We sell outsourced enrichment, so treat what follows with the scepticism it deserves. We also turn work away roughly once a quarter. The honest answer to in-house vs outsourced data enrichment is that neither model wins outright. It depends on how specifiable your work is, how steady the volume is, and whether the knowledge involved is worth keeping inside the building. Here is the comparison we actually use when a Head of Product Data asks us which way to go.
What each model looks like in practice
In-house means named people on your payroll, working in your PIM, against your schema. Usually one to six of them, often reporting into ecommerce or product management. They know your ranges, your suppliers and your awkward categories.
Outsourced means a supplier takes a defined scope and returns finished records to an agreed specification. You keep the schema and the acceptance test. They keep the capacity, the tooling and the throughput risk.
Hybrid is where most catalogues over about 20,000 SKUs end up. A small internal team owns the schema, the rules and the top-value lines. A supplier takes the backlog, the long tail, the spikes and the languages.
Very few businesses run either model in its pure form for long. The question is not really which one. It is where you draw the line between them.
The comparison at a glance
| Factor | In-house | Outsourced |
|---|---|---|
| Time to first output | Months | Weeks |
| Cost shape | Fixed salary | Variable per unit |
| Throughput ceiling | Headcount bound | Elastic |
| Category expertise | Deep, yours | Bought in |
| Quality control | Informal | Contractual |
| Knowledge retention | Retained | Leaks |
| Management overhead | Line management | Vendor management |
| Handling volume spikes | Poor | Good |
| Multiple languages | Rare | Standard |
Every row has exceptions. The rest of this article covers them.
Where in-house genuinely wins
This is the section most consultancies skip. These are real cases, and we have advised clients into all five of them.
The knowledge cannot be written down
Some enrichment depends on knowledge that lives in people rather than in documents. Which of your own-brand ranges supersedes which. Why a certain fitment is technically valid but commercially forbidden. Which supplier’s stated dimensions are reliably wrong by five millimetres.
If your best product manager cannot write the rule down, no supplier can follow it. Outsourcing that work produces plausible output that is quietly wrong, which is the most expensive kind. Keep it in-house until the rules are written, then reconsider.
Enrichment is really a merchandising decision
In some categories, filling the attribute is the easy part. Deciding the product hierarchy, the variant structure and which attributes drive the filters is a commercial judgement about how you want to sell.
Furniture, apparel and configurable ranges sit here often. If the person populating the field is also deciding the range architecture, that person should work for you. A supplier can execute the decision. They should not be making it.
The volume is a steady trickle
Two hundred new lines a month, one language, one channel, stable categories. That is a comfortable half-time job for one person who already knows the catalogue.
Wrapping a supplier around it adds specification writing, purchase orders, acceptance testing and quarterly reviews. The overhead can exceed the work. Below a certain volume, outsourcing is a worse deal and we say so.
The copy is the brand
Where product copy is marketing rather than specification, the voice matters more than the throughput. Own-label ranges, lifestyle categories and anything where the description is doing persuasive work belong with a writer who lives inside the brand.
We can and do write to a tone-of-voice guide. It is still second best to a copywriter who sits in the same room as the buying team. Be honest about which of your categories are actually like this. In most distributor catalogues it is far fewer than the marketing team claims.
You already have the capacity
You may already have two experienced product data specialists with genuine spare capacity. If they also have working tools, hiring a supplier alongside them rarely pays. Use them. The reason to look outside is usually that you do not have this, not that in-house is inferior.
Where outsourcing genuinely wins
The work is a backlog, not a flow
A one-off migration of 80,000 legacy records is the clearest case. Hiring six people for a nine-month task means six redundancies at the end, or six people with nothing to do. This is the single most common reason clients come to us. We have written separately about how a product description backlog builds up.
Backlogs also have a deadline attached, usually a replatform or a marketplace launch. Internal teams miss those deadlines because the day job keeps interrupting them.
The volume spikes and dips
Seasonal ranges, acquisition integrations, supplier onboarding waves. Fixed headcount is the wrong instrument for variable demand. You either overpay in the quiet months or miss the peaks.
An outsourced model converts a fixed cost into a variable one. That is worth real money to a finance director, and it is often the argument that gets the budget signed.
You need multiple languages
Running German, French and Dutch enrichment in-house means either three hires or three agencies. Suppliers who already run multilingual teams absorb this without a step change in your cost base. Very few internal teams can justify a full-time Dutch product data specialist.
The work is specifiable and repeatable
If you can write the rule down, someone else can follow it. Extracting torque figures from datasheets, normalising units, mapping supplier categories, populating a defined attribute set from a defined source. These are exactly the jobs where an outside team with better tooling beats an internal generalist.
The test is simple. Write the specification for fifty products. If two people follow it and produce the same output, the work is outsourceable. If they do not, the problem is the specification, not the supplier.
You need a method, not just hands
Sometimes the real gap is process. Nobody has defined completeness per category, nobody knows the acceptance threshold, and there is no repeatable route from supplier file to live PDP. Adding internal headcount to that produces faster chaos.
Buying a supplier who brings a method, a tooling stack and a QA regime fixes the process alongside the backlog. That is what most of our product content enrichment engagements actually deliver. The first six weeks look more like consulting than production.
The costs each side hides
What in-house costs that nobody budgets
- Recruitment lead time. Three months to hire, two months to become useful. Your backlog grows throughout.
- Attrition. Attribute population is repetitive work. Good people leave within eighteen months and take the category knowledge with them.
- Management time. Someone senior spends a day a week supervising, arbitrating and checking.
- Tooling. Extraction, translation and image processing licences are not free, and they are priced for teams larger than yours.
- Opportunity cost. Your best product manager doing data entry is the most expensive labour in the building.
- No elasticity. When the volume doubles, the answer is a six-month hiring cycle.
Fully loaded, an in-house enrichment specialist costs £45,000 to £65,000 a year once you include salary, on-costs, tooling and supervision. Work out your own figure before you compare anything, because the salary alone is not the number.
What outsourcing costs that nobody budgets
- Specification writing. Somebody internal has to define completeness per category. This is real work and it lands on your busiest person.
- Acceptance testing. You need a sampling plan and time to run it every batch. Suppliers who do not ask about this are a warning sign.
- Rework cycles. The first two batches will be wrong in ways nobody predicted. Budget the time and agree who pays.
- Vendor management. Contracts, reviews, invoices and escalations take a couple of days a month.
- Context loss. Every new batch team re-learns your quirks unless the supplier keeps the same people on the account. Ask about that in the tender.
- Exit. If the relationship ends, does the rule set come back to you in a usable form? Get it in writing at the start.
How to decide
The in-house vs outsourced data enrichment choice comes down to six questions. Answer them honestly. Each answer that lands on the right is a point for outsourcing.
- Can you write the enrichment rules down so two people produce the same output?
- Is the volume lumpy, or is it a steady predictable trickle?
- Is there a deadline attached, such as a replatform or marketplace launch?
- Do you need more than one language?
- Could you hire and train the people inside the timescale you have?
- Is the knowledge involved a competitive asset, or is it just laborious?
Four or more points for outsourcing and the case is usually clear. Two or fewer and you should build the internal capability. Three is the hybrid, which is where most of our clients sit.
The hybrid, and where to draw the line
The line that works is by decision type, not by product.
Keep in-house: the attribute schema, the taxonomy, the definition of complete, the acceptance criteria, the top revenue lines, and anything requiring commercial judgement. That is the intellectual property and it should never leave.
Send outside: the long tail, the backlog, the spikes, the languages, the asset processing, and the repeatable extraction from supplier documents. Onboarding new supplier ranges fits here too, which is why supplier data onboarding is usually the first thing we take on.
The split that fails is the one drawn by category. Handing an entire category to a supplier means handing over the judgement calls with it. You get output nobody internally believes in, and it gets quietly redone.
If you outsource, buy it properly
Run a paid pilot on 200 to 500 SKUs from your hardest category, not your easiest. Set the acceptance criteria before the pilot starts and score the output yourself. Ask which named people will be on your account and whether they stay.
Get the accuracy threshold, the rework terms and the schema change rate into the contract. Then measure treated lines against untreated ones on conversion, returns and on-site search. That approach is set out in our piece on product content performance.
If you keep it in-house, do these three things
Write the specification anyway. The discipline of defining completeness per category is worth more than the headcount decision, and it is what makes the work reviewable.
Buy the tooling. An internal team without extraction and normalisation tools is a team doing copy and paste, and that is where attrition comes from. Our view on where the machine helps and where it does not is in manual, AI or hybrid product descriptions.
Measure throughput per person per week. Most teams have no idea what theirs is. Without it you cannot tell whether outsourcing is cheaper, and you cannot forecast when the backlog clears.
Key takeaways
- In-house vs outsourced data enrichment is not a general question. It depends on specifiability, volume shape, languages and deadline.
- In-house wins when the knowledge cannot be written down or the work is a merchandising decision. It also wins on a steady low trickle, and where the copy is the brand.
- Outsourcing wins on backlogs, spikes, multiple languages, specifiable repeatable work, and when you need a method rather than hands.
- Both models hide costs. In-house hides recruitment, attrition and supervision. Outsourcing hides specification, acceptance testing and rework.
- Most catalogues over 20,000 SKUs end up hybrid. Draw the line by decision type, never by category.
- Whichever way you go, write the specification and measure throughput. Neither model works without them.
We will tell you to keep it in-house when that is the right answer. Book a thirty minute discovery call and get a second opinion. You can see the shape of the outsourced model on our product content enrichment page. The catalogues we have done it on are in our work.