- Growth operations is the recurring work around campaigns: checking accounts, pacing budgets, protecting stock, watching creatives and competitors, and reporting to founders or clients.
- When each job runs on its own tool, agent or spreadsheet, definitions and refresh times drift, and the team spends its mornings reconciling instead of deciding.
- Running every job on one shared context keeps the numbers consistent, lets agents build on each other’s findings, and makes approvals and history visible in one place.
Ask a growth lead what their team does all day and the answer is rarely “launch campaigns”. Most of the time goes into operations: the morning check across ad accounts, budget pacing, making sure ads don’t run on products that are out of stock, spotting tired creatives, keeping an eye on competitor prices, and turning all of it into an update for the founder or the client.
Each of those jobs has attracted its own tool. There is a pacing sheet, a creative dashboard, a marketplace report, a Slack bot for alerts and, increasingly, an AI agent or two. Each keeps its own copy of the data, refreshed on its own schedule, with its own idea of what “ROAS” and “yesterday” mean.
Why separate tools and agents drift apart
Agents built by different people, without a shared model of the business, end up with conflicting definitions and relearn the business every time. Growth teams feel this every morning.
- Different definitions. The pacing sheet uses platform-reported revenue; the founder report uses store revenue; the bidding agent uses a seven-day click window. All three are called ROAS.
- Different refresh times. The alert bot ran at 6 AM, the dashboard refreshed at 8 AM and the marketplace export is from last night. They disagree because they are describing different moments.
- Different identities. One tool knows the product by its Shopify ID, another by its ASIN, a third by a campaign naming convention. Stock-aware decisions break at the seams.
- Rules copied into prompts. Each agent has its own copy of “don’t scale under a week of cover”, and they diverge the first time someone updates only one.
The cost is easy to state: without a governed definition, an agent computes the same number differently each time. Multiply that across five jobs and the team spends its first hour every day working out which number is right.
- Growth jobsEach reads the same context and writes back what it did
- Shared contextOne definition, one identity, one version of now
- SourcesAds, marketplaces, store, sheets and team chat
The operating day on one context
Here is what the same day looks like when every job runs on one shared context. The data is synced once, definitions are applied once, and each job hands its findings to the next.
Two details make this work. First, the jobs share findings. When the stock guard notices the hero SKU is down to four days of cover, the morning brief and the pacing job both see it, so nobody proposes scaling that campaign. Second, the weekly update uses the same numbers the team worked with all week, so the founder or client never sees a figure the team has not seen first.
What goes into the shared context
The shared context does not need to hold every row from every platform. It needs the facts the recurring jobs depend on, prepared once:
- Definitions. One version of each metric the team steers by, with its window, currency and the platform label it maps to.
- Identities. Products matched across the store, ad catalogues and marketplaces; campaigns matched to the products they promote.
- Current state. Spend and pacing, conversions with recent days kept up to date, stock and restock dates, competitor prices.
- Targets and budgets. What each campaign or product line is measured against, and who owns it.
- Rules and instructions. The playbook, plus live instructions from the founder or client with their end dates.
- History. What each job found, what was proposed, what was approved and what happened.
Everything else can stay in the source systems and be fetched when a job needs it. Keeping the shared layer focused is what keeps it accurate.
What changes for D2C teams
In a brand, the people who touch growth numbers usually include a founder, a growth lead, a media buyer and someone running marketplaces. Each looks at a different screen. A shared context gives them one set of facts, and a way to route decisions to the right person: a small bid change to the media buyer, a budget shift above a threshold to the growth lead, a pricing response to the founder.
It also joins channels that normally live apart. Quick-commerce and marketplace ads, the brand’s own store and paid social draw on the same inventory. When stock is shared, the decision to push one channel affects the others, and only a context that sees all of them can weigh that.
What changes for agencies
Agencies run the same jobs across many clients, and each client has different targets, rules and approvers. Keeping a separate, isolated context per client — its own connections, definitions, rules and history — lets the agency reuse the same workflows everywhere while each client’s decisions follow that client’s rules. Access stays scoped: a media buyer sees the accounts they manage, and one client’s data never informs another client’s agent.
There is also a compounding effect. Models become cheaper and more interchangeable every year, while the context an organisation builds keeps gaining value. For an agency, each client’s resolved product catalogue, rules and decision history get more useful every month, and they are the reason a new agent can be useful on day one.
Mistakes to avoid when consolidating
- Moving everything at once. Start with the two or three jobs that disagree most often, usually the morning check and pacing, and move the rest once those numbers match.
- Keeping a shadow copy. If one old spreadsheet keeps its own formula, it will drift again. Retire it or point it at the shared definitions.
- Leaving rules in prompts. Rules copied into each agent’s instructions diverge the first time someone edits one. Keep them in the shared context with an owner.
- Skipping history. Without a record of what each job found and what was decided, the weekly review falls back on memory, and agents cannot learn from past decisions.
How to judge whether it is working
The benefits are operational, so measure operational things:
- Time to a clear morning picture. How long from opening the laptop to knowing what needs a decision today.
- Problems caught early. Stock-outs, disapprovals and spend spikes flagged before they cost a full day.
- Hours of reconciliation saved. Time no longer spent working out why two reports disagree.
- Approval rate and reversals. The share of agent proposals the team accepts, and how often an approved change had to be undone.
- Coverage. How many accounts or clients one person can operate with the same care.
Meerkats is built to be this shared context for growth teams and agencies: one live picture of ad platforms, marketplaces, the store and team knowledge that every workflow and agent reads from, with approvals and run history in the same place.