Article summary
An incremental Zalo experiment for Vietnam beauty brands that separates platform state from skincare need and calculates ROI with controls, contribution margin and negative experience.
A Vietnam beauty brand cannot calculate Zalo number-checking ROI as “registered phones found multiplied by average order value.” Registration is a channel observation. Incremental value comes from whether better routing and content create additional contribution margin after checking, creative, support, discount, returns and complaint cost.
Separate two segmentation axes
Axis one is Zalo channel observation: registration, unknown, error and time. Axis two is beauty need: sensitive-skin care, hydration, ingredient questions, color tutorial or gifting, expressed through voluntary behavior and first-party data. Do not infer skin need from gender, avatar or nickname.
Select a task and hypothesis
| Hypothesis | Zalo task | What must be measured |
|---|---|---|
| Reduce wasted channel handling | Registration | Does support time fall? |
| Recent queue ordering helps | Activity | Do useful replies increase? |
| Profile coverage research | Gender and Age | Aggregate coverage and bias |
Calculate who may enter the experiment
Include only customers with traceable source, consistent purpose, valid channel permission, no opt-out and a real beauty use case. An AIPUSH TXT contains one phone per line and no order, skin concern, language or permission. The internal crosswalk joins the result to experiment_id.
Create a randomized control
Within the eligible set, randomize between Zalo-observation-assisted routing and the current process. Keep offer, language, support skill and time comparable. A customer with an explicit channel choice is not randomized; honor the choice and exclude it from the primary experiment.
Design content from need, not appearance
| Customer expression | Content direction | Do not infer |
|---|---|---|
| Ingredient question | Formula and use guidance | Age |
| Sensitivity feedback | Gentle-use support | Medical diagnosis |
| Shade selection | Lighting, swatch and return guidance | Permanent skin-tone label |
| Gift need | Budget and occasion | Buyer’s gender |
Use contribution margin for ROI
Incremental return is experiment-arm contribution margin minus control contribution margin scaled to the same base. Net ROI equals incremental return minus checking, data handling, creative/support increment and negative-event cost, divided by total incremental investment. Revenue without returns and fulfillment exaggerates value.
Put customer experience into cost
Count opt-outs, complaints, duplicate contact, mistranslation and human escalation. A complaint costs more than handling time; it can require suppression, investigation and trust repair. Stop at a safety threshold rather than waiting for the monthly ROI report.
Analyze heterogeneity without post-hoc hunting
Preregister source, product need and new/existing customer strata. Do not keep slicing age, gender or city after seeing results until a positive number appears. Report intervals and uncertainty for small samples.
Prove the incremental value of Activity
Zalo activity time and days may simply describe platform use. Compare support efficiency and qualified conversation before and after activity ranking. If they do not improve, retain Registration or return to customer choice rather than relabel activity as purchase intent.
Four possible decisions after a month
The brand may expand one source, keep Registration only, stop the profile task or return completely to the prior process. Record evidence, cost, limitation and review date for each. No significant increment is itself a useful finding.
What a beauty brand is really optimizing
The goal is useful, accurate and timely skincare and purchase help in a channel the customer prefers. Zalo number checking is one variable in channel operations; ROI comes from the complete experience, not more personal labels in a phone table.
