bounded experiment.
- Revise
- The method is promising but a source, definition, sample or process needs repair before its result can be interpreted.
- Stop
- The method cannot meet the evidence, privacy, licensing, timing or value requirements. Stopping is a valid research outcome.
A method note, not an operating result. Any future performance claim requires a dated baseline, a defined ground-truth set and a disclosed comparison.