KENYA · GUIDE 0590 · EDITION 4.7
Data cleanup service in Kenya: Customers and Market Demand
The buyer and the buying decision
Businesses preparing data for software or reporting. The offer to test is: Clean and standardize a dataset without discarding unapproved records.
Clean a bounded dataset using agreed rules for names, dates, duplicates, missing values, and exceptions, preserving originals for comparison and approved recovery.
Separate interest from an accepted order
Process a copy of a small sample and report every rule.
The purpose of this test is to establish whether the buyer accepts the specified result, delivery arrangement and payment terms. An encouraging interview does not establish repeat demand. Ask what the buyer currently uses, what would make them change supplier and who has authority to approve the first order. Record a refusal as carefully as an enquiry so that unsuccessful offers do not disappear from the evidence.
Collect evidence specific to this offer
Keep original copies, transformation rules, changed-record counts, ambiguous cases, owner decisions, validation findings, and the delivered mapping between old and cleaned values.
| Question to resolve | Evidence to retain |
|---|---|
| What exactly will the buyer accept? | Have the owner review ambiguous matches and sample corrected records, checking that legitimate distinctions survive rather than treating every similar entry as duplication. |
| What is included in the first trial? | Clean a bounded dataset using agreed rules for names, dates, duplicates, missing values, and exceptions, preserving originals for comparison and approved recovery. |
| How will a prospect see the offer? | Show a fictional before-and-after sample with reversible decisions. |
| What must happen before repeating the trial? | Original records remain recoverable and uncertain matches are flagged. |
Decide whether to continue
Review the specification with the buyer before committing to tools or stock. A trial is commercially useful only if actual costs, rejected work, time spent and cash collected can be reconciled. Cleanup damages the dataset when two different customers sharing a name are merged or an empty field is replaced with an invented assumption.
If a buyer changes the required output, record a new scope and re-cost the test. Do not count a request for a different product as proof of demand for the original one. Keep personal contact details in your private records and use customer statements publicly only with permission.
Turn the example into a buyer test
Fictional cleanup CLN-90 standardises 500 contact rows while retaining the original file and a row-level change log. Allowed transformations are trim outer whitespace, normalise approved category spelling, preserve phone/code fields as text and flag candidate duplicates. Fixture: ' Asha ' becomes' Asha'; phone'0712345678' keeps its initial 0; ' Retail ' maps to approved' Retail'. Shared reception emails and similar names are not merged automatically. Output is cleaned copy, unresolved list and before/after counts. Marketing, identity investigation and deletion of originals are excluded.
| Buyer discussion | Evidence to retain |
|---|---|
| Specification to confirm | A person' s preferred name is not changed merely to match a guessed naming convention. |
| Output to inspect | No record vanishes without a documented rule/approval; reopen export through intended import route. |
| Failure that can defeat acceptance | Phone loses initial 0 |
Ask the buyer to describe one acceptable and one unacceptable result. Record the decision maker, price basis, quantity, handover point and payment date privately. A sample request, an encouraging conversation and an accepted order are different evidence. Keep the first commitment within your demonstrated capacity; no customer interview or successful trial is asserted here.
What a failed test should change
Check data type/import conversion.. If the offer cannot meet that check, revise the specification or process before promising a repeat. Keep rejected offers and reasons in the same register as positive replies so that your record does not exaggerate demand.
AI-assisted planning reference. Examples are illustrative. Editorial method and evidence limits.