A signed agency and a folder of exports
The contract is signed and the agency wants to go live. What arrives is a zip file: a CRM export with properties, a spreadsheet of landlords kept by the lettings manager, a CSV of applicants with a column called notes 2, and a list of contractors that someone typed into Word. Addresses are in one field in one file and split across five in another. Some landlords appear three times with slightly different spellings.
Your onboarding person spends days in spreadsheets, writing formulas, asking the agency questions by email and waiting for answers. The agency's enthusiasm from the sales call fades while nothing visible happens, and your developers get asked to write one-off import scripts that are thrown away afterwards.
Why every import is a new puzzle
The mess is not a sign of a badly run agency. Agency data builds up over years, across software changes, staff changes and branches that each did things their own way. The problem on your side is that there is no repeatable path from whatever arrives to what your product needs.
- Import scripts are written per agency, so nothing learned from one onboarding helps the next.
- There is no agreed list of what must be right before go-live and what can be fixed later.
- Duplicates and conflicts are resolved by your team, even though only the agency knows which landlord record is the real one.
- Errors are found after go-live, when negotiators see wrong data in front of clients.
What slow onboarding costs a proptech startup
Revenue usually starts at go-live, so every week spent cleaning data is a week without it. Onboarding staff become the constraint on how many agencies you can sign, and they burn out doing repetitive work. And an agency whose first experience is wrong landlord details or duplicate applicants will judge your product on that, however good the rest of it is.
How we build a repeatable import pipeline
What we build is a pipeline that treats agency data as an input to be checked, not a file to be massaged by hand.
- An upload area where the agency or your team drops files, with column mapping saved per source, so the second export from the same CRM maps itself.
- Normalisation rules for the common problems: address parsing and postcode checks, phone and email formats, name casing, dates in mixed formats.
- Duplicate detection for landlords, vendors, applicants and properties, using name, email, phone and address together, with a confidence level on each suggested match.
- Validation against what your product needs: which fields are required for go-live, which are nice to have, and which records break a rule, such as a tenancy with no property.
- A review queue that the agency's own admin works through: confirm or reject each suggested merge, fill required gaps, and decide on records that conflict.
- A dry-run report before anything is written, and a single committed import with a record of what was created, merged and skipped.
| Issue found | Handled by |
|---|---|
| Postcode missing but address complete | Automatic lookup, marked for spot check |
| Same landlord three times | Suggested merge, agency confirms |
| Tenancy with no matching property | Review queue, agency decides |
| Date in an unreadable format | Flagged, original kept |
| Free-text notes column | Imported as a note, not parsed into fields |
We deliberately do not guess where guessing could put wrong information in front of a landlord or tenant. Uncertain records go to a person at the agency.
Onboarding the next agency
The next agency uploads its files. Most columns map automatically because another agency used the same CRM. The dry run shows what will import, what needs a decision and what is blocking go-live. The agency's lettings manager spends an afternoon on the review queue, because only they know which of the two Mrs Patels is the landlord of Flat 3. Your onboarding person checks the report, commits the import and moves on to training.
Checklist: is data holding up your go-lives?
- Developers write a new import script for most new agencies.
- Your team decides which duplicate records to merge on the agency's behalf.
- There is no written list of what data must be right before go-live.
- Agencies find data errors in their first week live.
- The number of agencies you can onboard at once depends on one or two people.