How to Choose a LinkedIn Marketing Agency for Your Manufacturing or Export Business

Be precise first about what is actually being bought, because four unrelated skills sit inside this one service: building a target list that survives LinkedIn's unreliable industry data, writing messages a technical buyer will answer, running paid formats that cost several times what the equivalent Google Search click costs, and the CRM integration that turns conversations into pipeline instead of one salesperson's private inbox. Very few vendors are strong at all four. The ones that claim to be are usually strong at whichever one they demo first.

So the first decision is not which agency. It is which of three problems you are paying someone to solve, because each one needs a different kind of supplier — and one of them is a funnel optimization problem that no amount of outreach will fix.

Decide which of three problems you are paying someone to solve

Write down which of these three sentences is true of your company this quarter. If two feel true, the earlier one is the real constraint.

• "Nobody in our target market has heard of us." You need visibility work: a company and personal presence that a buyer can check, proof assets in the buyer's language, and paid reinforcement aimed at a small named account set. This is a content-and-media engagement, and it is measured in acceptance rates and inbound recognition, not in meetings.
• "We are known, but nobody is talking to us." You need an outbound programme: list building, sequenced outreach from real profiles, and enough reply capacity to honour the conversations it starts. This is where most exporters actually are, and it is the engagement type this article is mostly about.
• "We have conversations and they go nowhere." This is usually not an agency problem at all. It is a proof problem (a brochure where a specification sheet should be), a destination problem (a message about one product line pointing at a homepage), or a sales-process problem. Hiring an outreach vendor here buys more conversations that die in the same place.

The mechanics of what a competent programme looks like once it is running — the target list, the five-touch sequence, the stage metrics — are in our guide to LinkedIn for B2B manufacturers and exporters. This article is about the step before that: how to tell which vendor can actually run it.

Whose profiles, whose network, whose inbox — settle this in writing

There is no way to run outreach at scale from a company page. Pages cannot send connection requests and nobody replies to a logo, so every serious LinkedIn engagement runs out of your own people's personal profiles. That single fact creates the commercial issues nobody covers in the pitch.

You are lending the agency your team's professional identity. Three things need to be answered in the contract, not in an email:

• Which profiles, and who holds the session. If the agency's tool operates inside your export manager's logged-in browser session, that person's account carries the risk of whatever the tool does. If the agency proposes messaging from its own consultants' profiles instead, ask what network those profiles have in your sector — usually the answer is a network of other agencies.
• Where the conversation history lives. Require a weekly export into your CRM, with the contact, company, stage and the actual message thread. A monthly PDF summary is not a record. Connections themselves stay with the profile, which means they stay with the person — so if one salesperson's account is the whole programme, your buyer network leaves when they do.
• What you keep when it ends. The account and target lists with their filter logic, the message library with per-variant results, and the full conversation export. All three are cheap to hand over and routinely withheld, because they are what lets the next vendor — or your own team — continue without starting from zero.

The asymmetry worth naming out loud: if the approach gets an account restricted, it is your salesperson's account, not the agency's, and a restricted account loses access to the network it spent a year building.

Ask which tool they use, and whether it operates inside your account

Almost every vendor in this space uses automation of some kind, and almost none volunteer which. Ask directly, and ask for the product name. LinkedIn's own User Agreement prohibits using bots or other automated methods to access the service, scrape data or copy profiles — so the question is not whether a tool carries risk, but how much of that risk lands on your people's accounts.

Three broad approaches, in descending order of risk to you:

• Browser-extension injectors that act inside a logged-in session, sending requests and messages as if the user did. Highest restriction exposure, and the exposure sits entirely on your team's profiles.
• Cloud automation platforms that log into the account from the vendor's infrastructure. Slightly further from your browser, but a login from a different country than the account's usual location is itself a signal.
• Manual sending against saved Sales Navigator lists and a template library. Slower, more expensive per touch, and the only approach with no tooling risk at all. For a 40-to-80-account programme it is also perfectly feasible — volume is not the point at this scale.

Two numbers make the risk concrete. Connection requests are capped at roughly 100 per account per week, and the platform tightens that further when acceptance rates fall. So a vendor offering 1,000 new connections a month from one profile is either describing several profiles they have not mentioned, or describing a restriction that has not happened yet.

If your target markets are in the EU, add the data question. Where is the contact list stored, was any of it enriched from a third-party contacts database, what is the lawful basis for processing it, and who is controller versus processor? Put it in a data processing agreement. "We have a database of 40 million verified contacts" is presented as an asset in pitches and is closer to a liability on your side of the table.

The language and technical-depth test to run before you sign

This is the single most useful thing you can do during a selection, and it costs one meeting. Pick one real target account in one real target market. Give the vendor the company name, the product line and access to nothing else. Ask for three drafted messages: the connection note, the first real message, and the follow-up.

What you are reading for:

• Did they target the title in the buyer's language? A German list built only on "Procurement Manager" misses Einkaufsleiter, and for a German-market programme the local-language title set is most of the list, not a refinement of it.
• Is there one checkable, specific fact? A certification your market requires, a tolerance, a stock policy, a delivery time. Anything a quality engineer could verify or challenge. If the draft could be sent by any exporter in your sector with the company name swapped, it will perform like one.
• Does the connection note stay under about 200 characters, with no pitch and no attachment? A catalogue in the first touch is the most reliable tell that the vendor has not done this in an industrial category.
• Does the first real message ask a question rather than request a meeting? Asking for 30 minutes before earning it is what produces the 2% reply rates that later get blamed on "LinkedIn not working for our industry".
• Who wrote them? Ask for the name of the person who will write the live campaign, and whether that is the same person in the room. The most common failure in this category is a senior strategist selling the work and a junior executing it against a template library.

Then ask what they would need from your engineers to write the real thing. A vendor planning to write industrial outreach without a technical intake session is planning to guess, and buyers in this sector can tell within one sentence.

Read the deliverable, not the promise

Almost every dispute in these engagements traces back to a deliverable defined loosely enough that both sides read it differently. The wording matters more than the number attached to it.

Whatever the fee structure, the deliverable definition drives behaviour more than the price does. Per-meeting pricing reliably buys no-shows, because the vendor is paid for the booking and you absorb the empty slot. A flat retainer with agreed stage metrics is usually the honest structure for a category with a six-to-eighteen-month cycle, since nobody can be paid on orders that land three quarters after the work. If you do buy on qualified enquiries, write the definition — role, company size, market, stated need — and keep a rejection right with a monthly cap, so the definition has teeth without becoming a monthly argument.

Buy a paid 30-day proof, not a 12-month retainer

Almost every vendor will propose a six or twelve-month minimum on the grounds that the cycle is long. The cycle is long, but the question a pilot answers is not "did we get orders" — it is "can this vendor operate our profiles competently in our category". That takes four weeks.

• Scope: one market, 40 to 60 target accounts, two of your profiles, four weeks, a fixed fee you are willing to lose.
• Measured: connection acceptance rate, reply rate to the first real message, and conversations that reach a technical call or sample request. Agree the pass marks before it starts — something like 25% acceptance, 10% replies, and three to five conversations a salesperson would call real.
• Handover regardless of outcome: the account list with filters, every message variant with its numbers, and the conversation export. Written in, not implied.
• Explicitly not measured: orders, and anything involving follower growth.

A vendor who refuses a paid pilot outright is worth one follow-up question: is the objection the ramp — genuine, if their process front-loads research — or the scrutiny? If it is the ramp, agree a six-week pilot instead. If the answer is that their process only works over twelve months, they are describing a process that cannot be inspected, which is the whole problem with this category.

Reporting has to show what happened downstream

Ask for a sample monthly report from an existing client with the names removed. Four things should be in it: the stage numbers, per-variant message performance, the account list actually worked that month, and named enquiries with their current status in your system. Follower growth, impressions and Social Selling Index scores are not pipeline metrics, and their presence in a report usually signals the absence of the others.

The discipline transfers across verticals. In the university dental hospital programme we ran in Istanbul, roughly 6,000 leads became about 3,000 booked appointments at 6x ROAS — and the only reason those two numbers can be stated separately is that a CRM and a structured call-centre workflow recorded the handoff between them. A different industry, the same requirement. Your export equivalent is connection to conversation to sample or RFQ to order, and a vendor who cannot report past the first two stages is not measuring the part you care about. If you want an outside read on which stage is actually broken before you hire anyone, that is what a free growth audit is for.

Six questions, and the answers that should end the conversation

• "Which of my people's profiles will you message from, and who holds the session?" Vagueness here, or an answer that avoids naming the tool, is disqualifying.
• "Which software, and does it run inside a logged-in browser session?" A vendor who will not name the product is asking you to accept an unspecified risk on your own accounts.
• "Show me three target lists you built for a manufacturer, with the filter logic." Company names can be redacted; the method cannot be confidential, because the method is the thing you are buying.
• "Can I see three drafts for this real account, written by the person who will write the campaign?" "Our copywriters will handle that after onboarding" means no test before signature.
• "How is a qualified enquiry defined, and can I reject one?" No written definition, or no rejection right, means the reported number is the vendor's to decide.
• "If I stop after three months, what do I keep?" Anything less than lists, messages and conversation history is a lock-in clause dressed as a process.

Frequently asked questions

Should we hire an agency or an export salesperson instead? They solve different halves. An agency can buy you process, tooling and target-market language coverage faster than a hire can, but it cannot hold the technical conversation that follows a reply — and a four-person export office saturates at roughly 40 to 60 live conversations a month anyway. Sizing outreach above your own reply capacity is the most expensive mistake in this category, because an unanswered reply is worse than an unsent message.

Should the same agency run the paid campaigns? It helps, on one condition: that paid is reported against the same account set as the outreach rather than as a separate channel with its own dashboard. Sponsored content aimed at the same 40 to 80 accounts makes the outreach land warmer. A separate ads engagement optimising cost per form fill will quietly reward the cheapest, least useful contacts on the platform.

What should this realistically cost? For a single-market outreach programme across two profiles, monthly fees commonly sit in the low thousands of euros, with Sales Navigator seats and any ad budget on top rather than inside the fee — treat both extremes of the published range as questions to ask, not as bargains or premiums. The more useful comparison is not against other agencies: a European trade-fair stand runs roughly €25,000 to €60,000 once space, build, freight, travel and staff time are counted, and produces a comparable order of magnitude of serious conversations across three days.

How quickly should we expect to know whether it works? Stage metrics are readable in four to six weeks, qualified enquiries in roughly four to eight, and orders arrive on the buyer's tender calendar — commonly one to three quarters for a component or contract-manufacturing relationship. Any vendor promising orders in month one is describing a different business than yours.

The bottom line

Every agency in this category will show you the same capability list, so stop evaluating capability and evaluate three things you can actually inspect: who holds the risk on your accounts, whether they can write one message a technical buyer would answer, and what the deliverable says when read literally. Then buy four weeks of it before you buy a year.

If you would rather have the target-market mapping, the outreach system and the paid reinforcement built in that order — and reported as stage numbers instead of follower growth — that sequence is the core of our export and industrial growth work. The order matters more than the tooling, and it is the part most programmes skip.