Strategy
Part of Planning multilingual social marketing around self-described needs and reversible decisions
A six-part multilingual social marketing strategy framework for one England audience task
A six-part framework for one England task, worked through a council's Polish-language Facebook post: decision card, evidence, content chain, routes, gates, review.
Use this multilingual social marketing strategy framework for one reversible choice: whether to publish a capped target-language social post for a defined task in England. The cap is the most posts, spend or reach the owner approves in advance. One worked example runs through all six parts.
What to take away
- Name the task, area, population, channel and owner before naming a language.
- Label every evidence claim with its source, class and the inference it allows.
- Freeze the source claim and record whether the work is translation, localisation or transcreation.
- Compare routes on one source, one target version and one observation period.
- A failed gate means hold, whatever else looks good.
The worked example
This scenario is illustrative, not a case study. A London borough council wants to tell residents how to book a household recycling collection. Its service records show that some current users asked for Polish. The channel is the council's own Facebook page, and the owner is the waste-service content lead.
Polish is a sensible example because it is the most common main language other than English in England and Wales. In Census 2021 the figure is 1.1% of usual residents, or 612,000 people. That is context for the decision. It is not evidence that any resident of this borough wants a Polish post.
1. Write the decision card
Name the task, service, area, eligible population, channel and source-content owner. Name the language and locale only after participants or current service users have described their needs. Never derive it from a name, postcode, ethnicity, nationality or profile.
In the example, the card reads as follows. Task: book a recycling collection. Area: one London borough. Population: current users who asked for Polish. Channel: the council's Facebook page. Stop condition: the source claim changes or a gate owner records a hold.
Add two options, each with an owner, decision date, stop condition and recovery action. Here they are to post one controlled Polish version for a set period, or to keep the existing English-only route. Both must be reversible, or the card is a commitment, not a decision.
2. Label every evidence claim
Keep four columns: proposition, source, evidence class and permitted inference. Population context is not preference research. A reviewed translation is not proof of understanding. A platform event is not a service outcome, and an associated outcome does not show cause.
| Evidence class | Example in the borough case | Not permitted |
|---|---|---|
| Population context | Census figures for Polish speakers | A language preference |
| Target-text review | A Polish reviewer compared the post with the English source | Comprehension |
| Platform event | The page recorded views and clicks | A service outcome |
| Associated outcome | Bookings rose in the same weeks | Causation |
The ONS Census 2021 language bulletin shows why. English was a main language for 96.5% of North East residents and 78.4% of London residents. Those regional figures describe areas, not what one reader needs from a post.
For direct research, GOV.UK's guidance on finding participants says participants must be actual or likely users of the service. Pair that with audience research and disclose the limits of any volunteer or convenience sample.
3. Control the content chain
Freeze the English booking instructions under a version number. Record that the Polish work is localisation, not word-for-word translation, because the booking steps and the collection days must match the English page. Link the Polish post to its source version.
Assign a qualified Polish-speaking reviewer for the locale. Add a subject specialist for the collection rules and a separate reviewer for the rendered post, including image text and the alternative text for the image. Speaking Polish is not proof of professional competence, and nobody approves their own draft.
4. Compare routes on equal terms
Fix one comparison unit: the same source, the same Polish version, the same eligible audience and the same observation period. Do not rank routes by unlike platform counters. Leave cost fields blank until a dated quote states currency, VAT, unit, period, inclusions and exclusions.
The channel strategy decision here is between two routes on the same page.
| Field | Route A: organic post | Route B: paid promotion of the same post |
|---|---|---|
| Cost | Staff time only, from the quote or timesheet | Platform spend plus staff time, with VAT |
| Data flow | Page analytics, owner named | Added audience data, controller role recorded |
| Accessibility | Alternative text and readable captions | Same, plus ad preview |
| Exit | Delete or correct the post | Stop the promotion, then delete or correct |
Illustrative arithmetic: a quote of 1,000 pounds before VAT becomes 1,200 pounds with VAT at the standard 20% rate. Record both figures, because the buyer pays the total.
5. Set independent release gates
Give each gate its own evidence field and its own named reviewer. The UK regulations for England differ by topic, so the table names the source behind each row for the borough case.
| Gate | Source to check | What the check records |
|---|---|---|
| Data protection | UK GDPR; the ICO is the regulator | A written contract with any processor, covering the terms the UK GDPR lists |
| Direct marketing | PECR | Specific consent before marketing emails or texts go to individuals |
| Accessibility | WCAG 2.2 | Level AA as the minimum for the rendered post and its destination |
| Advertising and consumer claims | Digital Markets, Competition and Consumers Act 2024, section 225 | Unfair commercial practices are prohibited, so keep a file for each objective claim; ask the reviewer whether the CAP Code also applies |
| Equality | Equality Act 2010, section 149 | A public authority must have due regard to the need to eliminate discrimination, so record how the Polish post fits the council's duty |
| Intellectual property | Copyright law | Copyright arises automatically and there is no UK register, so record who owns the text, image and translation |
| Finance | VAT at the standard 20% rate | A dated quote with the VAT treatment and approver |
Platform approval, a supplier certificate or a signed contract does not settle the other decisions. A failed gate produces a hold even if another measure looks favourable. Some of the data pages are under review since the Data (Use and Access) Act, so a qualified reviewer must apply current law.
6. Predefine the review
For each measure, state the event, eligible population, numerator, denominator, period, timezone, source system, query version, exclusions, uncertainty and owner. In the example, one measure is bookings made from the Polish post's link. Agree it in the measurement and reporting plan.
Keep attribution methods separate from causal claims. A booking that followed the post is an association. At review, choose proceed, modify or hold. Proceed needs controlled language versions, accepted gates and interpretable evidence. Modify names one defect and the next test. Hold keeps the English-only route.
Before you act
- Write the decision card with task, channel, owner and stop condition.
- Label every evidence claim with its source and permitted inference.
- Assign a qualified native-level reviewer for the locale.
- Fix one comparison unit before ranking any route.
- Set independent gates for law, accessibility and finance.
- Predefine each measure's event, denominator, period and owner.
