MES or Video Analytics First? A Decision Guide for a Mid-Size Indian Factory

MES and camera analytics are not substitutes. An MES runs the plant's work: orders, routings, batch records, traceability. Camera analytics tells you where hours go on the floor as it is today. If a customer or regulator needs part-level or batch-level traceability, or your lines already talk over PLCs and your ERP data is clean, start with MES. If you can't yet say where yesterday's hours went, cameras usually answer that faster and cheaper, and the answer tells you what the MES should be built around.

By the Mama Editorial Team. Mama scopes camera and floor-visibility projects for Indian factories, so we have tried to be careful here: every claim about how MES or ERP projects go wrong comes from a named study or survey, and every rupee figure is either sourced or marked as an assumption.

Key takeaways

• They answer different questions. An MES records what was made, on which order, by which route, with which material. Cameras record when machines ran, stood or waited, and roughly why.
• MES first if you need traceability for an OEM, an export buyer or a regulator, if your machines are already networked, or if your ERP is live and trusted.
• Cameras first when the fleet is old and mixed, the ERP is used mainly for billing, and nobody can yet name the top three causes of lost hours.
• The documented failure modes in surveys and studies are cost, ERP integration, a shortage of in-house expertise and weak management ownership. Package choice ranks low.
• Year one (labelled estimate, 40-machine plant): an MES with machine connectivity can run to tens of lakhs. Camera analytics on CCTV you already have is closer to a few lakhs. The worked table shows the assumptions.

What each system actually is

The international standard for factory-level software, ISA-95, puts manufacturing operations management (the MES layer) between the business system (ERP) and the control system (PLCs, machine controls). An MES takes work orders down from ERP, sends recipes and instructions to the floor, and reports back what was produced, consumed, scrapped and inspected. It is the plant's system of record for production.

Camera analytics sits beside that stack rather than inside it. It reads the video streams of cameras on the floor and turns what it sees (a spindle turning, a line stopped, material missing at a station, a forklift queue) into time-stamped states. It doesn't know the work order number unless someone tells it. For a fuller primer, see what factory-floor video analytics is.

A simple way to hold the difference: MES knows what you made. Cameras show how the hours were spent making it.

What each one needs before it works

MES needs three things you may not have

1. Data-entry discipline. Any MES that is not fully machine-fed depends on people booking job start, job end, scrap and reasons at a terminal. If booking is skipped on a busy shift, the MES quietly fills with the same gaps your paper logsheet has. Our piece on manual vs automated production counting covers why self-reported stops undercount.
2. Machine connectivity. Feeding machine states automatically needs a PLC, a CNC with a data option such as FOCAS or MTConnect, or a retrofit gateway. A May 2026 guide from TeepTrak, a European vendor, puts a protocol gateway at €500–1,500 per machine before configuration, and external-sensor retrofits at €1,000–1,200 per machine. It is a vendor's figure, a sense of scale only. Older Indian fleets with mixed controls sit at the expensive end. We compare the options in tracking old CNCs without a PLC.
3. A working ERP. Work orders, item masters and routings have to come from somewhere. If the ERP is mainly a Tally-style billing system and routings live in a supervisor's head, the MES project turns into an ERP clean-up project first.

Cameras need fewer things, but specific ones

• Cameras that output a standard RTSP/ONVIF stream, or a recorder that does.
• Sightlines. Each machine or station needs a visible cue for "working", such as spindle motion, a cycle lamp or material flowing.
• A worker notice under India's data protection law, because people will appear in frame even when the analysis is about machines. Our DPDP worker CCTV notice guide has a template.

Cameras need nothing from the ERP and no terminal bookings, which is also why they can't give you traceability.

How MES rollouts go wrong: what the evidence says

• The obstacles Indian manufacturers report. In a Rockwell Automation survey of 1,560 decision-makers across 17 countries, reported by CXOToday in July 2026, Indian respondents named high total cost of ownership (56%), integration with ERP and other systems (49%), lack of internal MES expertise (47%) and cybersecurity or compliance concerns (45%) as their barriers. The same survey says 98% of Indian respondents have an MES somewhere but only 36% have one fully integrated. Treat these with care: an MES vendor ran the survey, and no India sample size or company-size split is given.
• What separates success from failure in smaller plants. A survey of 163 small and medium manufacturers in the International Journal of Production Research (Lee et al., 2012) found that MES success depended on top management support, sufficient investment, the system developer's capability and user participation.
• The Indian SME pattern from ERP. ERP is the closest Indian SME evidence we found, since peer-reviewed work on MES specifically in Indian SMEs is thin. A study in the Journal of Industrial Engineering and Management (Basu et al., 2012) ranked the issues affecting ERP implementation in Indian SMEs. Top management support came first, then clearly defined goals, project management and team competence. Package selection ranked near the bottom.
• The readiness gap. Researchers studying Industry 4.0 barriers note that "most SMEs in India lack the basic operational ability in terms of IT technology, real time monitoring of the processes" (Sahu and Gupta). A study of Indian manufacturing SMEs (Borana et al., Journal of Science and Technology Policy Management) ranked high investment, return on investment and multiskilled workforces as the most crucial barriers to digital transformation. A CII report prepared with KPMG, as summarised on CII's own blog, scores MSMEs at 2.4 out of 5 on Industry 4.0 maturity, against 2.9 for medium and 3.4 for large firms; we could not locate the report itself.

You will also see a much-quoted claim that "60–70% of ERP projects fail". Basu et al. repeat it, but they cite an online statistics page and a blog post, not a measured study. We don't rely on it, and you shouldn't either.

Taken together, the sources name ownership, investment, integration and skills far more often than software features.

Time to value

The sources above give no independent figure for how many months an MES takes to pay back, so we won't invent one. Our judgement from the structure of the work: an MES delivers value only after item masters, routings, terminals, connectivity and floor training are all in place, which is months of work, not weeks. Camera analytics on existing CCTV needs a stream, a placement check and a few days of baseline. The first useful picture of where hours leak can arrive in weeks. That picture is coarser than MES data and carries no order or batch identity.

A worked year-one ₹ comparison

Assume a 350-worker machine shop in an engineering cluster with 40 machines on two shifts: 25 older CNCs with mixed controls and 15 conventional machines. Its ERP handles billing and purchase. It has 12 IP CCTV cameras that stream RTSP, enough to see all 40 machines. Every figure below is a labelled estimate for comparison, not a quote.

The two columns don't buy the same thing. The MES column buys work-order tracking, material consumption, scrap by order and traceability, none of which the camera column delivers. The camera column buys a ranked, costed list of where machine hours were lost. That list is the input that shows whether a ₹50-lakh-plus system will pay for itself. To put a rupee value on a recovered hour, use your machine hour rate.

When MES should come first

Plainly, in these cases:

• Traceability is a condition of the business. Your OEM customer, export buyer or regulator needs a batch or serial history: which material lot, which machine, which operator's shift, which inspection. Cameras can't produce a genealogy record.
• Your lines are already networked. If most machines have PLCs or CNC data options on Ethernet, the expensive connectivity step is largely done, and an MES can read states directly with more precision than video.
• The ERP is live, trusted and used for production planning. Then the MES has clean orders and routings to work from, and the integration risk the surveys highlight is smaller.
• Process plants where the truth is in tags (temperatures, pressures, flows) and the PLC historian already holds most of it. A camera adds little there.
• Someone owns it. The studies above keep naming management ownership. If a director will personally chair the MES rollout, that removes the most-cited failure factor.

When cameras first makes sense

• Mixed, older fleets where connectivity would be the biggest cost line.
• Labour-intensive floors (assembly, garments, packing) where the losses are waiting, starving and late starts, and there is no controller to read.
• The owner can't yet rank the loss buckets. Buying a system of record before you know what to record often produces a costly MES that records the wrong things.
• You need an answer this quarter, for a capacity decision, a new order or a second-shift question.

A sensible path for many plants is both, in sequence: cameras for a few months to find and size the losses, then an MES scoped around the two or three processes where traceability or order-level control actually pays.

What cameras can't tell you here

A camera doesn't know which work order a machine is on, what material lot is loaded, whether a part passed inspection, or the exact cycle count on an enclosed machine. A powered machine idling with nothing moving can look like it is running. A pillar can hide a station. Downtime reasons are inferred from what is visible, such as an empty station, a waiting crane or an open panel, and some will stay "unknown" until a supervisor confirms them. None of this replaces an MES where an MES is genuinely required.

Where Mama fits

Mama reads the cameras a plant already has (or a few added ones) and sends the owner a short morning note on where machine and line hours leaked yesterday. It watches machines and flows and counts people only in aggregate, with no names and no operator ranking. That makes it a "cameras first" option. Mama can draw a camera plan for the lines an MES would be built around, so the losses are sized before the bigger decision. Leave an email or phone in the form below.

FAQ

Can camera analytics replace an MES? No. Cameras give you machine and flow states over time. An MES gives you work orders, material consumption, scrap by order and traceability. If you need those records, you need an MES or an ERP production module.

Is there a reliable MES failure rate for Indian SMEs? We found none. The studies we could trace identify the causes (management ownership, investment, developer capability, user participation, ERP integration) rather than a failure percentage. The widely repeated "60–70% fail" figure for ERP traces back to a web page and a blog post.

Will the camera data feed an MES later? Usually it can. Machine run, stop and wait states with timestamps can be exported and mapped to downtime codes.

Are the cost figures above quotes? No. They are labelled estimates built from one published cloud-MES price, one vendor's connectivity range and our own camera retrofit cost guide. Indian integrator quotes vary widely with fleet age, control types and scope.