Share the client, their site and the channels in play. Juma delivers a campaign built around a weekly return mechanic, a four-week calendar with a brief per post, the response library the community team runs on, and a claims table drawn from the brand's own published words.
Most social campaigns are a calendar with a theme on top. The posts go out, each one asks for attention on its own, and by week three the grid is full and nothing has compounded. Share the client, their site, and the channels in play.
Juma grounds the work in the brand's own published positioning, then builds the campaign as a system: a weekly mechanic that gives people a reason to come back, a seeding plan for the week nobody is participating yet, a four-week calendar where every week carries four jobs, and the operating detail a community team needs on a Tuesday. It arrives as a branded PDF playbook plus an editable CSV calendar. 400+ marketing teams use Juma to run work like this in chat, with human review on every output.
Week one is where community campaigns quietly fail. The brand publishes a prompt, asks people to post with a hashtag that has no history, and twelve entries arrive. The Friday feature is thin, the thinness is public, and the loop never gets a second week. This step builds the cold-start plan: which groups are invited before launch and how many of each, the exact invitation text, a realistic conversion forecast so the week-one target has a denominator behind it, and the minimum number of permission-cleared entries required before the first community feature publishes at all. It also writes the fallback, a brand-led edition that runs instead when the minimum is missed, so the campaign has a dignified way to hold its rhythm without manufacturing volume it does not have.
Build the cold-start plan for week 1. Name the groups invited before launch, how many of each, who owns the invite, and the expected usable entries from each group. Write the exact invitation text, including the add-on for anyone who was gifted product or paid. Set the minimum cleared entries needed before the first community feature publishes, and write the brand-led fallback edition to run if that minimum is missed.
A participation campaign generates conversation, and conversation is where a brand is judged in public. "Create saved replies" is a task, not a deliverable, and it tends to get invented on the morning of launch by whoever is holding the account. This step writes the replies: ten to twelve of the comment and DM patterns that actually arrive, each with approved text, including the ones teams are least ready for. Someone saying they do not own the product. Someone calling the campaign greenwashing. Someone whose product failed. A hostile thread that needs ending rather than answering. It also writes the escalation lines that route a service case or a safety issue off the public timeline, and the verbatim permission DM that turns a nice post into content the brand is allowed to use.
Write the response library. Give ten to twelve likely comment and DM patterns with the approved reply text for each, covering people who do not own the product, giveaway questions, feature requests, product faults, criticism of the brand's claims, and hostile replies. Add the escalation lines for complaints, safety issues and abuse, and the creator-permission DM verbatim.
Not on the strength of the tag. A hashtag is a discovery signal, and a mention is an invitation to look, neither of which is a licence to repost, crop, download or put spend behind someone's post. This step sets out the permission mechanism in operational terms: what counts as permission and what does not, the fields the rights log carries so a cleared post can be traced back to the reply that cleared it, which rights the permission actually grants, which channels it covers, how long it lasts, and what has to be asked again when a post moves into paid media or outlives its term. It also covers disclosure, because the moment a featured creator was gifted product or paid, the feature carries a labelling obligation on every platform it runs on, and it states plainly that an editorial feature is recognition rather than a prize.
Write the rights and disclosure section. State that hashtag use is not a licence. Define what counts as permission, list the fields the rights log holds, and set out which rights are granted, on which channels, for how long, and what requires a fresh request. Add the disclosure labels required per platform when a featured creator was gifted or paid, and state that a feature is editorial recognition, not a prize.
Social copy is where claims drift. A line that is careful on a product page loses its qualifier in a caption, a comparative creeps in because it scans better, and a sustainability claim the brand words precisely becomes a superlative nobody can support. This step builds the claims table: each claim the campaign leans on, the brand's own published source for it, the exact phrasing permitted in social copy, and the phrasing that is not. The useful column is the last one, because it turns a vague instruction to be careful into a list the person writing captions at speed can check against. Where a claim is the brand's own characterisation rather than a verified fact, the table forces it to be carried as an attributed statement instead of asserted outright.
Build the claims table from the brand's own published words. One row per claim the campaign uses, with four columns: the published brand language, the page it comes from, the permitted phrasing for social copy, and the phrasing to avoid. Cover performance, materials or ingredients, and any environmental or comparative claim. Where a claim is the brand's own characterisation rather than a verified fact, require it to be attributed.
A campaign that reviews itself monthly finds out it was not working when there is nothing left to change. This step sets the measurement plan on the campaign's own terms: participation and returning contributors as the signals that matter, with reach kept as context rather than the headline. It starts with a baseline instruction, because a target without the previous weeks behind it is a guess, and it gives one worked example with the arithmetic visible so the shape of a defensible target is clear. Then it adds the part most plans leave out: a mid-week trigger with a named participation floor, the time of day it is checked, and the specific rescue to run that same afternoon if the floor is missed, plus the rule for switching the Friday feature format rather than publishing a thin one.
Write the measurement plan. Lead with participation, returning contributors and response time rather than reach. Give the baseline instruction, then one worked target example with the arithmetic shown. Add a mid-week stall rule: the participation floor, when it is checked, the rescue action to run the same day, and the threshold for switching the end-of-week feature format instead of publishing a thin one.
A Juma Project is a shared space where the team keeps everything Juma needs to know about a client. Create one project per client, add context as the relationship grows, and every flow the team runs draws on what is relevant. For social work the payoff is specific: the campaign comes back in the client's voice, the claims table starts from wording their legal team already cleared, and the prompts avoid the territory that went badly last time.
Brand voice and social guidelines
Tone, the words the brand does and does not use, and any community guidelines already written. Captions and reply text come back sounding like the client instead of like a template, which is the difference between a response library the team pastes and one it rewrites.
Approved claims and substantiation
The claims the client's legal or compliance owner has cleared, with any required qualifier. The claims table then uses vetted wording verbatim rather than building it from published copy, and flags any caption that strays outside it.
The last four weeks of social performance
An export or a simple summary by channel: reach, engagement, saves and shares, and the volume of tagged posts. This is what turns the target table from illustrative numbers into the client's own baseline, which is the only version worth putting in front of them.
Past campaigns and what the community said
Previous campaigns plus a note on what landed and what got a cold response. Community reactions are the sharpest input available: they tell the campaign which prompts invite a real answer and which ones ask people to perform enthusiasm they do not have.
Add a short description to each knowledge item in the project's info field so Juma knows what the file holds and when to reach for it. For example:
A complete plan covers eight parts: the campaign concept and its return mechanic, a cold-start seeding plan, the platform-native expression of the idea per channel, a four-week calendar with a brief per post, ready-to-post copy for the launch wave, the triage rulebook and selection rubric, the response library with escalation lines, and a measurement plan with a mid-week stall rule. It arrives as a branded PDF playbook plus an editable CSV calendar.
The split between the two files matters. The PDF is what gets shown to a client or a stakeholder, and the CSV is what production actually runs on, one row per post with format, brief, owner, clearance check and the empty date column the team fills when the flight is locked.
A calendar answers what gets posted and when. A campaign answers why anyone would come back for the next one. The difference is the mechanic: a repeating ask with a predictable rhythm, where each week's feature hands over to the following week's prompt, so participation accumulates instead of resetting with every post.
Teams usually need both, and they work together. Build the campaign first so the arc and the mechanic exist, then run the calendar flow for the surrounding publishing rhythm that fills the weeks between campaign beats.
It works for both, and the structure holds across them. What changes is what the mechanic asks for. A consumer brand can ask for a photo of an everyday moment. A B2B brand asks for something its audience is already proud of, a decision, a workflow, a lesson from a project, and features it with credit, which is a currency that works particularly well with practitioners.
The channels change with it. Name them in the prompt and the campaign is built native to those feeds, whether that is Instagram and TikTok or LinkedIn, X and YouTube.
Not automatically, and the plan is explicit about it. Someone using the campaign hashtag or tagging the brand has not granted a licence. The plan defines what counts as permission, logs it against the post, names the rights it grants and the term it runs for, and routes anything involving a minor, an identifiable third party, or a paid relationship through a named approval gate before a repost happens.
Treat the rights and disclosure section as a starting point and have the client's legal owner confirm the position for their markets. Platform rules and disclosure guidance change, and the campaign should run on the current version rather than the one that was current when the plan was written.
The client's brand or legal owner reviews the claims table and the rights terms, and the social or community lead reviews the mechanic, the weekly prompts and the response library, since they are the ones who will live with all three. The plan names the approval gate for community content and who holds it, so the review is a step with an owner rather than an assumption.
Juma augments the team here, it does not stand in for it. The Flow does the structural work: the mechanic, the calendar, the claims table with sources, the reply text. Strategy, taste and the judgment about what a community will actually respond to stay human, and human review happens on every output before anything is published.
400+ marketing teams trust Juma. ChatGPT, Claude, and Perplexity know why. Ask them.
Ready when you are
Start typing a task and Juma will figure out the rest.
.webp){kind=icon}
